chore(release): v0.6.0-nightly.20251225.9f65bd3b

2026-01-06 17:09:14 +00:00 · 2025-12-25 00:12:49 +00:00
109 changed files with 2496 additions and 5788 deletions
--- a/.github/workflows/release-sdk.yml
+++ b/.github/workflows/release-sdk.yml
@@ -91,8 +91,6 @@ jobs:
        with:
          node-version-file: '.nvmrc'
          cache: 'npm'
-          registry-url: 'https://registry.npmjs.org'
-          scope: '@qwen-code'

      - name: 'Install Dependencies'
        run: |-
@@ -128,14 +126,6 @@ jobs:
          IS_PREVIEW: '${{ steps.vars.outputs.is_preview }}'
          MANUAL_VERSION: '${{ inputs.version }}'

-      - name: 'Set SDK package version (local only)'
-        env:
-          RELEASE_VERSION: '${{ steps.version.outputs.RELEASE_VERSION }}'
-        run: |-
-          # Ensure the package version matches the computed release version.
-          # This is required for nightly/preview because npm does not allow re-publishing the same version.
-          npm version -w @qwen-code/sdk "${RELEASE_VERSION}" --no-git-tag-version --allow-same-version
-
      - name: 'Build CLI Bundle'
        run: |
          npm run build
@@ -168,21 +158,7 @@ jobs:
          git config user.name "github-actions[bot]"
          git config user.email "github-actions[bot]@users.noreply.github.com"

-      - name: 'Build SDK'
-        working-directory: 'packages/sdk-typescript'
-        run: |-
-          npm run build
-
-      - name: 'Publish @qwen-code/sdk'
-        working-directory: 'packages/sdk-typescript'
-        run: |-
-          npm publish --access public --tag=${{ steps.version.outputs.NPM_TAG }} ${{ steps.vars.outputs.is_dry_run == 'true' && '--dry-run' || '' }}
-        env:
-          NODE_AUTH_TOKEN: '${{ secrets.NPM_TOKEN }}'
-
      - name: 'Create and switch to a release branch'
-        if: |-
-          ${{ steps.vars.outputs.is_dry_run == 'false' && steps.vars.outputs.is_nightly == 'false' && steps.vars.outputs.is_preview == 'false' }}
        id: 'release_branch'
        env:
          RELEASE_TAG: '${{ steps.version.outputs.RELEASE_TAG }}'
@@ -191,22 +167,50 @@ jobs:
          git switch -c "${BRANCH_NAME}"
          echo "BRANCH_NAME=${BRANCH_NAME}" >> "${GITHUB_OUTPUT}"

-      - name: 'Commit and Push package version (stable only)'
-        if: |-
-          ${{ steps.vars.outputs.is_dry_run == 'false' && steps.vars.outputs.is_nightly == 'false' && steps.vars.outputs.is_preview == 'false' }}
+      - name: 'Update package version'
+        env:
+          RELEASE_VERSION: '${{ steps.version.outputs.RELEASE_VERSION }}'
+        run: |-
+          # Use npm workspaces so the root lockfile is updated consistently.
+          npm version -w @qwen-code/sdk "${RELEASE_VERSION}" --no-git-tag-version --allow-same-version
+
+      - name: 'Commit and Conditionally Push package version'
        env:
          BRANCH_NAME: '${{ steps.release_branch.outputs.BRANCH_NAME }}'
+          IS_DRY_RUN: '${{ steps.vars.outputs.is_dry_run }}'
          RELEASE_TAG: '${{ steps.version.outputs.RELEASE_TAG }}'
        run: |-
-          # Only persist version bumps after a successful publish.
          git add packages/sdk-typescript/package.json package-lock.json
          if git diff --staged --quiet; then
            echo "No version changes to commit"
          else
            git commit -m "chore(release): sdk-typescript ${RELEASE_TAG}"
          fi
-          echo "Pushing release branch to remote..."
-          git push --set-upstream origin "${BRANCH_NAME}" --follow-tags
+          if [[ "${IS_DRY_RUN}" == "false" ]]; then
+            echo "Pushing release branch to remote..."
+            git push --set-upstream origin "${BRANCH_NAME}" --follow-tags
+          else
+            echo "Dry run enabled. Skipping push."
+          fi
+
+      - name: 'Build SDK'
+        working-directory: 'packages/sdk-typescript'
+        run: |-
+          npm run build
+
+      - name: 'Configure npm for publishing'
+        uses: 'actions/setup-node@49933ea5288caeca8642d1e84afbd3f7d6820020' # ratchet:actions/setup-node@v4
+        with:
+          node-version-file: '.nvmrc'
+          registry-url: 'https://registry.npmjs.org'
+          scope: '@qwen-code'
+
+      - name: 'Publish @qwen-code/sdk'
+        working-directory: 'packages/sdk-typescript'
+        run: |-
+          npm publish --access public --tag=${{ steps.version.outputs.NPM_TAG }} ${{ steps.vars.outputs.is_dry_run == 'true' && '--dry-run' || '' }}
+        env:
+          NODE_AUTH_TOKEN: '${{ secrets.NPM_TOKEN }}'

      - name: 'Create GitHub Release and Tag'
        if: |-
@@ -216,29 +220,16 @@ jobs:
          RELEASE_BRANCH: '${{ steps.release_branch.outputs.BRANCH_NAME }}'
          RELEASE_TAG: '${{ steps.version.outputs.RELEASE_TAG }}'
          PREVIOUS_RELEASE_TAG: '${{ steps.version.outputs.PREVIOUS_RELEASE_TAG }}'
-          IS_NIGHTLY: '${{ steps.vars.outputs.is_nightly }}'
-          IS_PREVIEW: '${{ steps.vars.outputs.is_preview }}'
-          REF: '${{ github.event.inputs.ref || github.sha }}'
        run: |-
-          # For stable releases, use the release branch; for nightly/preview, use the current ref
-          if [[ "${IS_NIGHTLY}" == "true" || "${IS_PREVIEW}" == "true" ]]; then
-            TARGET="${REF}"
-            PRERELEASE_FLAG="--prerelease"
-          else
-            TARGET="${RELEASE_BRANCH}"
-            PRERELEASE_FLAG=""
-          fi
-
          gh release create "sdk-typescript-${RELEASE_TAG}" \
-            --target "${TARGET}" \
+            --target "$RELEASE_BRANCH" \
            --title "SDK TypeScript Release ${RELEASE_TAG}" \
            --notes-start-tag "sdk-typescript-${PREVIOUS_RELEASE_TAG}" \
-            --generate-notes \
-            ${PRERELEASE_FLAG}
+            --generate-notes

      - name: 'Create PR to merge release branch into main'
        if: |-
-          ${{ steps.vars.outputs.is_dry_run == 'false' && steps.vars.outputs.is_nightly == 'false' && steps.vars.outputs.is_preview == 'false' }}
+          ${{ steps.vars.outputs.is_dry_run == 'false' }}
        id: 'pr'
        env:
          GITHUB_TOKEN: '${{ secrets.GITHUB_TOKEN }}'
@@ -260,7 +251,7 @@ jobs:

      - name: 'Wait for CI checks to complete'
        if: |-
-          ${{ steps.vars.outputs.is_dry_run == 'false' && steps.vars.outputs.is_nightly == 'false' && steps.vars.outputs.is_preview == 'false' }}
+          ${{ steps.vars.outputs.is_dry_run == 'false' }}
        env:
          GITHUB_TOKEN: '${{ secrets.GITHUB_TOKEN }}'
          PR_URL: '${{ steps.pr.outputs.PR_URL }}'
@@ -271,7 +262,7 @@ jobs:

      - name: 'Enable auto-merge for release PR'
        if: |-
-          ${{ steps.vars.outputs.is_dry_run == 'false' && steps.vars.outputs.is_nightly == 'false' && steps.vars.outputs.is_preview == 'false' }}
+          ${{ steps.vars.outputs.is_dry_run == 'false' }}
        env:
          GITHUB_TOKEN: '${{ secrets.GITHUB_TOKEN }}'
          PR_URL: '${{ steps.pr.outputs.PR_URL }}'
--- a/docs/users/overview.md
+++ b/docs/users/overview.md
@@ -1,6 +1,4 @@
 # Qwen Code overview
-[![@qwen-code/qwen-code downloads](https://img.shields.io/npm/dw/@qwen-code/qwen-code.svg)](https://npm-compare.com/@qwen-code/qwen-code) 
-[![@qwen-code/qwen-code version](https://img.shields.io/npm/v/@qwen-code/qwen-code.svg)](https://www.npmjs.com/package/@qwen-code/qwen-code)

 > Learn about Qwen Code, Qwen's agentic coding tool that lives in your terminal and helps you turn ideas into code faster than ever before.

@@ -48,7 +46,7 @@ You'll be prompted to log in on first use. That's it! [Continue with Quickstart

 > [!note]
 >
-> **New VS Code Extension (Beta)**: Prefer a graphical interface? Our new **VS Code extension** provides an easy-to-use native IDE experience without requiring terminal familiarity. Simply install from the marketplace and start coding with Qwen Code directly in your sidebar. Download and install the [Qwen Code Companion](https://marketplace.visualstudio.com/items?itemName=qwenlm.qwen-code-vscode-ide-companion) now.
+> **New VS Code Extension (Beta)**: Prefer a graphical interface? Our new **VS Code extension** provides an easy-to-use native IDE experience without requiring terminal familiarity. Simply install from the marketplace and start coding with Qwen Code directly in your sidebar. You can search for **Qwen Code** in the VS Code Marketplace and download it.

 ## What Qwen Code does for you

--- a/integration-tests/file-system.test.ts
+++ b/integration-tests/file-system.test.ts
@@ -5,6 +5,8 @@
 */

 import { describe, it, expect } from 'vitest';
+import { existsSync } from 'node:fs';
+import * as path from 'node:path';
 import { TestRig, printDebugInfo, validateModelOutput } from './test-helper.js';

 describe('file-system', () => {
@@ -200,8 +202,8 @@ describe('file-system', () => {
    const readAttempt = toolLogs.find(
      (log) => log.toolRequest.name === 'read_file',
    );
-    const editAttempt = toolLogs.find(
-      (log) => log.toolRequest.name === 'edit_file',
+    const writeAttempt = toolLogs.find(
+      (log) => log.toolRequest.name === 'write_file',
    );
    const successfulReplace = toolLogs.find(
      (log) => log.toolRequest.name === 'replace' && log.toolRequest.success,
@@ -224,15 +226,15 @@ describe('file-system', () => {

    // CRITICAL: Verify that no matter what the model did, it never successfully
    // wrote or replaced anything.
-    if (editAttempt) {
+    if (writeAttempt) {
      console.error(
-        'A edit_file attempt was made when no file should be written.',
+        'A write_file attempt was made when no file should be written.',
      );
      printDebugInfo(rig, result);
    }
    expect(
-      editAttempt,
-      'edit_file should not have been called',
+      writeAttempt,
+      'write_file should not have been called',
    ).toBeUndefined();

    if (successfulReplace) {
@@ -243,5 +245,12 @@ describe('file-system', () => {
      successfulReplace,
      'A successful replace should not have occurred',
    ).toBeUndefined();
+
+    // Final verification: ensure the file was not created.
+    const filePath = path.join(rig.testDir!, fileName);
+    const fileExists = existsSync(filePath);
+    expect(fileExists, 'The non-existent file should not be created').toBe(
+      false,
+    );
  });
 });
--- a/integration-tests/sdk-typescript/permission-control.test.ts
+++ b/integration-tests/sdk-typescript/permission-control.test.ts
@@ -952,8 +952,7 @@ describe('Permission Control (E2E)', () => {
        TEST_TIMEOUT,
      );

-      // FIXME: This test is flaky and sometimes fails with no tool calls.
-      it.skip(
+      it(
        'should allow read-only tools without restrictions',
        async () => {
          // Create test files for the model to read
--- a/integration-tests/sdk-typescript/system-control.test.ts
+++ b/integration-tests/sdk-typescript/system-control.test.ts
@@ -314,88 +314,4 @@ describe('System Control (E2E)', () => {
      );
    });
  });
-
-  describe('supportedCommands API', () => {
-    it('should return list of supported slash commands', async () => {
-      const sessionId = crypto.randomUUID();
-      const generator = (async function* () {
-        yield {
-          type: 'user',
-          session_id: sessionId,
-          message: { role: 'user', content: 'Hello' },
-          parent_tool_use_id: null,
-        } as SDKUserMessage;
-      })();
-
-      const q = query({
-        prompt: generator,
-        options: {
-          ...SHARED_TEST_OPTIONS,
-          cwd: testDir,
-          model: 'qwen3-max',
-          debug: false,
-        },
-      });
-
-      try {
-        const result = await q.supportedCommands();
-        // Start consuming messages to trigger initialization
-        const messageConsumer = (async () => {
-          try {
-            for await (const _message of q) {
-              // Just consume messages
-            }
-          } catch (error) {
-            // Ignore errors from query being closed
-            if (error instanceof Error && error.message !== 'Query is closed') {
-              throw error;
-            }
-          }
-        })();
-
-        // Verify result structure
-        expect(result).toBeDefined();
-        expect(result).toHaveProperty('commands');
-        expect(Array.isArray(result?.['commands'])).toBe(true);
-
-        const commands = result?.['commands'] as string[];
-
-        // Verify default allowed built-in commands are present
-        expect(commands).toContain('init');
-        expect(commands).toContain('summary');
-        expect(commands).toContain('compress');
-
-        // Verify commands are sorted
-        const sortedCommands = [...commands].sort();
-        expect(commands).toEqual(sortedCommands);
-
-        // Verify all commands are strings
-        commands.forEach((cmd) => {
-          expect(typeof cmd).toBe('string');
-          expect(cmd.length).toBeGreaterThan(0);
-        });
-
-        await q.close();
-        await messageConsumer;
-      } catch (error) {
-        await q.close();
-        throw error;
-      }
-    });
-
-    it('should throw error when supportedCommands is called on closed query', async () => {
-      const q = query({
-        prompt: 'Hello',
-        options: {
-          ...SHARED_TEST_OPTIONS,
-          cwd: testDir,
-          model: 'qwen3-max',
-        },
-      });
-
-      await q.close();
-
-      await expect(q.supportedCommands()).rejects.toThrow('Query is closed');
-    });
-  });
 });
--- a/package-lock.json
+++ b/package-lock.json
@@ -1,12 +1,12 @@
 {
  "name": "@qwen-code/qwen-code",
-  "version": "0.6.0-nightly.20251226.17eb20c1",
+  "version": "0.6.0-nightly.20251225.9f65bd3b",
  "lockfileVersion": 3,
  "requires": true,
  "packages": {
    "": {
      "name": "@qwen-code/qwen-code",
-      "version": "0.6.0-nightly.20251226.17eb20c1",
+      "version": "0.6.0-nightly.20251225.9f65bd3b",
      "workspaces": [
        "packages/*"
      ],
@@ -134,36 +134,6 @@
        "node": ">=6.0.0"
      }
    },
-    "node_modules/@anthropic-ai/sdk": {
-      "version": "0.36.3",
-      "resolved": "https://registry.npmjs.org/@anthropic-ai/sdk/-/sdk-0.36.3.tgz",
-      "integrity": "sha512-+c0mMLxL/17yFZ4P5+U6bTWiCSFZUKJddrv01ud2aFBWnTPLdRncYV76D3q1tqfnL7aCnhRtykFnoCFzvr4U3Q==",
-      "license": "MIT",
-      "dependencies": {
-        "@types/node": "^18.11.18",
-        "@types/node-fetch": "^2.6.4",
-        "abort-controller": "^3.0.0",
-        "agentkeepalive": "^4.2.1",
-        "form-data-encoder": "1.7.2",
-        "formdata-node": "^4.3.2",
-        "node-fetch": "^2.6.7"
-      }
-    },
-    "node_modules/@anthropic-ai/sdk/node_modules/@types/node": {
-      "version": "18.19.130",
-      "resolved": "https://registry.npmjs.org/@types/node/-/node-18.19.130.tgz",
-      "integrity": "sha512-GRaXQx6jGfL8sKfaIDD6OupbIHBr9jv7Jnaml9tB7l4v068PAOXqfcujMMo5PhbIs6ggR1XODELqahT2R8v0fg==",
-      "license": "MIT",
-      "dependencies": {
-        "undici-types": "~5.26.4"
-      }
-    },
-    "node_modules/@anthropic-ai/sdk/node_modules/undici-types": {
-      "version": "5.26.5",
-      "resolved": "https://registry.npmjs.org/undici-types/-/undici-types-5.26.5.tgz",
-      "integrity": "sha512-JlCMO+ehdEIKqlFxk6IfVoAUVmgz7cU7zD/h9XZ0qzeosSHmUJVOzSQvvYSYWXkFXC+IfLKSIffhv0sVZup6pA==",
-      "license": "MIT"
-    },
    "node_modules/@asamuzakjp/css-color": {
      "version": "3.2.0",
      "resolved": "https://registry.npmjs.org/@asamuzakjp/css-color/-/css-color-3.2.0.tgz",
@@ -3852,16 +3822,6 @@
        "undici-types": "~6.21.0"
      }
    },
-    "node_modules/@types/node-fetch": {
-      "version": "2.6.13",
-      "resolved": "https://registry.npmjs.org/@types/node-fetch/-/node-fetch-2.6.13.tgz",
-      "integrity": "sha512-QGpRVpzSaUs30JBSGPjOg4Uveu384erbHBoT1zeONvyCfwQxIkUshLAOqN/k9EjGviPRmWTTe6aH2qySWKTVSw==",
-      "license": "MIT",
-      "dependencies": {
-        "@types/node": "*",
-        "form-data": "^4.0.4"
-      }
-    },
    "node_modules/@types/normalize-package-data": {
      "version": "2.4.4",
      "resolved": "https://registry.npmjs.org/@types/normalize-package-data/-/normalize-package-data-2.4.4.tgz",
@@ -4860,6 +4820,7 @@
      "version": "3.0.0",
      "resolved": "https://registry.npmjs.org/abort-controller/-/abort-controller-3.0.0.tgz",
      "integrity": "sha512-h8lQ8tacZYnR3vNQTgibj+tODHI5/+l06Au2Pcriv/Gmet0eaj4TwWH41sO9wnHDiQsEj19q0drzdWdeAHtweg==",
+      "dev": true,
      "license": "MIT",
      "dependencies": {
        "event-target-shim": "^5.0.0"
@@ -4946,18 +4907,6 @@
        "node": ">= 14"
      }
    },
-    "node_modules/agentkeepalive": {
-      "version": "4.6.0",
-      "resolved": "https://registry.npmjs.org/agentkeepalive/-/agentkeepalive-4.6.0.tgz",
-      "integrity": "sha512-kja8j7PjmncONqaTsB8fQ+wE2mSU2DJ9D4XKoJ5PFWIdRMa6SLSN1ff4mOr4jCbfRSsxR4keIiySJU0N9T5hIQ==",
-      "license": "MIT",
-      "dependencies": {
-        "humanize-ms": "^1.2.1"
-      },
-      "engines": {
-        "node": ">= 8.0.0"
-      }
-    },
    "node_modules/ajv": {
      "version": "6.12.6",
      "resolved": "https://registry.npmjs.org/ajv/-/ajv-6.12.6.tgz",
@@ -5529,6 +5478,7 @@
      "version": "0.4.0",
      "resolved": "https://registry.npmjs.org/asynckit/-/asynckit-0.4.0.tgz",
      "integrity": "sha512-Oei9OH4tRh0YqU3GxhX79dM/mwVgvbZJaSNaRk+bshkj0S5cfHcgYakreBjrHwatXKbz+IoIdYLxrKim2MjW0Q==",
+      "dev": true,
      "license": "MIT"
    },
    "node_modules/atomically": {
@@ -6487,6 +6437,7 @@
      "version": "1.0.8",
      "resolved": "https://registry.npmjs.org/combined-stream/-/combined-stream-1.0.8.tgz",
      "integrity": "sha512-FQN4MRfuJeHf7cBbBMJFXhKSDq+2kAArBlmRBvcvFE5BB1HZKXtSFASDhdlz9zOYwxh8lDdnvmMOe/+5cdoEdg==",
+      "dev": true,
      "license": "MIT",
      "dependencies": {
        "delayed-stream": "~1.0.0"
@@ -7112,6 +7063,7 @@
      "version": "1.0.0",
      "resolved": "https://registry.npmjs.org/delayed-stream/-/delayed-stream-1.0.0.tgz",
      "integrity": "sha512-ZySD7Nf91aLB0RxL4KGrKHBXl7Eds1DAmEdcoVawXnLD7SDhpNgtuII2aAkg7a7QS41jxPSZ17p4VdGnMHk3MQ==",
+      "dev": true,
      "license": "MIT",
      "engines": {
        "node": ">=0.4.0"
@@ -7624,6 +7576,7 @@
      "version": "2.1.0",
      "resolved": "https://registry.npmjs.org/es-set-tostringtag/-/es-set-tostringtag-2.1.0.tgz",
      "integrity": "sha512-j6vWzfrGVfyXxge+O0x5sh6cvxAog0a/4Rdd2K36zCMV5eJ+/+tOAngRO8cODMNWbVRdVlmGZQL2YS3yR8bIUA==",
+      "dev": true,
      "license": "MIT",
      "dependencies": {
        "es-errors": "^1.3.0",
@@ -8153,6 +8106,7 @@
      "version": "5.0.1",
      "resolved": "https://registry.npmjs.org/event-target-shim/-/event-target-shim-5.0.1.tgz",
      "integrity": "sha512-i/2XbnSz/uxRCU6+NdVJgKWDTM427+MqYbkQzD321DuCQJUqOuJKIA0IM2+W2xtYHdKOmZ4dR6fExsd4SXL+WQ==",
+      "dev": true,
      "license": "MIT",
      "engines": {
        "node": ">=6"
@@ -8698,6 +8652,7 @@
      "version": "4.0.4",
      "resolved": "https://registry.npmjs.org/form-data/-/form-data-4.0.4.tgz",
      "integrity": "sha512-KrGhL9Q4zjj0kiUt5OO4Mr/A/jlI2jDYs5eHBpYHPcBEVSiipAvn2Ko2HnPe20rmcuuvMHNdZFp+4IlGTMF0Ow==",
+      "dev": true,
      "license": "MIT",
      "dependencies": {
        "asynckit": "^0.4.0",
@@ -8710,16 +8665,11 @@
        "node": ">= 6"
      }
    },
-    "node_modules/form-data-encoder": {
-      "version": "1.7.2",
-      "resolved": "https://registry.npmjs.org/form-data-encoder/-/form-data-encoder-1.7.2.tgz",
-      "integrity": "sha512-qfqtYan3rxrnCk1VYaA4H+Ms9xdpPqvLZa6xmMgFvhO32x7/3J/ExcTd6qpxM0vH2GdMI+poehyBZvqfMTto8A==",
-      "license": "MIT"
-    },
    "node_modules/form-data/node_modules/mime-types": {
      "version": "2.1.35",
      "resolved": "https://registry.npmjs.org/mime-types/-/mime-types-2.1.35.tgz",
      "integrity": "sha512-ZDY+bPm5zTTF+YpCrAU9nK0UgICYPT0QtT1NZWFv4s++TNkcgVaT0g6+4R2uI4MjQjzysHB1zxuWL50hzaeXiw==",
+      "dev": true,
      "license": "MIT",
      "dependencies": {
        "mime-db": "1.52.0"
@@ -8728,28 +8678,6 @@
        "node": ">= 0.6"
      }
    },
-    "node_modules/formdata-node": {
-      "version": "4.4.1",
-      "resolved": "https://registry.npmjs.org/formdata-node/-/formdata-node-4.4.1.tgz",
-      "integrity": "sha512-0iirZp3uVDjVGt9p49aTaqjk84TrglENEDuqfdlZQ1roC9CWlPk6Avf8EEnZNcAqPonwkG35x4n3ww/1THYAeQ==",
-      "license": "MIT",
-      "dependencies": {
-        "node-domexception": "1.0.0",
-        "web-streams-polyfill": "4.0.0-beta.3"
-      },
-      "engines": {
-        "node": ">= 12.20"
-      }
-    },
-    "node_modules/formdata-node/node_modules/web-streams-polyfill": {
-      "version": "4.0.0-beta.3",
-      "resolved": "https://registry.npmjs.org/web-streams-polyfill/-/web-streams-polyfill-4.0.0-beta.3.tgz",
-      "integrity": "sha512-QW95TCTaHmsYfHDybGMwO5IJIM93I/6vTRk+daHTWFPhwh+C8Cg7j7XyKrwrj8Ib6vYXe0ocYNrmzY4xAAN6ug==",
-      "license": "MIT",
-      "engines": {
-        "node": ">= 14"
-      }
-    },
    "node_modules/formdata-polyfill": {
      "version": "4.0.10",
      "resolved": "https://registry.npmjs.org/formdata-polyfill/-/formdata-polyfill-4.0.10.tgz",
@@ -9334,6 +9262,7 @@
      "version": "1.0.2",
      "resolved": "https://registry.npmjs.org/has-tostringtag/-/has-tostringtag-1.0.2.tgz",
      "integrity": "sha512-NqADB8VjPFLM2V0VvHUewwwsw0ZWBaIdgo+ieHtK3hasLz4qeCRjYcqfB6AQrBggRKppKF8L52/VqdVsO47Dlw==",
+      "dev": true,
      "license": "MIT",
      "dependencies": {
        "has-symbols": "^1.0.3"
@@ -9512,15 +9441,6 @@
        "node": ">=16.17.0"
      }
    },
-    "node_modules/humanize-ms": {
-      "version": "1.2.1",
-      "resolved": "https://registry.npmjs.org/humanize-ms/-/humanize-ms-1.2.1.tgz",
-      "integrity": "sha512-Fl70vYtsAFb/C06PTS9dZBo7ihau+Tu/DNCk/OyHhea07S+aeMWpFFkUaXRa8fI+ScZbEI8dfSxwY7gxZ9SAVQ==",
-      "license": "MIT",
-      "dependencies": {
-        "ms": "^2.0.0"
-      }
-    },
    "node_modules/husky": {
      "version": "9.1.7",
      "resolved": "https://registry.npmjs.org/husky/-/husky-9.1.7.tgz",
@@ -12020,48 +11940,6 @@
        "node": ">=10.5.0"
      }
    },
-    "node_modules/node-fetch": {
-      "version": "2.7.0",
-      "resolved": "https://registry.npmjs.org/node-fetch/-/node-fetch-2.7.0.tgz",
-      "integrity": "sha512-c4FRfUm/dbcWZ7U+1Wq0AwCyFL+3nt2bEw05wfxSz+DWpWsitgmSgYmy2dQdWyKC1694ELPqMs/YzUSNozLt8A==",
-      "license": "MIT",
-      "dependencies": {
-        "whatwg-url": "^5.0.0"
-      },
-      "engines": {
-        "node": "4.x || >=6.0.0"
-      },
-      "peerDependencies": {
-        "encoding": "^0.1.0"
-      },
-      "peerDependenciesMeta": {
-        "encoding": {
-          "optional": true
-        }
-      }
-    },
-    "node_modules/node-fetch/node_modules/tr46": {
-      "version": "0.0.3",
-      "resolved": "https://registry.npmjs.org/tr46/-/tr46-0.0.3.tgz",
-      "integrity": "sha512-N3WMsuqV66lT30CrXNbEjx4GEwlow3v6rr4mCcv6prnfwhS01rkgyFdjPNBYd9br7LpXV1+Emh01fHnq2Gdgrw==",
-      "license": "MIT"
-    },
-    "node_modules/node-fetch/node_modules/webidl-conversions": {
-      "version": "3.0.1",
-      "resolved": "https://registry.npmjs.org/webidl-conversions/-/webidl-conversions-3.0.1.tgz",
-      "integrity": "sha512-2JAn3z8AR6rjK8Sm8orRC0h/bcl/DqL7tRPdGZ4I1CjdF+EaMLmYxBHyXuKL849eucPFhvBoxMsflfOb8kxaeQ==",
-      "license": "BSD-2-Clause"
-    },
-    "node_modules/node-fetch/node_modules/whatwg-url": {
-      "version": "5.0.0",
-      "resolved": "https://registry.npmjs.org/whatwg-url/-/whatwg-url-5.0.0.tgz",
-      "integrity": "sha512-saE57nupxk6v3HY35+jzBwYa0rKSy0XR8JSxZPwgLr7ys0IBzhGviA1/TUGJLmSVqs8pb9AnvICXEuOHLprYTw==",
-      "license": "MIT",
-      "dependencies": {
-        "tr46": "~0.0.3",
-        "webidl-conversions": "^3.0.0"
-      }
-    },
    "node_modules/node-pty": {
      "version": "1.0.0",
      "resolved": "https://registry.npmjs.org/node-pty/-/node-pty-1.0.0.tgz",
@@ -17316,7 +17194,7 @@
    },
    "packages/cli": {
      "name": "@qwen-code/qwen-code",
-      "version": "0.6.0-nightly.20251226.17eb20c1",
+      "version": "0.6.0-nightly.20251225.9f65bd3b",
      "dependencies": {
        "@google/genai": "1.30.0",
        "@iarna/toml": "^2.2.5",
@@ -17953,10 +17831,9 @@
    },
    "packages/core": {
      "name": "@qwen-code/qwen-code-core",
-      "version": "0.6.0-nightly.20251226.17eb20c1",
+      "version": "0.6.0-nightly.20251225.9f65bd3b",
      "hasInstallScript": true,
      "dependencies": {
-        "@anthropic-ai/sdk": "^0.36.1",
        "@google/genai": "1.30.0",
        "@modelcontextprotocol/sdk": "^1.25.1",
        "@opentelemetry/api": "^1.9.0",
@@ -18593,7 +18470,7 @@
    },
    "packages/sdk-typescript": {
      "name": "@qwen-code/sdk",
-      "version": "0.1.0",
+      "version": "0.6.0-nightly.20251225.9f65bd3b",
      "license": "Apache-2.0",
      "dependencies": {
        "@modelcontextprotocol/sdk": "^1.25.1",
@@ -21413,7 +21290,7 @@
    },
    "packages/test-utils": {
      "name": "@qwen-code/qwen-code-test-utils",
-      "version": "0.6.0-nightly.20251226.17eb20c1",
+      "version": "0.6.0-nightly.20251225.9f65bd3b",
      "dev": true,
      "license": "Apache-2.0",
      "devDependencies": {
@@ -21425,7 +21302,7 @@
    },
    "packages/vscode-ide-companion": {
      "name": "qwen-code-vscode-ide-companion",
-      "version": "0.6.0-nightly.20251226.17eb20c1",
+      "version": "0.6.0-nightly.20251225.9f65bd3b",
      "license": "LICENSE",
      "dependencies": {
        "@modelcontextprotocol/sdk": "^1.25.1",
--- a/package.json
+++ b/package.json
@@ -1,6 +1,6 @@
 {
  "name": "@qwen-code/qwen-code",
-  "version": "0.6.0-nightly.20251226.17eb20c1",
+  "version": "0.6.0-nightly.20251225.9f65bd3b",
  "engines": {
    "node": ">=20.0.0"
  },
@@ -13,7 +13,7 @@
    "url": "git+https://github.com/QwenLM/qwen-code.git"
  },
  "config": {
-    "sandboxImageUri": "ghcr.io/qwenlm/qwen-code:0.6.0-nightly.20251226.17eb20c1"
+    "sandboxImageUri": "ghcr.io/qwenlm/qwen-code:0.6.0-nightly.20251225.9f65bd3b"
  },
  "scripts": {
    "start": "cross-env node scripts/start.js",
--- a/packages/cli/package.json
+++ b/packages/cli/package.json
@@ -1,6 +1,6 @@
 {
  "name": "@qwen-code/qwen-code",
-  "version": "0.6.0-nightly.20251226.17eb20c1",
+  "version": "0.6.0-nightly.20251225.9f65bd3b",
  "description": "Qwen Code",
  "repository": {
    "type": "git",
@@ -33,7 +33,7 @@
    "dist"
  ],
  "config": {
-    "sandboxImageUri": "ghcr.io/qwenlm/qwen-code:0.6.0-nightly.20251226.17eb20c1"
+    "sandboxImageUri": "ghcr.io/qwenlm/qwen-code:0.6.0-nightly.20251225.9f65bd3b"
  },
  "dependencies": {
    "@google/genai": "1.30.0",
--- a/packages/cli/src/acp-integration/acp.ts
+++ b/packages/cli/src/acp-integration/acp.ts
@@ -98,14 +98,6 @@ export class AgentSideConnection implements Client {
    );
  }

-  /**
-   * Sends a custom notification to the client.
-   * Used for extension-specific notifications that are not part of the core ACP protocol.
-   */
-  async sendCustomNotification<T>(method: string, params: T): Promise<void> {
-    return await this.#connection.sendNotification(method, params);
-  }
-
  /**
   * Request permission before running a tool
   *
@@ -382,7 +374,6 @@ export interface Client {
  ): Promise<schema.RequestPermissionResponse>;
  sessionUpdate(params: schema.SessionNotification): Promise<void>;
  authenticateUpdate(params: schema.AuthenticateUpdate): Promise<void>;
-  sendCustomNotification<T>(method: string, params: T): Promise<void>;
  writeTextFile(
    params: schema.WriteTextFileRequest,
  ): Promise<schema.WriteTextFileResponse>;
--- a/packages/cli/src/acp-integration/acpAgent.ts
+++ b/packages/cli/src/acp-integration/acpAgent.ts
@@ -15,10 +15,10 @@ import {
  qwenOAuth2Events,
  MCPServerConfig,
  SessionService,
+  buildApiHistoryFromConversation,
  type Config,
  type ConversationRecord,
  type DeviceAuthorizationData,
-  tokenLimit,
 } from '@qwen-code/qwen-code-core';
 import type { ApprovalModeValue } from './schema.js';
 import * as acp from './acp.js';
@@ -165,30 +165,9 @@ class GeminiAgent {
    this.setupFileSystem(config);

    const session = await this.createAndStoreSession(config);
-    const configuredModel = (
-      config.getModel() ||
-      this.config.getModel() ||
-      ''
-    ).trim();
-    const modelId = configuredModel || 'default';
-    const modelName = configuredModel || modelId;

    return {
      sessionId: session.getId(),
-      models: {
-        currentModelId: modelId,
-        availableModels: [
-          {
-            modelId,
-            name: modelName,
-            description: null,
-            _meta: {
-              contextLimit: tokenLimit(modelId),
-            },
-          },
-        ],
-        _meta: null,
-      },
    };
  }

@@ -348,20 +327,12 @@ class GeminiAgent {
    const sessionId = config.getSessionId();
    const geminiClient = config.getGeminiClient();

-    // Use GeminiClient to manage chat lifecycle properly
-    // This ensures geminiClient.chat is in sync with the session's chat
-    //
-    // Note: When loading a session, config.initialize() has already been called
-    // in newSessionConfig(), which in turn calls geminiClient.initialize().
-    // The GeminiClient.initialize() method checks config.getResumedSessionData()
-    // and automatically loads the conversation history into the chat instance.
-    // So we only need to initialize if it hasn't been done yet.
-    if (!geminiClient.isInitialized()) {
-      await geminiClient.initialize();
-    }
-
-    // Now get the chat instance that's managed by GeminiClient
-    const chat = geminiClient.getChat();
+    const history = conversation
+      ? buildApiHistoryFromConversation(conversation)
+      : undefined;
+    const chat = history
+      ? await geminiClient.startChat(history)
+      : await geminiClient.startChat();

    const session = new Session(
      sessionId,
--- a/packages/cli/src/acp-integration/schema.ts
+++ b/packages/cli/src/acp-integration/schema.ts
@@ -93,7 +93,6 @@ export type ModeInfo = z.infer<typeof modeInfoSchema>;
 export type ModesData = z.infer<typeof modesDataSchema>;

 export type AgentInfo = z.infer<typeof agentInfoSchema>;
-export type ModelInfo = z.infer<typeof modelInfoSchema>;

 export type PromptCapabilities = z.infer<typeof promptCapabilitiesSchema>;

@@ -255,26 +254,8 @@ export const authenticateUpdateSchema = z.object({

 export type AuthenticateUpdate = z.infer<typeof authenticateUpdateSchema>;

-export const acpMetaSchema = z.record(z.unknown()).nullable().optional();
-
-export const modelIdSchema = z.string();
-
-export const modelInfoSchema = z.object({
-  _meta: acpMetaSchema,
-  description: z.string().nullable().optional(),
-  modelId: modelIdSchema,
-  name: z.string(),
-});
-
-export const sessionModelStateSchema = z.object({
-  _meta: acpMetaSchema,
-  availableModels: z.array(modelInfoSchema),
-  currentModelId: modelIdSchema,
-});
-
 export const newSessionResponseSchema = z.object({
  sessionId: z.string(),
-  models: sessionModelStateSchema,
 });

 export const loadSessionResponseSchema = z.null();
@@ -533,13 +514,6 @@ export const currentModeUpdateSchema = z.object({

 export type CurrentModeUpdate = z.infer<typeof currentModeUpdateSchema>;

-export const currentModelUpdateSchema = z.object({
-  sessionUpdate: z.literal('current_model_update'),
-  model: modelInfoSchema,
-});
-
-export type CurrentModelUpdate = z.infer<typeof currentModelUpdateSchema>;
-
 export const sessionUpdateSchema = z.union([
  z.object({
    content: contentBlockSchema,
@@ -581,7 +555,6 @@ export const sessionUpdateSchema = z.union([
    sessionUpdate: z.literal('plan'),
  }),
  currentModeUpdateSchema,
-  currentModelUpdateSchema,
  availableCommandsUpdateSchema,
 ]);

--- a/packages/cli/src/acp-integration/session/Session.ts
+++ b/packages/cli/src/acp-integration/session/Session.ts
@@ -41,11 +41,9 @@ import * as fs from 'node:fs/promises';
 import * as path from 'node:path';
 import { z } from 'zod';
 import { getErrorMessage } from '../../utils/errors.js';
-import { normalizePartList } from '../../utils/nonInteractiveHelpers.js';
 import {
  handleSlashCommand,
  getAvailableCommands,
-  type NonInteractiveSlashCommandResult,
 } from '../../nonInteractiveCliCommands.js';
 import type {
  AvailableCommand,
@@ -65,6 +63,12 @@ import { PlanEmitter } from './emitters/PlanEmitter.js';
 import { MessageEmitter } from './emitters/MessageEmitter.js';
 import { SubAgentTracker } from './SubAgentTracker.js';

+/**
+ * Built-in commands that are allowed in ACP integration mode.
+ * Only safe, read-only commands that don't require interactive UI.
+ */
+export const ALLOWED_BUILTIN_COMMANDS_FOR_ACP = ['init'];
+
 /**
 * Session represents an active conversation session with the AI model.
 * It uses modular components for consistent event emission:
@@ -163,26 +167,24 @@ export class Session implements SessionContext {
    const firstTextBlock = params.prompt.find((block) => block.type === 'text');
    const inputText = firstTextBlock?.text || '';

-    let parts: Part[] | null;
+    let parts: Part[];

    if (isSlashCommand(inputText)) {
-      // Handle slash command - uses default allowed commands (init, summary, compress)
+      // Handle slash command - allow specific built-in commands for ACP integration
      const slashCommandResult = await handleSlashCommand(
        inputText,
        pendingSend,
        this.config,
        this.settings,
+        ALLOWED_BUILTIN_COMMANDS_FOR_ACP,
      );

-      parts = await this.#processSlashCommandResult(
-        slashCommandResult,
-        params.prompt,
-      );
-
-      // If parts is null, the command was fully handled (e.g., /summary completed)
-      // Return early without sending to the model
-      if (parts === null) {
-        return { stopReason: 'end_turn' };
+      if (slashCommandResult) {
+        // Use the result from the slash command
+        parts = slashCommandResult as Part[];
+      } else {
+        // Slash command didn't return a prompt, continue with normal processing
+        parts = await this.#resolvePrompt(params.prompt, pendingSend.signal);
      }
    } else {
      // Normal processing for non-slash commands
@@ -293,10 +295,11 @@ export class Session implements SessionContext {
  async sendAvailableCommandsUpdate(): Promise<void> {
    const abortController = new AbortController();
    try {
-      // Use default allowed commands from getAvailableCommands
      const slashCommands = await getAvailableCommands(
        this.config,
+        this.settings,
        abortController.signal,
+        ALLOWED_BUILTIN_COMMANDS_FOR_ACP,
      );

      // Convert SlashCommand[] to AvailableCommand[] format for ACP protocol
@@ -644,103 +647,6 @@ export class Session implements SessionContext {
    }
  }

-  /**
-   * Processes the result of a slash command execution.
-   *
-   * Supported result types in ACP mode:
-   * - submit_prompt: Submits content to the model
-   * - stream_messages: Streams multiple messages to the client (ACP-specific)
-   * - unsupported: Command cannot be executed in ACP mode
-   * - no_command: No command was found, use original prompt
-   *
-   * Note: 'message' type is not supported in ACP mode - commands should use
-   * 'stream_messages' instead for consistent async handling.
-   *
-   * @param result The result from handleSlashCommand
-   * @param originalPrompt The original prompt blocks
-   * @returns Parts to use for the prompt, or null if command was handled without needing model interaction
-   */
-  async #processSlashCommandResult(
-    result: NonInteractiveSlashCommandResult,
-    originalPrompt: acp.ContentBlock[],
-  ): Promise<Part[] | null> {
-    switch (result.type) {
-      case 'submit_prompt':
-        // Command wants to submit a prompt to the model
-        // Convert PartListUnion to Part[]
-        return normalizePartList(result.content);
-
-      case 'message': {
-        // 'message' type is not ideal for ACP mode, but we handle it for compatibility
-        // by converting it to a stream_messages-like notification
-        await this.client.sendCustomNotification('_qwencode/slash_command', {
-          sessionId: this.sessionId,
-          command: originalPrompt
-            .filter((block) => block.type === 'text')
-            .map((block) => (block.type === 'text' ? block.text : ''))
-            .join(' '),
-          messageType: result.messageType,
-          message: result.content || '',
-        });
-
-        if (result.messageType === 'error') {
-          // Throw error to stop execution
-          throw new Error(result.content || 'Slash command failed.');
-        }
-        // For info messages, return null to indicate command was handled
-        return null;
-      }
-
-      case 'stream_messages': {
-        // Command returns multiple messages via async generator (ACP-preferred)
-        const command = originalPrompt
-          .filter((block) => block.type === 'text')
-          .map((block) => (block.type === 'text' ? block.text : ''))
-          .join(' ');
-
-        // Stream all messages to the client
-        for await (const msg of result.messages) {
-          await this.client.sendCustomNotification('_qwencode/slash_command', {
-            sessionId: this.sessionId,
-            command,
-            messageType: msg.messageType,
-            message: msg.content,
-          });
-
-          // If we encounter an error message, throw after sending
-          if (msg.messageType === 'error') {
-            throw new Error(msg.content || 'Slash command failed.');
-          }
-        }
-
-        // All messages sent successfully, return null to indicate command was handled
-        return null;
-      }
-
-      case 'unsupported': {
-        // Command returned an unsupported result type
-        const unsupportedError = `Slash command not supported in ACP integration: ${result.reason}`;
-        throw new Error(unsupportedError);
-      }
-
-      case 'no_command':
-        // No command was found or executed, use original prompt
-        return originalPrompt.map((block) => {
-          if (block.type === 'text') {
-            return { text: block.text };
-          }
-          throw new Error(`Unsupported block type: ${block.type}`);
-        });
-
-      default: {
-        // Exhaustiveness check
-        const _exhaustive: never = result;
-        const unknownError = `Unknown slash command result type: ${(_exhaustive as NonInteractiveSlashCommandResult).type}`;
-        throw new Error(unknownError);
-      }
-    }
-  }
-
  async #resolvePrompt(
    message: acp.ContentBlock[],
    abortSignal: AbortSignal,
--- a/packages/cli/src/config/auth.ts
+++ b/packages/cli/src/config/auth.ts
@@ -26,20 +26,6 @@ export function validateAuthMethod(authMethod: string): string | null {
    return null;
  }

-  if (authMethod === AuthType.USE_ANTHROPIC) {
-    const hasApiKey = process.env['ANTHROPIC_API_KEY'];
-    if (!hasApiKey) {
-      return 'ANTHROPIC_API_KEY environment variable not found.';
-    }
-
-    const hasBaseUrl = process.env['ANTHROPIC_BASE_URL'];
-    if (!hasBaseUrl) {
-      return 'ANTHROPIC_BASE_URL environment variable not found.';
-    }
-
-    return null;
-  }
-
  if (authMethod === AuthType.USE_GEMINI) {
    const hasApiKey = process.env['GEMINI_API_KEY'];
    if (!hasApiKey) {
--- a/packages/cli/src/config/config.test.ts
+++ b/packages/cli/src/config/config.test.ts
@@ -2114,14 +2114,7 @@ describe('loadCliConfig model selection', () => {
  });

  it('always prefers model from argvs', async () => {
-    process.argv = [
-      'node',
-      'script.js',
-      '--auth-type',
-      'openai',
-      '--model',
-      'qwen3-coder-plus',
-    ];
+    process.argv = ['node', 'script.js', '--model', 'qwen3-coder-plus'];
    const argv = await parseArguments({} as Settings);
    const config = await loadCliConfig(
      {
@@ -2141,14 +2134,7 @@ describe('loadCliConfig model selection', () => {
  });

  it('selects the model from argvs if provided', async () => {
-    process.argv = [
-      'node',
-      'script.js',
-      '--auth-type',
-      'openai',
-      '--model',
-      'qwen3-coder-plus',
-    ];
+    process.argv = ['node', 'script.js', '--model', 'qwen3-coder-plus'];
    const argv = await parseArguments({} as Settings);
    const config = await loadCliConfig(
      {
--- a/packages/cli/src/config/config.ts
+++ b/packages/cli/src/config/config.ts
@@ -468,7 +468,6 @@ export async function parseArguments(settings: Settings): Promise<CliArgs> {
          type: 'string',
          choices: [
            AuthType.USE_OPENAI,
-            AuthType.USE_ANTHROPIC,
            AuthType.QWEN_OAUTH,
            AuthType.USE_GEMINI,
            AuthType.USE_VERTEX_AI,
@@ -877,30 +876,11 @@ export async function loadCliConfig(
    );
  }

-  const selectedAuthType =
-    (argv.authType as AuthType | undefined) ||
-    settings.security?.auth?.selectedType;
-
-  const apiKey =
-    (selectedAuthType === AuthType.USE_OPENAI
-      ? argv.openaiApiKey ||
-        process.env['OPENAI_API_KEY'] ||
-        settings.security?.auth?.apiKey
-      : '') || '';
-  const baseUrl =
-    (selectedAuthType === AuthType.USE_OPENAI
-      ? argv.openaiBaseUrl ||
-        process.env['OPENAI_BASE_URL'] ||
-        settings.security?.auth?.baseUrl
-      : '') || '';
  const resolvedModel =
    argv.model ||
-    (selectedAuthType === AuthType.USE_OPENAI
-      ? process.env['OPENAI_MODEL'] ||
-        process.env['QWEN_MODEL'] ||
-        settings.model?.name
-      : '') ||
-    '';
+    process.env['OPENAI_MODEL'] ||
+    process.env['QWEN_MODEL'] ||
+    settings.model?.name;

  const sandboxConfig = await loadSandboxConfig(settings, argv);
  const screenReader =
@@ -987,15 +967,23 @@ export async function loadCliConfig(
    extensions: allExtensions,
    blockedMcpServers,
    noBrowser: !!process.env['NO_BROWSER'],
-    authType: selectedAuthType,
+    authType:
+      (argv.authType as AuthType | undefined) ||
+      settings.security?.auth?.selectedType,
    inputFormat,
    outputFormat,
    includePartialMessages,
    generationConfig: {
      ...(settings.model?.generationConfig || {}),
      model: resolvedModel,
-      apiKey,
-      baseUrl,
+      apiKey:
+        argv.openaiApiKey ||
+        process.env['OPENAI_API_KEY'] ||
+        settings.security?.auth?.apiKey,
+      baseUrl:
+        argv.openaiBaseUrl ||
+        process.env['OPENAI_BASE_URL'] ||
+        settings.security?.auth?.baseUrl,
      enableOpenAILogging:
        (typeof argv.openaiLogging === 'undefined'
          ? settings.model?.enableOpenAILogging
--- a/packages/cli/src/i18n/locales/en.js
+++ b/packages/cli/src/i18n/locales/en.js
@@ -258,8 +258,6 @@ export default {
  ', Tab to change focus': ', Tab to change focus',
  'To see changes, Qwen Code must be restarted. Press r to exit and apply changes now.':
    'To see changes, Qwen Code must be restarted. Press r to exit and apply changes now.',
-  'The command "/{{command}}" is not supported in non-interactive mode.':
-    'The command "/{{command}}" is not supported in non-interactive mode.',
  // ============================================================================
  // Settings Labels
  // ============================================================================
@@ -592,12 +590,6 @@ export default {
  'No conversation found to summarize.': 'No conversation found to summarize.',
  'Failed to generate project context summary: {{error}}':
    'Failed to generate project context summary: {{error}}',
-  'Saved project summary to {{filePathForDisplay}}.':
-    'Saved project summary to {{filePathForDisplay}}.',
-  'Saving project summary...': 'Saving project summary...',
-  'Generating project summary...': 'Generating project summary...',
-  'Failed to generate summary - no text content received from LLM response':
-    'Failed to generate summary - no text content received from LLM response',

  // ============================================================================
  // Commands - Model
--- a/packages/cli/src/i18n/locales/ru.js
+++ b/packages/cli/src/i18n/locales/ru.js
@@ -260,8 +260,7 @@ export default {
  ', Tab to change focus': ', Tab для смены фокуса',
  'To see changes, Qwen Code must be restarted. Press r to exit and apply changes now.':
    'Для применения изменений необходимо перезапустить Qwen Code. Нажмите r для выхода и применения изменений.',
-  'The command "/{{command}}" is not supported in non-interactive mode.':
-    'Команда "/{{command}}" не поддерживается в неинтерактивном режиме.',
+
  // ============================================================================
  // Метки настроек
  // ============================================================================
@@ -605,12 +604,6 @@ export default {
    'Не найдено диалогов для создания сводки.',
  'Failed to generate project context summary: {{error}}':
    'Не удалось сгенерировать сводку контекста проекта: {{error}}',
-  'Saved project summary to {{filePathForDisplay}}.':
-    'Сводка проекта сохранена в {{filePathForDisplay}}',
-  'Saving project summary...': 'Сохранение сводки проекта...',
-  'Generating project summary...': 'Генерация сводки проекта...',
-  'Failed to generate summary - no text content received from LLM response':
-    'Не удалось сгенерировать сводку - не получен текстовый контент из ответа LLM',

  // ============================================================================
  // Команды - Модель
--- a/packages/cli/src/i18n/locales/zh.js
+++ b/packages/cli/src/i18n/locales/zh.js
@@ -249,8 +249,6 @@ export default {
  ', Tab to change focus': '，Tab 切换焦点',
  'To see changes, Qwen Code must be restarted. Press r to exit and apply changes now.':
    '要查看更改，必须重启 Qwen Code。按 r 退出并立即应用更改。',
-  'The command "/{{command}}" is not supported in non-interactive mode.':
-    '不支持在非交互模式下使用命令 "/{{command}}"。',
  // ============================================================================
  // Settings Labels
  // ============================================================================
@@ -562,12 +560,6 @@ export default {
  'No conversation found to summarize.': '未找到要总结的对话',
  'Failed to generate project context summary: {{error}}':
    '生成项目上下文摘要失败：{{error}}',
-  'Saved project summary to {{filePathForDisplay}}.':
-    '项目摘要已保存到 {{filePathForDisplay}}',
-  'Saving project summary...': '正在保存项目摘要...',
-  'Generating project summary...': '正在生成项目摘要...',
-  'Failed to generate summary - no text content received from LLM response':
-    '生成摘要失败 - 未从 LLM 响应中接收到文本内容',

  // ============================================================================
  // Commands - Model
--- a/packages/cli/src/nonInteractive/control/controllers/systemController.ts
+++ b/packages/cli/src/nonInteractive/control/controllers/systemController.ts
@@ -20,7 +20,8 @@ import type {
  CLIControlSetModelRequest,
  CLIMcpServerConfig,
 } from '../../types.js';
-import { getAvailableCommands } from '../../../nonInteractiveCliCommands.js';
+import { CommandService } from '../../../services/CommandService.js';
+import { BuiltinCommandLoader } from '../../../services/BuiltinCommandLoader.js';
 import {
  MCPServerConfig,
  AuthProviderType,
@@ -406,7 +407,7 @@ export class SystemController extends BaseController {
  }

  /**
-   * Load slash command names using getAvailableCommands
+   * Load slash command names using CommandService
   *
   * @param signal - AbortSignal to respect for cancellation
   * @returns Promise resolving to array of slash command names
@@ -417,14 +418,21 @@ export class SystemController extends BaseController {
    }

    try {
-      const commands = await getAvailableCommands(this.context.config, signal);
+      const service = await CommandService.create(
+        [new BuiltinCommandLoader(this.context.config)],
+        signal,
+      );

      if (signal.aborted) {
        return [];
      }

-      // Extract command names and sort
-      return commands.map((cmd) => cmd.name).sort();
+      const names = new Set<string>();
+      const commands = service.getCommands();
+      for (const command of commands) {
+        names.add(command.name);
+      }
+      return Array.from(names).sort();
    } catch (error) {
      // Check if the error is due to abort
      if (signal.aborted) {
--- a/packages/cli/src/nonInteractiveCli.test.ts
+++ b/packages/cli/src/nonInteractiveCli.test.ts
@@ -68,7 +68,6 @@ describe('runNonInteractive', () => {
  let mockShutdownTelemetry: Mock;
  let consoleErrorSpy: MockInstance;
  let processStdoutSpy: MockInstance;
-  let processStderrSpy: MockInstance;
  let mockGeminiClient: {
    sendMessageStream: Mock;
    getChatRecordingService: Mock;
@@ -87,9 +86,6 @@ describe('runNonInteractive', () => {
    processStdoutSpy = vi
      .spyOn(process.stdout, 'write')
      .mockImplementation(() => true);
-    processStderrSpy = vi
-      .spyOn(process.stderr, 'write')
-      .mockImplementation(() => true);
    vi.spyOn(process, 'exit').mockImplementation((code) => {
      throw new Error(`process.exit(${code}) called`);
    });
@@ -143,8 +139,6 @@ describe('runNonInteractive', () => {
      setModel: vi.fn(async (model: string) => {
        currentModel = model;
      }),
-      getExperimentalZedIntegration: vi.fn().mockReturnValue(false),
-      isInteractive: vi.fn().mockReturnValue(false),
    } as unknown as Config;

    mockSettings = {
@@ -858,7 +852,7 @@ describe('runNonInteractive', () => {
    expect(processStdoutSpy).toHaveBeenCalledWith('Response from command');
  });

-  it('should handle command that requires confirmation by returning early', async () => {
+  it('should throw FatalInputError if a command requires confirmation', async () => {
    const mockCommand = {
      name: 'confirm',
      description: 'a command that needs confirmation',
@@ -870,16 +864,15 @@ describe('runNonInteractive', () => {
    };
    mockGetCommands.mockReturnValue([mockCommand]);

-    await runNonInteractive(
-      mockConfig,
-      mockSettings,
-      '/confirm',
-      'prompt-id-confirm',
-    );
-
-    // Should write error message to stderr
-    expect(processStderrSpy).toHaveBeenCalledWith(
-      'Shell command confirmation is not supported in non-interactive mode. Use YOLO mode or pre-approve commands.\n',
+    await expect(
+      runNonInteractive(
+        mockConfig,
+        mockSettings,
+        '/confirm',
+        'prompt-id-confirm',
+      ),
+    ).rejects.toThrow(
+      'Exiting due to a confirmation prompt requested by the command.',
    );
  });

@@ -916,30 +909,7 @@ describe('runNonInteractive', () => {
    expect(processStdoutSpy).toHaveBeenCalledWith('Response to unknown');
  });

-  it('should handle known but unsupported slash commands like /help by returning early', async () => {
-    // Mock a built-in command that exists but is not in the allowed list
-    const mockHelpCommand = {
-      name: 'help',
-      description: 'Show help',
-      kind: CommandKind.BUILT_IN,
-      action: vi.fn(),
-    };
-    mockGetCommands.mockReturnValue([mockHelpCommand]);
-
-    await runNonInteractive(
-      mockConfig,
-      mockSettings,
-      '/help',
-      'prompt-id-help',
-    );
-
-    // Should write error message to stderr
-    expect(processStderrSpy).toHaveBeenCalledWith(
-      'The command "/help" is not supported in non-interactive mode.\n',
-    );
-  });
-
-  it('should handle unhandled command result types by returning early with error', async () => {
+  it('should throw for unhandled command result types', async () => {
    const mockCommand = {
      name: 'noaction',
      description: 'unhandled type',
@@ -950,16 +920,15 @@ describe('runNonInteractive', () => {
    };
    mockGetCommands.mockReturnValue([mockCommand]);

-    await runNonInteractive(
-      mockConfig,
-      mockSettings,
-      '/noaction',
-      'prompt-id-unhandled',
-    );
-
-    // Should write error message to stderr
-    expect(processStderrSpy).toHaveBeenCalledWith(
-      'Unknown command result type: unhandled\n',
+    await expect(
+      runNonInteractive(
+        mockConfig,
+        mockSettings,
+        '/noaction',
+        'prompt-id-unhandled',
+      ),
+    ).rejects.toThrow(
+      'Exiting due to command result that is not supported in non-interactive mode.',
    );
  });

--- a/packages/cli/src/nonInteractiveCli.ts
+++ b/packages/cli/src/nonInteractiveCli.ts
@@ -42,55 +42,6 @@ import {
  computeUsageFromMetrics,
 } from './utils/nonInteractiveHelpers.js';

-/**
- * Emits a final message for slash command results.
- * Note: systemMessage should already be emitted before calling this function.
- */
-async function emitNonInteractiveFinalMessage(params: {
-  message: string;
-  isError: boolean;
-  adapter?: JsonOutputAdapterInterface;
-  config: Config;
-  startTimeMs: number;
-}): Promise<void> {
-  const { message, isError, adapter, config } = params;
-
-  if (!adapter) {
-    // Text output mode: write directly to stdout/stderr
-    const target = isError ? process.stderr : process.stdout;
-    target.write(`${message}\n`);
-    return;
-  }
-
-  // JSON output mode: emit assistant message and result
-  // (systemMessage should already be emitted by caller)
-  adapter.startAssistantMessage();
-  adapter.processEvent({
-    type: GeminiEventType.Content,
-    value: message,
-  } as unknown as Parameters<JsonOutputAdapterInterface['processEvent']>[0]);
-  adapter.finalizeAssistantMessage();
-
-  const metrics = uiTelemetryService.getMetrics();
-  const usage = computeUsageFromMetrics(metrics);
-  const outputFormat = config.getOutputFormat();
-  const stats =
-    outputFormat === OutputFormat.JSON
-      ? uiTelemetryService.getMetrics()
-      : undefined;
-
-  adapter.emitResult({
-    isError,
-    durationMs: Date.now() - params.startTimeMs,
-    apiDurationMs: 0,
-    numTurns: 0,
-    errorMessage: isError ? message : undefined,
-    usage,
-    stats,
-    summary: message,
-  });
-}
-
 /**
 * Provides optional overrides for `runNonInteractive` execution.
 *
@@ -164,16 +115,6 @@ export async function runNonInteractive(
      process.on('SIGINT', shutdownHandler);
      process.on('SIGTERM', shutdownHandler);

-      // Emit systemMessage first (always the first message in JSON mode)
-      if (adapter) {
-        const systemMessage = await buildSystemMessage(
-          config,
-          sessionId,
-          permissionMode,
-        );
-        adapter.emitMessage(systemMessage);
-      }
-
      let initialPartList: PartListUnion | null = extractPartsFromUserMessage(
        options.userMessage,
      );
@@ -187,45 +128,10 @@ export async function runNonInteractive(
            config,
            settings,
          );
-          switch (slashCommandResult.type) {
-            case 'submit_prompt':
-              // A slash command can replace the prompt entirely; fall back to @-command processing otherwise.
-              initialPartList = slashCommandResult.content;
-              slashHandled = true;
-              break;
-            case 'message': {
-              // systemMessage already emitted above
-              await emitNonInteractiveFinalMessage({
-                message: slashCommandResult.content,
-                isError: slashCommandResult.messageType === 'error',
-                adapter,
-                config,
-                startTimeMs: startTime,
-              });
-              return;
-            }
-            case 'stream_messages':
-              throw new FatalInputError(
-                'Stream messages mode is not supported in non-interactive CLI',
-              );
-            case 'unsupported': {
-              await emitNonInteractiveFinalMessage({
-                message: slashCommandResult.reason,
-                isError: true,
-                adapter,
-                config,
-                startTimeMs: startTime,
-              });
-              return;
-            }
-            case 'no_command':
-              break;
-            default: {
-              const _exhaustive: never = slashCommandResult;
-              throw new FatalInputError(
-                `Unhandled slash command result type: ${(_exhaustive as { type: string }).type}`,
-              );
-            }
+          if (slashCommandResult) {
+            // A slash command can replace the prompt entirely; fall back to @-command processing otherwise.
+            initialPartList = slashCommandResult as PartListUnion;
+            slashHandled = true;
          }
        }

@@ -257,6 +163,15 @@ export async function runNonInteractive(
      const initialParts = normalizePartList(initialPartList);
      let currentMessages: Content[] = [{ role: 'user', parts: initialParts }];

+      if (adapter) {
+        const systemMessage = await buildSystemMessage(
+          config,
+          sessionId,
+          permissionMode,
+        );
+        adapter.emitMessage(systemMessage);
+      }
+
      let isFirstTurn = true;
      while (true) {
        turnCount++;
--- a/packages/cli/src/nonInteractiveCliCommands.test.ts
+++ b/packages/cli/src/nonInteractiveCliCommands.test.ts
@@ -1,242 +0,0 @@
-/**
- * @license
- * Copyright 2025 Qwen Team
- * SPDX-License-Identifier: Apache-2.0
- */
-
-import { describe, it, expect, vi, beforeEach } from 'vitest';
-import { handleSlashCommand } from './nonInteractiveCliCommands.js';
-import type { Config } from '@qwen-code/qwen-code-core';
-import type { LoadedSettings } from './config/settings.js';
-import { CommandKind } from './ui/commands/types.js';
-
-// Mock the CommandService
-const mockGetCommands = vi.hoisted(() => vi.fn());
-const mockCommandServiceCreate = vi.hoisted(() => vi.fn());
-vi.mock('./services/CommandService.js', () => ({
-  CommandService: {
-    create: mockCommandServiceCreate,
-  },
-}));
-
-describe('handleSlashCommand', () => {
-  let mockConfig: Config;
-  let mockSettings: LoadedSettings;
-  let abortController: AbortController;
-
-  beforeEach(() => {
-    mockCommandServiceCreate.mockResolvedValue({
-      getCommands: mockGetCommands,
-    });
-
-    mockConfig = {
-      getExperimentalZedIntegration: vi.fn().mockReturnValue(false),
-      isInteractive: vi.fn().mockReturnValue(false),
-      getSessionId: vi.fn().mockReturnValue('test-session'),
-      getFolderTrustFeature: vi.fn().mockReturnValue(false),
-      getFolderTrust: vi.fn().mockReturnValue(false),
-      getProjectRoot: vi.fn().mockReturnValue('/test/project'),
-      storage: {},
-    } as unknown as Config;
-
-    mockSettings = {
-      system: { path: '', settings: {} },
-      systemDefaults: { path: '', settings: {} },
-      user: { path: '', settings: {} },
-      workspace: { path: '', settings: {} },
-    } as LoadedSettings;
-
-    abortController = new AbortController();
-  });
-
-  it('should return no_command for non-slash input', async () => {
-    const result = await handleSlashCommand(
-      'regular text',
-      abortController,
-      mockConfig,
-      mockSettings,
-    );
-
-    expect(result.type).toBe('no_command');
-  });
-
-  it('should return no_command for unknown slash commands', async () => {
-    mockGetCommands.mockReturnValue([]);
-
-    const result = await handleSlashCommand(
-      '/unknowncommand',
-      abortController,
-      mockConfig,
-      mockSettings,
-    );
-
-    expect(result.type).toBe('no_command');
-  });
-
-  it('should return unsupported for known built-in commands not in allowed list', async () => {
-    const mockHelpCommand = {
-      name: 'help',
-      description: 'Show help',
-      kind: CommandKind.BUILT_IN,
-      action: vi.fn(),
-    };
-    mockGetCommands.mockReturnValue([mockHelpCommand]);
-
-    const result = await handleSlashCommand(
-      '/help',
-      abortController,
-      mockConfig,
-      mockSettings,
-      [], // Empty allowed list
-    );
-
-    expect(result.type).toBe('unsupported');
-    if (result.type === 'unsupported') {
-      expect(result.reason).toContain('/help');
-      expect(result.reason).toContain('not supported');
-    }
-  });
-
-  it('should return unsupported for /help when using default allowed list', async () => {
-    const mockHelpCommand = {
-      name: 'help',
-      description: 'Show help',
-      kind: CommandKind.BUILT_IN,
-      action: vi.fn(),
-    };
-    mockGetCommands.mockReturnValue([mockHelpCommand]);
-
-    const result = await handleSlashCommand(
-      '/help',
-      abortController,
-      mockConfig,
-      mockSettings,
-      // Default allowed list: ['init', 'summary', 'compress']
-    );
-
-    expect(result.type).toBe('unsupported');
-    if (result.type === 'unsupported') {
-      expect(result.reason).toBe(
-        'The command "/help" is not supported in non-interactive mode.',
-      );
-    }
-  });
-
-  it('should execute allowed built-in commands', async () => {
-    const mockInitCommand = {
-      name: 'init',
-      description: 'Initialize project',
-      kind: CommandKind.BUILT_IN,
-      action: vi.fn().mockResolvedValue({
-        type: 'message',
-        messageType: 'info',
-        content: 'Project initialized',
-      }),
-    };
-    mockGetCommands.mockReturnValue([mockInitCommand]);
-
-    const result = await handleSlashCommand(
-      '/init',
-      abortController,
-      mockConfig,
-      mockSettings,
-      ['init'], // init is in the allowed list
-    );
-
-    expect(result.type).toBe('message');
-    if (result.type === 'message') {
-      expect(result.content).toBe('Project initialized');
-    }
-  });
-
-  it('should execute file commands regardless of allowed list', async () => {
-    const mockFileCommand = {
-      name: 'custom',
-      description: 'Custom file command',
-      kind: CommandKind.FILE,
-      action: vi.fn().mockResolvedValue({
-        type: 'submit_prompt',
-        content: [{ text: 'Custom prompt' }],
-      }),
-    };
-    mockGetCommands.mockReturnValue([mockFileCommand]);
-
-    const result = await handleSlashCommand(
-      '/custom',
-      abortController,
-      mockConfig,
-      mockSettings,
-      [], // Empty allowed list, but FILE commands should still work
-    );
-
-    expect(result.type).toBe('submit_prompt');
-    if (result.type === 'submit_prompt') {
-      expect(result.content).toEqual([{ text: 'Custom prompt' }]);
-    }
-  });
-
-  it('should return unsupported for other built-in commands like /quit', async () => {
-    const mockQuitCommand = {
-      name: 'quit',
-      description: 'Quit application',
-      kind: CommandKind.BUILT_IN,
-      action: vi.fn(),
-    };
-    mockGetCommands.mockReturnValue([mockQuitCommand]);
-
-    const result = await handleSlashCommand(
-      '/quit',
-      abortController,
-      mockConfig,
-      mockSettings,
-    );
-
-    expect(result.type).toBe('unsupported');
-    if (result.type === 'unsupported') {
-      expect(result.reason).toContain('/quit');
-      expect(result.reason).toContain('not supported');
-    }
-  });
-
-  it('should handle command with no action', async () => {
-    const mockCommand = {
-      name: 'noaction',
-      description: 'Command without action',
-      kind: CommandKind.FILE,
-      // No action property
-    };
-    mockGetCommands.mockReturnValue([mockCommand]);
-
-    const result = await handleSlashCommand(
-      '/noaction',
-      abortController,
-      mockConfig,
-      mockSettings,
-    );
-
-    expect(result.type).toBe('no_command');
-  });
-
-  it('should return message when command returns void', async () => {
-    const mockCommand = {
-      name: 'voidcmd',
-      description: 'Command that returns void',
-      kind: CommandKind.FILE,
-      action: vi.fn().mockResolvedValue(undefined),
-    };
-    mockGetCommands.mockReturnValue([mockCommand]);
-
-    const result = await handleSlashCommand(
-      '/voidcmd',
-      abortController,
-      mockConfig,
-      mockSettings,
-    );
-
-    expect(result.type).toBe('message');
-    if (result.type === 'message') {
-      expect(result.content).toBe('Command executed successfully.');
-      expect(result.messageType).toBe('info');
-    }
-  });
-});
--- a/packages/cli/src/nonInteractiveCliCommands.ts
+++ b/packages/cli/src/nonInteractiveCliCommands.ts
@@ -7,6 +7,7 @@
 import type { PartListUnion } from '@google/genai';
 import { parseSlashCommand } from './utils/commands.js';
 import {
+  FatalInputError,
  Logger,
  uiTelemetryService,
  type Config,
@@ -18,164 +19,10 @@ import {
  CommandKind,
  type CommandContext,
  type SlashCommand,
-  type SlashCommandActionReturn,
 } from './ui/commands/types.js';
 import { createNonInteractiveUI } from './ui/noninteractive/nonInteractiveUi.js';
 import type { LoadedSettings } from './config/settings.js';
 import type { SessionStatsState } from './ui/contexts/SessionContext.js';
-import { t } from './i18n/index.js';
-
-/**
- * Built-in commands that are allowed in non-interactive modes (CLI and ACP).
- * Only safe, read-only commands that don't require interactive UI.
- *
- * These commands are:
- * - init: Initialize project configuration
- * - summary: Generate session summary
- * - compress: Compress conversation history
- */
-export const ALLOWED_BUILTIN_COMMANDS_NON_INTERACTIVE = [
-  'init',
-  'summary',
-  'compress',
-] as const;
-
-/**
- * Result of handling a slash command in non-interactive mode.
- *
- * Supported types:
- * - 'submit_prompt': Submits content to the model (supports all modes)
- * - 'message': Returns a single message (supports non-interactive JSON/text only)
- * - 'stream_messages': Streams multiple messages (supports ACP only)
- * - 'unsupported': Command cannot be executed in this mode
- * - 'no_command': No command was found or executed
- */
-export type NonInteractiveSlashCommandResult =
-  | {
-      type: 'submit_prompt';
-      content: PartListUnion;
-    }
-  | {
-      type: 'message';
-      messageType: 'info' | 'error';
-      content: string;
-    }
-  | {
-      type: 'stream_messages';
-      messages: AsyncGenerator<
-        { messageType: 'info' | 'error'; content: string },
-        void,
-        unknown
-      >;
-    }
-  | {
-      type: 'unsupported';
-      reason: string;
-      originalType: string;
-    }
-  | {
-      type: 'no_command';
-    };
-
-/**
- * Converts a SlashCommandActionReturn to a NonInteractiveSlashCommandResult.
- *
- * Only the following result types are supported in non-interactive mode:
- * - submit_prompt: Submits content to the model (all modes)
- * - message: Returns a single message (non-interactive JSON/text only)
- * - stream_messages: Streams multiple messages (ACP only)
- *
- * All other result types are converted to 'unsupported'.
- *
- * @param result The result from executing a slash command action
- * @returns A NonInteractiveSlashCommandResult describing the outcome
- */
-function handleCommandResult(
-  result: SlashCommandActionReturn,
-): NonInteractiveSlashCommandResult {
-  switch (result.type) {
-    case 'submit_prompt':
-      return {
-        type: 'submit_prompt',
-        content: result.content,
-      };
-
-    case 'message':
-      return {
-        type: 'message',
-        messageType: result.messageType,
-        content: result.content,
-      };
-
-    case 'stream_messages':
-      return {
-        type: 'stream_messages',
-        messages: result.messages,
-      };
-
-    /**
-     * Currently return types below are never generated due to the
-     * whitelist of allowed slash commands in ACP and non-interactive mode.
-     * We'll try to add more supported return types in the future.
-     */
-    case 'tool':
-      return {
-        type: 'unsupported',
-        reason:
-          'Tool execution from slash commands is not supported in non-interactive mode.',
-        originalType: 'tool',
-      };
-
-    case 'quit':
-      return {
-        type: 'unsupported',
-        reason:
-          'Quit command is not supported in non-interactive mode. The process will exit naturally after completion.',
-        originalType: 'quit',
-      };
-
-    case 'dialog':
-      return {
-        type: 'unsupported',
-        reason: `Dialog '${result.dialog}' cannot be opened in non-interactive mode.`,
-        originalType: 'dialog',
-      };
-
-    case 'load_history':
-      return {
-        type: 'unsupported',
-        reason:
-          'Loading history is not supported in non-interactive mode. Each invocation starts with a fresh context.',
-        originalType: 'load_history',
-      };
-
-    case 'confirm_shell_commands':
-      return {
-        type: 'unsupported',
-        reason:
-          'Shell command confirmation is not supported in non-interactive mode. Use YOLO mode or pre-approve commands.',
-        originalType: 'confirm_shell_commands',
-      };
-
-    case 'confirm_action':
-      return {
-        type: 'unsupported',
-        reason:
-          'Action confirmation is not supported in non-interactive mode. Commands requiring confirmation cannot be executed.',
-        originalType: 'confirm_action',
-      };
-
-    default: {
-      // Exhaustiveness check
-      const _exhaustive: never = result;
-      return {
-        type: 'unsupported',
-        reason: `Unknown command result type: ${(_exhaustive as SlashCommandActionReturn).type}`,
-        originalType: 'unknown',
-      };
-    }
-  }
-}

 /**
 * Filters commands based on the allowed built-in command names.
@@ -215,146 +62,122 @@ function filterCommandsForNonInteractive(
 * @param config The configuration object
 * @param settings The loaded settings
 * @param allowedBuiltinCommandNames Optional array of built-in command names that are
- *   allowed. Defaults to ALLOWED_BUILTIN_COMMANDS_NON_INTERACTIVE (init, summary, compress).
- *   Pass an empty array to only allow file commands.
- * @returns A Promise that resolves to a `NonInteractiveSlashCommandResult` describing
- *   the outcome of the command execution.
+ *   allowed. If not provided or empty, only file commands are available.
+ * @returns A Promise that resolves to `PartListUnion` if a valid command is
+ *   found and results in a prompt, or `undefined` otherwise.
+ * @throws {FatalInputError} if the command result is not supported in
+ *   non-interactive mode.
 */
 export const handleSlashCommand = async (
  rawQuery: string,
  abortController: AbortController,
  config: Config,
  settings: LoadedSettings,
-  allowedBuiltinCommandNames: string[] = [
-    ...ALLOWED_BUILTIN_COMMANDS_NON_INTERACTIVE,
-  ],
-): Promise<NonInteractiveSlashCommandResult> => {
+  allowedBuiltinCommandNames?: string[],
+): Promise<PartListUnion | undefined> => {
  const trimmed = rawQuery.trim();
  if (!trimmed.startsWith('/')) {
-    return { type: 'no_command' };
+    return;
  }

-  const isAcpMode = config.getExperimentalZedIntegration();
-  const isInteractive = config.isInteractive();
-
-  const executionMode = isAcpMode
-    ? 'acp'
-    : isInteractive
-      ? 'interactive'
-      : 'non_interactive';
-
  const allowedBuiltinSet = new Set(allowedBuiltinCommandNames ?? []);

-  // Load all commands to check if the command exists but is not allowed
-  const allLoaders = [
-    new BuiltinCommandLoader(config),
-    new FileCommandLoader(config),
-  ];
+  // Only load BuiltinCommandLoader if there are allowed built-in commands
+  const loaders =
+    allowedBuiltinSet.size > 0
+      ? [new BuiltinCommandLoader(config), new FileCommandLoader(config)]
+      : [new FileCommandLoader(config)];

  const commandService = await CommandService.create(
-    allLoaders,
+    loaders,
    abortController.signal,
  );
-  const allCommands = commandService.getCommands();
+  const commands = commandService.getCommands();
  const filteredCommands = filterCommandsForNonInteractive(
-    allCommands,
+    commands,
    allowedBuiltinSet,
  );

-  // First, try to parse with filtered commands
  const { commandToExecute, args } = parseSlashCommand(
    rawQuery,
    filteredCommands,
  );

-  if (!commandToExecute) {
-    // Check if this is a known command that's just not allowed
-    const { commandToExecute: knownCommand } = parseSlashCommand(
-      rawQuery,
-      allCommands,
-    );
-
-    if (knownCommand) {
-      // Command exists but is not allowed in non-interactive mode
-      return {
-        type: 'unsupported',
-        reason: t(
-          'The command "/{{command}}" is not supported in non-interactive mode.',
-          { command: knownCommand.name },
-        ),
-        originalType: 'filtered_command',
+  if (commandToExecute) {
+    if (commandToExecute.action) {
+      // Not used by custom commands but may be in the future.
+      const sessionStats: SessionStatsState = {
+        sessionId: config?.getSessionId(),
+        sessionStartTime: new Date(),
+        metrics: uiTelemetryService.getMetrics(),
+        lastPromptTokenCount: 0,
+        promptCount: 1,
      };
+
+      const logger = new Logger(config?.getSessionId() || '', config?.storage);
+
+      const context: CommandContext = {
+        services: {
+          config,
+          settings,
+          git: undefined,
+          logger,
+        },
+        ui: createNonInteractiveUI(),
+        session: {
+          stats: sessionStats,
+          sessionShellAllowlist: new Set(),
+        },
+        invocation: {
+          raw: trimmed,
+          name: commandToExecute.name,
+          args,
+        },
+      };
+
+      const result = await commandToExecute.action(context, args);
+
+      if (result) {
+        switch (result.type) {
+          case 'submit_prompt':
+            return result.content;
+          case 'confirm_shell_commands':
+            // This result indicates a command attempted to confirm shell commands.
+            // However note that currently, ShellTool is excluded in non-interactive
+            // mode unless 'YOLO mode' is active, so confirmation actually won't
+            // occur because of YOLO mode.
+            // This ensures that if a command *does* request confirmation (e.g.
+            // in the future with more granular permissions), it's handled appropriately.
+            throw new FatalInputError(
+              'Exiting due to a confirmation prompt requested by the command.',
+            );
+          default:
+            throw new FatalInputError(
+              'Exiting due to command result that is not supported in non-interactive mode.',
+            );
+        }
+      }
    }
-
-    return { type: 'no_command' };
  }

-  if (!commandToExecute.action) {
-    return { type: 'no_command' };
-  }
-
-  // Not used by custom commands but may be in the future.
-  const sessionStats: SessionStatsState = {
-    sessionId: config?.getSessionId(),
-    sessionStartTime: new Date(),
-    metrics: uiTelemetryService.getMetrics(),
-    lastPromptTokenCount: 0,
-    promptCount: 1,
-  };
-
-  const logger = new Logger(config?.getSessionId() || '', config?.storage);
-
-  const context: CommandContext = {
-    executionMode,
-    services: {
-      config,
-      settings,
-      git: undefined,
-      logger,
-    },
-    ui: createNonInteractiveUI(),
-    session: {
-      stats: sessionStats,
-      sessionShellAllowlist: new Set(),
-    },
-    invocation: {
-      raw: trimmed,
-      name: commandToExecute.name,
-      args,
-    },
-  };
-
-  const result = await commandToExecute.action(context, args);
-
-  if (!result) {
-    // Command executed but returned no result (e.g., void return)
-    return {
-      type: 'message',
-      messageType: 'info',
-      content: 'Command executed successfully.',
-    };
-  }
-
-  // Handle different result types
-  return handleCommandResult(result);
+  return;
 };

 /**
 * Retrieves all available slash commands for the current configuration.
 *
 * @param config The configuration object
+ * @param settings The loaded settings
 * @param abortSignal Signal to cancel the loading process
 * @param allowedBuiltinCommandNames Optional array of built-in command names that are
- *   allowed. Defaults to ALLOWED_BUILTIN_COMMANDS_NON_INTERACTIVE (init, summary, compress).
- *   Pass an empty array to only include file commands.
+ *   allowed. If not provided or empty, only file commands are available.
 * @returns A Promise that resolves to an array of SlashCommand objects
 */
 export const getAvailableCommands = async (
  config: Config,
+  settings: LoadedSettings,
  abortSignal: AbortSignal,
-  allowedBuiltinCommandNames: string[] = [
-    ...ALLOWED_BUILTIN_COMMANDS_NON_INTERACTIVE,
-  ],
+  allowedBuiltinCommandNames?: string[],
 ): Promise<SlashCommand[]> => {
  try {
    const allowedBuiltinSet = new Set(allowedBuiltinCommandNames ?? []);
--- a/packages/cli/src/ui/auth/useAuth.ts
+++ b/packages/cli/src/ui/auth/useAuth.ts
@@ -228,7 +228,6 @@ export const useAuthCommand = (
      ![
        AuthType.QWEN_OAUTH,
        AuthType.USE_OPENAI,
-        AuthType.USE_ANTHROPIC,
        AuthType.USE_GEMINI,
        AuthType.USE_VERTEX_AI,
      ].includes(defaultAuthType as AuthType)
@@ -241,7 +240,6 @@ export const useAuthCommand = (
            validValues: [
              AuthType.QWEN_OAUTH,
              AuthType.USE_OPENAI,
-              AuthType.USE_ANTHROPIC,
              AuthType.USE_GEMINI,
              AuthType.USE_VERTEX_AI,
            ].join(', '),
--- a/packages/cli/src/ui/commands/compressCommand.ts
+++ b/packages/cli/src/ui/commands/compressCommand.ts
@@ -19,9 +19,7 @@ export const compressCommand: SlashCommand = {
  kind: CommandKind.BUILT_IN,
  action: async (context) => {
    const { ui } = context;
-    const executionMode = context.executionMode ?? 'interactive';
-
-    if (executionMode === 'interactive' && ui.pendingItem) {
+    if (ui.pendingItem) {
      ui.addItem(
        {
          type: MessageType.ERROR,
@@ -42,80 +40,13 @@ export const compressCommand: SlashCommand = {
      },
    };

-    const config = context.services.config;
-    const geminiClient = config?.getGeminiClient();
-    if (!config || !geminiClient) {
-      return {
-        type: 'message',
-        messageType: 'error',
-        content: t('Config not loaded.'),
-      };
-    }
-
-    const doCompress = async () => {
-      const promptId = `compress-${Date.now()}`;
-      return await geminiClient.tryCompressChat(promptId, true);
-    };
-
-    if (executionMode === 'acp') {
-      const messages = async function* () {
-        try {
-          yield {
-            messageType: 'info' as const,
-            content: 'Compressing context...',
-          };
-          const compressed = await doCompress();
-          if (!compressed) {
-            yield {
-              messageType: 'error' as const,
-              content: t('Failed to compress chat history.'),
-            };
-            return;
-          }
-          yield {
-            messageType: 'info' as const,
-            content: `Context compressed (${compressed.originalTokenCount} -> ${compressed.newTokenCount}).`,
-          };
-        } catch (e) {
-          yield {
-            messageType: 'error' as const,
-            content: t('Failed to compress chat history: {{error}}', {
-              error: e instanceof Error ? e.message : String(e),
-            }),
-          };
-        }
-      };
-
-      return { type: 'stream_messages', messages: messages() };
-    }
-
    try {
-      if (executionMode === 'interactive') {
-        ui.setPendingItem(pendingMessage);
-      }
-
-      const compressed = await doCompress();
-
-      if (!compressed) {
-        if (executionMode === 'interactive') {
-          ui.addItem(
-            {
-              type: MessageType.ERROR,
-              text: t('Failed to compress chat history.'),
-            },
-            Date.now(),
-          );
-          return;
-        }
-
-        return {
-          type: 'message',
-          messageType: 'error',
-          content: t('Failed to compress chat history.'),
-        };
-      }
-
-      if (executionMode === 'interactive') {
+      ui.setPendingItem(pendingMessage);
+      const promptId = `compress-${Date.now()}`;
+      const compressed = await context.services.config
+        ?.getGeminiClient()
+        ?.tryCompressChat(promptId, true);
+      if (compressed) {
        ui.addItem(
          {
            type: MessageType.COMPRESSION,
@@ -128,39 +59,27 @@ export const compressCommand: SlashCommand = {
          } as HistoryItemCompression,
          Date.now(),
        );
-        return;
-      }
-
-      return {
-        type: 'message',
-        messageType: 'info',
-        content: `Context compressed (${compressed.originalTokenCount} -> ${compressed.newTokenCount}).`,
-      };
-    } catch (e) {
-      if (executionMode === 'interactive') {
+      } else {
        ui.addItem(
          {
            type: MessageType.ERROR,
-            text: t('Failed to compress chat history: {{error}}', {
-              error: e instanceof Error ? e.message : String(e),
-            }),
+            text: t('Failed to compress chat history.'),
          },
          Date.now(),
        );
-        return;
      }
-
-      return {
-        type: 'message',
-        messageType: 'error',
-        content: t('Failed to compress chat history: {{error}}', {
-          error: e instanceof Error ? e.message : String(e),
-        }),
-      };
+    } catch (e) {
+      ui.addItem(
+        {
+          type: MessageType.ERROR,
+          text: t('Failed to compress chat history: {{error}}', {
+            error: e instanceof Error ? e.message : String(e),
+          }),
+        },
+        Date.now(),
+      );
    } finally {
-      if (executionMode === 'interactive') {
-        ui.setPendingItem(null);
-      }
+      ui.setPendingItem(null);
    }
  },
 };
--- a/packages/cli/src/ui/commands/summaryCommand.ts
+++ b/packages/cli/src/ui/commands/summaryCommand.ts
@@ -26,8 +26,6 @@ export const summaryCommand: SlashCommand = {
  action: async (context): Promise<SlashCommandActionReturn> => {
    const { config } = context.services;
    const { ui } = context;
-    const executionMode = context.executionMode ?? 'interactive';
-
    if (!config) {
      return {
        type: 'message',
@@ -45,8 +43,8 @@ export const summaryCommand: SlashCommand = {
      };
    }

-    // Check if already generating summary (interactive UI only)
-    if (executionMode === 'interactive' && ui.pendingItem) {
+    // Check if already generating summary
+    if (ui.pendingItem) {
      ui.addItem(
        {
          type: 'error' as const,
@@ -65,22 +63,29 @@ export const summaryCommand: SlashCommand = {
      };
    }

-    const getChatHistory = () => {
+    try {
+      // Get the current chat history
      const chat = geminiClient.getChat();
-      return chat.getHistory();
-    };
+      const history = chat.getHistory();

-    const validateChatHistory = (
-      history: ReturnType<typeof getChatHistory>,
-    ) => {
      if (history.length <= 2) {
-        throw new Error(t('No conversation found to summarize.'));
+        return {
+          type: 'message',
+          messageType: 'info',
+          content: t('No conversation found to summarize.'),
+        };
      }
-    };

-    const generateSummaryMarkdown = async (
-      history: ReturnType<typeof getChatHistory>,
-    ): Promise<string> => {
+      // Show loading state
+      const pendingMessage: HistoryItemSummary = {
+        type: 'summary',
+        summary: {
+          isPending: true,
+          stage: 'generating',
+        },
+      };
+      ui.setPendingItem(pendingMessage);
+
      // Build the conversation context for summary generation
      const conversationContext = history.map((message) => ({
        role: message.role,
@@ -116,21 +121,19 @@ export const summaryCommand: SlashCommand = {

      if (!markdownSummary) {
        throw new Error(
-          t(
-            'Failed to generate summary - no text content received from LLM response',
-          ),
+          'Failed to generate summary - no text content received from LLM response',
        );
      }

-      return markdownSummary;
-    };
+      // Update loading message to show saving progress
+      ui.setPendingItem({
+        type: 'summary',
+        summary: {
+          isPending: true,
+          stage: 'saving',
+        },
+      });

-    const saveSummaryToDisk = async (
-      markdownSummary: string,
-    ): Promise<{
-      filePathForDisplay: string;
-      fullPath: string;
-    }> => {
      // Ensure .qwen directory exists
      const projectRoot = config.getProjectRoot();
      const qwenDir = path.join(projectRoot, '.qwen');
@@ -152,163 +155,45 @@ export const summaryCommand: SlashCommand = {

      await fsPromises.writeFile(summaryPath, summaryContent, 'utf8');

-      return {
-        filePathForDisplay: '.qwen/PROJECT_SUMMARY.md',
-        fullPath: summaryPath,
-      };
-    };
-
-    const emitInteractivePending = (stage: 'generating' | 'saving') => {
-      if (executionMode !== 'interactive') {
-        return;
-      }
-      const pendingMessage: HistoryItemSummary = {
-        type: 'summary',
-        summary: {
-          isPending: true,
-          stage,
-        },
-      };
-      ui.setPendingItem(pendingMessage);
-    };
-
-    const completeInteractive = (filePathForDisplay: string) => {
-      if (executionMode !== 'interactive') {
-        return;
-      }
+      // Clear pending item and show success message
      ui.setPendingItem(null);
      const completedSummaryItem: HistoryItemSummary = {
        type: 'summary',
        summary: {
          isPending: false,
          stage: 'completed',
-          filePath: filePathForDisplay,
+          filePath: '.qwen/PROJECT_SUMMARY.md',
        },
      };
      ui.addItem(completedSummaryItem, Date.now());
-    };

-    const formatErrorMessage = (error: unknown): string =>
-      t('Failed to generate project context summary: {{error}}', {
-        error: error instanceof Error ? error.message : String(error),
-      });
-
-    const failInteractive = (error: unknown) => {
-      if (executionMode !== 'interactive') {
-        return;
-      }
+      return {
+        type: 'message',
+        messageType: 'info',
+        content: '', // Empty content since we show the message in UI component
+      };
+    } catch (error) {
+      // Clear pending item on error
      ui.setPendingItem(null);
      ui.addItem(
        {
          type: 'error' as const,
-          text: `❌ ${formatErrorMessage(error)}`,
+          text: `❌ ${t(
+            'Failed to generate project context summary: {{error}}',
+            {
+              error: error instanceof Error ? error.message : String(error),
+            },
+          )}`,
        },
        Date.now(),
      );
-    };
-
-    const formatSuccessMessage = (filePathForDisplay: string): string =>
-      t('Saved project summary to {{filePathForDisplay}}.', {
-        filePathForDisplay,
-      });
-
-    const returnNoConversationMessage = (): SlashCommandActionReturn => {
-      const msg = t('No conversation found to summarize.');
-      if (executionMode === 'acp') {
-        const messages = async function* () {
-          yield {
-            messageType: 'info' as const,
-            content: msg,
-          };
-        };
-        return {
-          type: 'stream_messages',
-          messages: messages(),
-        };
-      }
-      return {
-        type: 'message',
-        messageType: 'info',
-        content: msg,
-      };
-    };
-
-    const executeSummaryGeneration = async (
-      history: ReturnType<typeof getChatHistory>,
-    ): Promise<{
-      markdownSummary: string;
-      filePathForDisplay: string;
-    }> => {
-      emitInteractivePending('generating');
-      const markdownSummary = await generateSummaryMarkdown(history);
-      emitInteractivePending('saving');
-      const { filePathForDisplay } = await saveSummaryToDisk(markdownSummary);
-      completeInteractive(filePathForDisplay);
-      return { markdownSummary, filePathForDisplay };
-    };
-
-    // Validate chat history once at the beginning
-    const history = getChatHistory();
-    try {
-      validateChatHistory(history);
-    } catch (_error) {
-      return returnNoConversationMessage();
-    }
-
-    if (executionMode === 'acp') {
-      const messages = async function* () {
-        try {
-          yield {
-            messageType: 'info' as const,
-            content: t('Generating project summary...'),
-          };
-
-          const { filePathForDisplay } =
-            await executeSummaryGeneration(history);
-
-          yield {
-            messageType: 'info' as const,
-            content: formatSuccessMessage(filePathForDisplay),
-          };
-        } catch (error) {
-          failInteractive(error);
-          yield {
-            messageType: 'error' as const,
-            content: formatErrorMessage(error),
-          };
-        }
-      };
-
-      return {
-        type: 'stream_messages',
-        messages: messages(),
-      };
-    }
-
-    try {
-      const { filePathForDisplay } = await executeSummaryGeneration(history);
-
-      if (executionMode === 'non_interactive') {
-        return {
-          type: 'message',
-          messageType: 'info',
-          content: formatSuccessMessage(filePathForDisplay),
-        };
-      }
-
-      // Interactive mode: UI components already display progress and completion.
-      return {
-        type: 'message',
-        messageType: 'info',
-        content: '',
-      };
-    } catch (error) {
-      failInteractive(error);

      return {
        type: 'message',
        messageType: 'error',
-        content: formatErrorMessage(error),
+        content: t('Failed to generate project context summary: {{error}}', {
+          error: error instanceof Error ? error.message : String(error),
+        }),
      };
    }
  },
--- a/packages/cli/src/ui/commands/types.ts
+++ b/packages/cli/src/ui/commands/types.ts
@@ -22,14 +22,6 @@ import type {

 // Grouped dependencies for clarity and easier mocking
 export interface CommandContext {
-  /**
-   * Execution mode for the current invocation.
-   *
-   * - interactive: React/Ink UI mode
-   * - non_interactive: non-interactive CLI mode (text/json)
-   * - acp: ACP/Zed integration mode
-   */
-  executionMode?: 'interactive' | 'non_interactive' | 'acp';
  // Invocation properties for when commands are called.
  invocation?: {
    /** The raw, untrimmed input string from the user. */
@@ -116,19 +108,6 @@ export interface MessageActionReturn {
  content: string;
 }

-/**
- * The return type for a command action that streams multiple messages.
- * Used for long-running operations that need to send progress updates.
- */
-export interface StreamMessagesActionReturn {
-  type: 'stream_messages';
-  messages: AsyncGenerator<
-    { messageType: 'info' | 'error'; content: string },
-    void,
-    unknown
-  >;
-}
-
 /**
 * The return type for a command action that needs to open a dialog.
 */
@@ -195,7 +174,6 @@ export interface ConfirmActionReturn {
 export type SlashCommandActionReturn =
  | ToolActionReturn
  | MessageActionReturn
-  | StreamMessagesActionReturn
  | QuitActionReturn
  | OpenDialogActionReturn
  | LoadHistoryActionReturn
--- a/packages/cli/src/ui/hooks/slashCommandProcessor.ts
+++ b/packages/cli/src/ui/hooks/slashCommandProcessor.ts
@@ -520,13 +520,6 @@ export const useSlashCommandProcessor = (
                    true,
                  );
                }
-                case 'stream_messages': {
-                  // stream_messages is only used in ACP/Zed integration mode
-                  // and should not be returned in interactive UI mode
-                  throw new Error(
-                    'stream_messages result type is not supported in interactive mode',
-                  );
-                }
                default: {
                  const unhandled: never = result;
                  throw new Error(
--- a/packages/cli/src/ui/hooks/useGeminiStream.ts
+++ b/packages/cli/src/ui/hooks/useGeminiStream.ts
@@ -526,15 +526,10 @@ export const useGeminiStream = (
        return currentThoughtBuffer;
      }

-      let newThoughtBuffer = currentThoughtBuffer + thoughtText;
-
-      const pendingType = pendingHistoryItemRef.current?.type;
-      const isPendingThought =
-        pendingType === 'gemini_thought' ||
-        pendingType === 'gemini_thought_content';
+      const newThoughtBuffer = currentThoughtBuffer + thoughtText;

      // If we're not already showing a thought, start a new one
-      if (!isPendingThought) {
+      if (pendingHistoryItemRef.current?.type !== 'gemini_thought') {
        // If there's a pending non-thought item, finalize it first
        if (pendingHistoryItemRef.current) {
          addItem(pendingHistoryItemRef.current, userMessageTimestamp);
@@ -542,37 +537,11 @@ export const useGeminiStream = (
        setPendingHistoryItem({ type: 'gemini_thought', text: '' });
      }

-      // Split large thought messages for better rendering performance (same rationale
-      // as regular content streaming). This helps avoid terminal flicker caused by
-      // constantly re-rendering an ever-growing "pending" block.
-      const splitPoint = findLastSafeSplitPoint(newThoughtBuffer);
-      const nextPendingType: 'gemini_thought' | 'gemini_thought_content' =
-        isPendingThought && pendingType === 'gemini_thought_content'
-          ? 'gemini_thought_content'
-          : 'gemini_thought';
-
-      if (splitPoint === newThoughtBuffer.length) {
-        // Update the existing thought message with accumulated content
-        setPendingHistoryItem({
-          type: nextPendingType,
-          text: newThoughtBuffer,
-        });
-      } else {
-        const beforeText = newThoughtBuffer.substring(0, splitPoint);
-        const afterText = newThoughtBuffer.substring(splitPoint);
-        addItem(
-          {
-            type: nextPendingType,
-            text: beforeText,
-          },
-          userMessageTimestamp,
-        );
-        setPendingHistoryItem({
-          type: 'gemini_thought_content',
-          text: afterText,
-        });
-        newThoughtBuffer = afterText;
-      }
+      // Update the existing thought message with accumulated content
+      setPendingHistoryItem({
+        type: 'gemini_thought',
+        text: newThoughtBuffer,
+      });

      // Also update the thought state for the loading indicator
      mergeThought(eventValue);
--- a/packages/cli/src/ui/models/availableModels.ts
+++ b/packages/cli/src/ui/models/availableModels.ts
@@ -60,11 +60,6 @@ export function getOpenAIAvailableModelFromEnv(): AvailableModel | null {
  return id ? { id, label: id } : null;
 }

-export function getAnthropicAvailableModelFromEnv(): AvailableModel | null {
-  const id = process.env['ANTHROPIC_MODEL']?.trim();
-  return id ? { id, label: id } : null;
-}
-
 export function getAvailableModelsForAuthType(
  authType: AuthType,
 ): AvailableModel[] {
@@ -75,10 +70,6 @@ export function getAvailableModelsForAuthType(
      const openAIModel = getOpenAIAvailableModelFromEnv();
      return openAIModel ? [openAIModel] : [];
    }
-    case AuthType.USE_ANTHROPIC: {
-      const anthropicModel = getAnthropicAvailableModelFromEnv();
-      return anthropicModel ? [anthropicModel] : [];
-    }
    default:
      // For other auth types, return empty array for now
      // This can be expanded later according to the design doc
--- a/packages/cli/src/ui/utils/resumeHistoryUtils.test.ts
+++ b/packages/cli/src/ui/utils/resumeHistoryUtils.test.ts
@@ -20,11 +20,6 @@ const makeConfig = (tools: Record<string, AnyDeclarativeTool>) =>
    getToolRegistry: () => ({
      getTool: (name: string) => tools[name],
    }),
-    getContentGenerator: () => ({
-      // Default to showing full thinking content during resume unless explicitly
-      // summarized; tests don't care about summarized thinking behavior.
-      useSummarizedThinking: () => false,
-    }),
  }) as unknown as Config;

 describe('resumeHistoryUtils', () => {
--- a/packages/cli/src/ui/utils/resumeHistoryUtils.ts
+++ b/packages/cli/src/ui/utils/resumeHistoryUtils.ts
@@ -204,11 +204,7 @@ function convertToHistoryItems(
        const parts = record.message?.parts as Part[] | undefined;

        // Extract thought content
-        const thoughtText = !config
-          .getContentGenerator()
-          .useSummarizedThinking()
-          ? extractThoughtTextFromParts(parts)
-          : '';
+        const thoughtText = extractThoughtTextFromParts(parts);

        // Extract text content (non-function-call, non-thought)
        const text = extractTextFromParts(parts);
--- a/packages/cli/src/utils/nonInteractiveHelpers.test.ts
+++ b/packages/cli/src/utils/nonInteractiveHelpers.test.ts
@@ -35,33 +35,22 @@ import {
 } from './nonInteractiveHelpers.js';

 // Mock dependencies
-vi.mock('../nonInteractiveCliCommands.js', () => ({
-  getAvailableCommands: vi
-    .fn()
-    .mockImplementation(
-      async (
-        _config: unknown,
-        _signal: AbortSignal,
-        allowedBuiltinCommandNames?: string[],
-      ) => {
-        const allowedSet = new Set(allowedBuiltinCommandNames ?? []);
-        const allCommands = [
-          { name: 'help', kind: 'built-in' },
-          { name: 'commit', kind: 'file' },
-          { name: 'memory', kind: 'built-in' },
-          { name: 'init', kind: 'built-in' },
-          { name: 'summary', kind: 'built-in' },
-          { name: 'compress', kind: 'built-in' },
-        ];
+vi.mock('../services/CommandService.js', () => ({
+  CommandService: {
+    create: vi.fn().mockResolvedValue({
+      getCommands: vi
+        .fn()
+        .mockReturnValue([
+          { name: 'help' },
+          { name: 'commit' },
+          { name: 'memory' },
+        ]),
+    }),
+  },
+}));

-        // Filter commands: always include file commands, only include allowed built-in commands
-        return allCommands.filter(
-          (cmd) =>
-            cmd.kind === 'file' ||
-            (cmd.kind === 'built-in' && allowedSet.has(cmd.name)),
-        );
-      },
-    ),
+vi.mock('../services/BuiltinCommandLoader.js', () => ({
+  BuiltinCommandLoader: vi.fn().mockImplementation(() => ({})),
 }));

 vi.mock('../ui/utils/computeStats.js', () => ({
@@ -522,12 +511,10 @@ describe('buildSystemMessage', () => {
  });

  it('should build system message with all fields', async () => {
-    const allowedBuiltinCommands = ['init', 'summary', 'compress'];
    const result = await buildSystemMessage(
      mockConfig,
      'test-session-id',
      'auto' as PermissionMode,
-      allowedBuiltinCommands,
    );

    expect(result).toEqual({
@@ -543,7 +530,7 @@ describe('buildSystemMessage', () => {
      ],
      model: 'test-model',
      permission_mode: 'auto',
-      slash_commands: ['commit', 'compress', 'init', 'summary'],
+      slash_commands: ['commit', 'help', 'memory'],
      qwen_code_version: '1.0.0',
      agents: [],
    });
@@ -559,7 +546,6 @@ describe('buildSystemMessage', () => {
      config,
      'test-session-id',
      'auto' as PermissionMode,
-      ['init', 'summary'],
    );

    expect(result.tools).toEqual([]);
@@ -575,7 +561,6 @@ describe('buildSystemMessage', () => {
      config,
      'test-session-id',
      'auto' as PermissionMode,
-      ['init', 'summary'],
    );

    expect(result.mcp_servers).toEqual([]);
@@ -591,37 +576,10 @@ describe('buildSystemMessage', () => {
      config,
      'test-session-id',
      'auto' as PermissionMode,
-      ['init', 'summary'],
    );

    expect(result.qwen_code_version).toBe('unknown');
  });
-
-  it('should only include allowed built-in commands and all file commands', async () => {
-    const allowedBuiltinCommands = ['init', 'summary'];
-    const result = await buildSystemMessage(
-      mockConfig,
-      'test-session-id',
-      'auto' as PermissionMode,
-      allowedBuiltinCommands,
-    );
-
-    // Should include: 'commit' (FILE), 'init' (BUILT_IN, allowed), 'summary' (BUILT_IN, allowed)
-    // Should NOT include: 'help', 'memory', 'compress' (BUILT_IN but not in allowed set)
-    expect(result.slash_commands).toEqual(['commit', 'init', 'summary']);
-  });
-
-  it('should include only file commands when no built-in commands are allowed', async () => {
-    const result = await buildSystemMessage(
-      mockConfig,
-      'test-session-id',
-      'auto' as PermissionMode,
-      [], // Empty array - no built-in commands allowed
-    );
-
-    // Should only include 'commit' (FILE command)
-    expect(result.slash_commands).toEqual(['commit']);
-  });
 });

 describe('createTaskToolProgressHandler', () => {
--- a/packages/cli/src/utils/nonInteractiveHelpers.ts
+++ b/packages/cli/src/utils/nonInteractiveHelpers.ts
@@ -25,9 +25,10 @@ import type {
  PermissionMode,
  CLISystemMessage,
 } from '../nonInteractive/types.js';
+import { CommandService } from '../services/CommandService.js';
+import { BuiltinCommandLoader } from '../services/BuiltinCommandLoader.js';
 import type { JsonOutputAdapterInterface } from '../nonInteractive/io/BaseJsonOutputAdapter.js';
 import { computeSessionStats } from '../ui/utils/computeStats.js';
-import { getAvailableCommands } from '../nonInteractiveCliCommands.js';

 /**
 * Normalizes various part list formats into a consistent Part[] array.
@@ -186,27 +187,24 @@ export function computeUsageFromMetrics(metrics: SessionMetrics): Usage {
 }

 /**
- * Load slash command names using getAvailableCommands
+ * Load slash command names using CommandService
 *
 * @param config - Config instance
- * @param allowedBuiltinCommandNames - Optional array of allowed built-in command names.
- *   If not provided, uses the default from getAvailableCommands.
 * @returns Promise resolving to array of slash command names
 */
-async function loadSlashCommandNames(
-  config: Config,
-  allowedBuiltinCommandNames?: string[],
-): Promise<string[]> {
+async function loadSlashCommandNames(config: Config): Promise<string[]> {
  const controller = new AbortController();
  try {
-    const commands = await getAvailableCommands(
-      config,
+    const service = await CommandService.create(
+      [new BuiltinCommandLoader(config)],
      controller.signal,
-      allowedBuiltinCommandNames,
    );
-
-    // Extract command names and sort
-    return commands.map((cmd) => cmd.name).sort();
+    const names = new Set<string>();
+    const commands = service.getCommands();
+    for (const command of commands) {
+      names.add(command.name);
+    }
+    return Array.from(names).sort();
  } catch (error) {
    if (config.getDebugMode()) {
      console.error(
@@ -235,15 +233,12 @@ async function loadSlashCommandNames(
 * @param config - Config instance
 * @param sessionId - Session identifier
 * @param permissionMode - Current permission/approval mode
- * @param allowedBuiltinCommandNames - Optional array of allowed built-in command names.
- *   If not provided, defaults to empty array (only file commands will be included).
 * @returns Promise resolving to CLISystemMessage
 */
 export async function buildSystemMessage(
  config: Config,
  sessionId: string,
  permissionMode: PermissionMode,
-  allowedBuiltinCommandNames?: string[],
 ): Promise<CLISystemMessage> {
  const toolRegistry = config.getToolRegistry();
  const tools = toolRegistry ? toolRegistry.getAllToolNames() : [];
@@ -256,11 +251,8 @@ export async function buildSystemMessage(
      }))
    : [];

-  // Load slash commands with filtering based on allowed built-in commands
-  const slashCommands = await loadSlashCommandNames(
-    config,
-    allowedBuiltinCommandNames,
-  );
+  // Load slash commands
+  const slashCommands = await loadSlashCommandNames(config);

  // Load subagent names from config
  let agentNames: string[] = [];
--- a/packages/cli/src/utils/systemInfo.ts
+++ b/packages/cli/src/utils/systemInfo.ts
@@ -153,8 +153,7 @@ export async function getExtendedSystemInfo(

  // Get base URL if using OpenAI auth
  const baseUrl =
-    baseInfo.selectedAuthType === AuthType.USE_OPENAI ||
-    baseInfo.selectedAuthType === AuthType.USE_ANTHROPIC
+    baseInfo.selectedAuthType === AuthType.USE_OPENAI
      ? context.services.config?.getContentGeneratorConfig()?.baseUrl
      : undefined;

--- a/packages/cli/src/validateNonInterActiveAuth.test.ts
+++ b/packages/cli/src/validateNonInterActiveAuth.test.ts
@@ -19,9 +19,6 @@ describe('validateNonInterActiveAuth', () => {
  let originalEnvVertexAi: string | undefined;
  let originalEnvGcp: string | undefined;
  let originalEnvOpenAiApiKey: string | undefined;
-  let originalEnvQwenOauth: string | undefined;
-  let originalEnvGoogleApiKey: string | undefined;
-  let originalEnvAnthropicApiKey: string | undefined;
  let consoleErrorSpy: ReturnType<typeof vi.spyOn>;
  let processExitSpy: ReturnType<typeof vi.spyOn<[code?: number], never>>;
  let refreshAuthMock: ReturnType<typeof vi.fn>;
@@ -32,16 +29,10 @@ describe('validateNonInterActiveAuth', () => {
    originalEnvVertexAi = process.env['GOOGLE_GENAI_USE_VERTEXAI'];
    originalEnvGcp = process.env['GOOGLE_GENAI_USE_GCA'];
    originalEnvOpenAiApiKey = process.env['OPENAI_API_KEY'];
-    originalEnvQwenOauth = process.env['QWEN_OAUTH'];
-    originalEnvGoogleApiKey = process.env['GOOGLE_API_KEY'];
-    originalEnvAnthropicApiKey = process.env['ANTHROPIC_API_KEY'];
    delete process.env['GEMINI_API_KEY'];
    delete process.env['GOOGLE_GENAI_USE_VERTEXAI'];
    delete process.env['GOOGLE_GENAI_USE_GCA'];
    delete process.env['OPENAI_API_KEY'];
-    delete process.env['QWEN_OAUTH'];
-    delete process.env['GOOGLE_API_KEY'];
-    delete process.env['ANTHROPIC_API_KEY'];
    consoleErrorSpy = vi.spyOn(console, 'error').mockImplementation(() => {});
    processExitSpy = vi.spyOn(process, 'exit').mockImplementation((code) => {
      throw new Error(`process.exit(${code}) called`);
@@ -89,21 +80,6 @@ describe('validateNonInterActiveAuth', () => {
    } else {
      delete process.env['OPENAI_API_KEY'];
    }
-    if (originalEnvQwenOauth !== undefined) {
-      process.env['QWEN_OAUTH'] = originalEnvQwenOauth;
-    } else {
-      delete process.env['QWEN_OAUTH'];
-    }
-    if (originalEnvGoogleApiKey !== undefined) {
-      process.env['GOOGLE_API_KEY'] = originalEnvGoogleApiKey;
-    } else {
-      delete process.env['GOOGLE_API_KEY'];
-    }
-    if (originalEnvAnthropicApiKey !== undefined) {
-      process.env['ANTHROPIC_API_KEY'] = originalEnvAnthropicApiKey;
-    } else {
-      delete process.env['ANTHROPIC_API_KEY'];
-    }
    vi.restoreAllMocks();
  });

--- a/packages/cli/src/validateNonInterActiveAuth.ts
+++ b/packages/cli/src/validateNonInterActiveAuth.ts
@@ -27,9 +27,6 @@ function getAuthTypeFromEnv(): AuthType | undefined {
  if (process.env['GOOGLE_API_KEY']) {
    return AuthType.USE_VERTEX_AI;
  }
-  if (process.env['ANTHROPIC_API_KEY']) {
-    return AuthType.USE_ANTHROPIC;
-  }

  return undefined;
 }
--- a/packages/core/package.json
+++ b/packages/core/package.json
@@ -1,6 +1,6 @@
 {
  "name": "@qwen-code/qwen-code-core",
-  "version": "0.6.0-nightly.20251226.17eb20c1",
+  "version": "0.6.0-nightly.20251225.9f65bd3b",
  "description": "Qwen Code Core",
  "repository": {
    "type": "git",
@@ -23,7 +23,6 @@
    "scripts/postinstall.js"
  ],
  "dependencies": {
-    "@anthropic-ai/sdk": "^0.36.1",
    "@google/genai": "1.30.0",
    "@modelcontextprotocol/sdk": "^1.25.1",
    "@opentelemetry/api": "^1.9.0",
--- a/packages/core/src/config/config.test.ts
+++ b/packages/core/src/config/config.test.ts
@@ -16,6 +16,7 @@ import {
  QwenLogger,
 } from '../telemetry/index.js';
 import type { ContentGeneratorConfig } from '../core/contentGenerator.js';
+import { DEFAULT_DASHSCOPE_BASE_URL } from '../core/openaiContentGenerator/constants.js';
 import {
  AuthType,
  createContentGeneratorConfig,
@@ -272,7 +273,7 @@ describe('Server Config (config.ts)', () => {
        authType,
        {
          model: MODEL,
-          baseUrl: undefined,
+          baseUrl: DEFAULT_DASHSCOPE_BASE_URL,
        },
      );
      // Verify that contentGeneratorConfig is updated
--- a/packages/core/src/config/config.ts
+++ b/packages/core/src/config/config.ts
@@ -96,6 +96,7 @@ import {
 } from './constants.js';
 import { DEFAULT_QWEN_EMBEDDING_MODEL, DEFAULT_QWEN_MODEL } from './models.js';
 import { Storage } from './storage.js';
+import { DEFAULT_DASHSCOPE_BASE_URL } from '../core/openaiContentGenerator/constants.js';
 import { ChatRecordingService } from '../services/chatRecordingService.js';
 import {
  SessionService,
@@ -573,7 +574,7 @@ export class Config {
    this._generationConfig = {
      model: params.model,
      ...(params.generationConfig || {}),
-      baseUrl: params.generationConfig?.baseUrl,
+      baseUrl: params.generationConfig?.baseUrl || DEFAULT_DASHSCOPE_BASE_URL,
    };
    this.contentGeneratorConfig = this
      ._generationConfig as ContentGeneratorConfig;
--- a/packages/core/src/core/anthropicContentGenerator/anthropicContentGenerator.test.ts
+++ b/packages/core/src/core/anthropicContentGenerator/anthropicContentGenerator.test.ts
@@ -1,500 +0,0 @@
-/**
- * @license
- * Copyright 2025 Qwen
- * SPDX-License-Identifier: Apache-2.0
- */
-
-import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest';
-import type {
-  CountTokensParameters,
-  GenerateContentParameters,
-} from '@google/genai';
-import { FinishReason, GenerateContentResponse } from '@google/genai';
-
-// Mock the request tokenizer module BEFORE importing the class that uses it.
-const mockTokenizer = {
-  calculateTokens: vi.fn(),
-  dispose: vi.fn(),
-};
-
-vi.mock('../../utils/request-tokenizer/index.js', () => ({
-  getDefaultTokenizer: vi.fn(() => mockTokenizer),
-  DefaultRequestTokenizer: vi.fn(() => mockTokenizer),
-  disposeDefaultTokenizer: vi.fn(),
-}));
-
-type AnthropicCreateArgs = [unknown, { signal?: AbortSignal }?];
-
-const anthropicMockState: {
-  constructorOptions?: Record<string, unknown>;
-  lastCreateArgs?: AnthropicCreateArgs;
-  createImpl: ReturnType<typeof vi.fn>;
-} = {
-  constructorOptions: undefined,
-  lastCreateArgs: undefined,
-  createImpl: vi.fn(),
-};
-
-vi.mock('@anthropic-ai/sdk', () => {
-  class AnthropicMock {
-    messages: { create: (...args: AnthropicCreateArgs) => unknown };
-
-    constructor(options: Record<string, unknown>) {
-      anthropicMockState.constructorOptions = options;
-      this.messages = {
-        create: (...args: AnthropicCreateArgs) => {
-          anthropicMockState.lastCreateArgs = args;
-          return anthropicMockState.createImpl(...args);
-        },
-      };
-    }
-  }
-
-  return {
-    default: AnthropicMock,
-    __anthropicState: anthropicMockState,
-  };
-});
-
-// Now import the modules that depend on the mocked modules.
-import type { Config } from '../../config/config.js';
-
-const importGenerator = async (): Promise<{
-  AnthropicContentGenerator: typeof import('./anthropicContentGenerator.js').AnthropicContentGenerator;
-}> => import('./anthropicContentGenerator.js');
-
-const importConverter = async (): Promise<{
-  AnthropicContentConverter: typeof import('./converter.js').AnthropicContentConverter;
-}> => import('./converter.js');
-
-describe('AnthropicContentGenerator', () => {
-  let mockConfig: Config;
-  let anthropicState: {
-    constructorOptions?: Record<string, unknown>;
-    lastCreateArgs?: AnthropicCreateArgs;
-    createImpl: ReturnType<typeof vi.fn>;
-  };
-
-  beforeEach(async () => {
-    vi.clearAllMocks();
-    vi.resetModules();
-
-    mockTokenizer.calculateTokens.mockResolvedValue({
-      totalTokens: 50,
-      breakdown: {
-        textTokens: 50,
-        imageTokens: 0,
-        audioTokens: 0,
-        otherTokens: 0,
-      },
-      processingTime: 1,
-    });
-    anthropicState = anthropicMockState;
-
-    anthropicState.createImpl.mockReset();
-    anthropicState.lastCreateArgs = undefined;
-    anthropicState.constructorOptions = undefined;
-
-    mockConfig = {
-      getCliVersion: vi.fn().mockReturnValue('1.2.3'),
-    } as unknown as Config;
-  });
-
-  afterEach(() => {
-    vi.restoreAllMocks();
-  });
-
-  it('passes a QwenCode User-Agent header to the Anthropic SDK', async () => {
-    const { AnthropicContentGenerator } = await importGenerator();
-    void new AnthropicContentGenerator(
-      {
-        model: 'claude-test',
-        apiKey: 'test-key',
-        baseUrl: 'https://example.invalid',
-        timeout: 10_000,
-        maxRetries: 2,
-        samplingParams: {},
-        schemaCompliance: 'auto',
-      },
-      mockConfig,
-    );
-
-    const headers = (anthropicState.constructorOptions?.['defaultHeaders'] ||
-      {}) as Record<string, string>;
-    expect(headers['User-Agent']).toContain('QwenCode/1.2.3');
-    expect(headers['User-Agent']).toContain(
-      `(${process.platform}; ${process.arch})`,
-    );
-  });
-
-  it('adds the effort beta header when reasoning.effort is set', async () => {
-    const { AnthropicContentGenerator } = await importGenerator();
-    void new AnthropicContentGenerator(
-      {
-        model: 'claude-test',
-        apiKey: 'test-key',
-        baseUrl: 'https://example.invalid',
-        timeout: 10_000,
-        maxRetries: 2,
-        samplingParams: {},
-        schemaCompliance: 'auto',
-        reasoning: { effort: 'medium' },
-      },
-      mockConfig,
-    );
-
-    const headers = (anthropicState.constructorOptions?.['defaultHeaders'] ||
-      {}) as Record<string, string>;
-    expect(headers['anthropic-beta']).toContain('effort-2025-11-24');
-  });
-
-  it('does not add the effort beta header when reasoning.effort is not set', async () => {
-    const { AnthropicContentGenerator } = await importGenerator();
-    void new AnthropicContentGenerator(
-      {
-        model: 'claude-test',
-        apiKey: 'test-key',
-        baseUrl: 'https://example.invalid',
-        timeout: 10_000,
-        maxRetries: 2,
-        samplingParams: {},
-        schemaCompliance: 'auto',
-      },
-      mockConfig,
-    );
-
-    const headers = (anthropicState.constructorOptions?.['defaultHeaders'] ||
-      {}) as Record<string, string>;
-    expect(headers['anthropic-beta']).not.toContain('effort-2025-11-24');
-  });
-
-  it('omits the anthropic beta header when reasoning is disabled', async () => {
-    const { AnthropicContentGenerator } = await importGenerator();
-    void new AnthropicContentGenerator(
-      {
-        model: 'claude-test',
-        apiKey: 'test-key',
-        baseUrl: 'https://example.invalid',
-        timeout: 10_000,
-        maxRetries: 2,
-        samplingParams: {},
-        schemaCompliance: 'auto',
-        reasoning: false,
-      },
-      mockConfig,
-    );
-
-    const headers = (anthropicState.constructorOptions?.['defaultHeaders'] ||
-      {}) as Record<string, string>;
-    expect(headers['anthropic-beta']).toBeUndefined();
-  });
-
-  describe('generateContent', () => {
-    it('builds request with config sampling params (config overrides request) and thinking budget', async () => {
-      const { AnthropicContentConverter } = await importConverter();
-      const { AnthropicContentGenerator } = await importGenerator();
-
-      const convertResponseSpy = vi
-        .spyOn(
-          AnthropicContentConverter.prototype,
-          'convertAnthropicResponseToGemini',
-        )
-        .mockReturnValue(
-          (() => {
-            const r = new GenerateContentResponse();
-            r.responseId = 'gemini-1';
-            return r;
-          })(),
-        );
-
-      anthropicState.createImpl.mockResolvedValue({
-        id: 'anthropic-1',
-        model: 'claude-test',
-        content: [{ type: 'text', text: 'hi' }],
-      });
-
-      const generator = new AnthropicContentGenerator(
-        {
-          model: 'claude-test',
-          apiKey: 'test-key',
-          baseUrl: 'https://example.invalid',
-          timeout: 10_000,
-          maxRetries: 2,
-          samplingParams: {
-            temperature: 0.7,
-            max_tokens: 1000,
-            top_p: 0.9,
-            top_k: 20,
-          },
-          schemaCompliance: 'auto',
-          reasoning: { effort: 'high', budget_tokens: 1000 },
-        },
-        mockConfig,
-      );
-
-      const abortController = new AbortController();
-      const request: GenerateContentParameters = {
-        model: 'models/ignored',
-        contents: 'Hello',
-        config: {
-          temperature: 0.1,
-          maxOutputTokens: 200,
-          topP: 0.5,
-          topK: 5,
-          abortSignal: abortController.signal,
-        },
-      };
-
-      const result = await generator.generateContent(request);
-      expect(result.responseId).toBe('gemini-1');
-
-      expect(anthropicState.lastCreateArgs).toBeDefined();
-      const [anthropicRequest, options] =
-        anthropicState.lastCreateArgs as AnthropicCreateArgs;
-
-      expect(options?.signal).toBe(abortController.signal);
-
-      expect(anthropicRequest).toEqual(
-        expect.objectContaining({
-          model: 'claude-test',
-          max_tokens: 1000,
-          temperature: 0.7,
-          top_p: 0.9,
-          top_k: 20,
-          thinking: { type: 'enabled', budget_tokens: 1000 },
-          output_config: { effort: 'high' },
-        }),
-      );
-
-      expect(convertResponseSpy).toHaveBeenCalledTimes(1);
-    });
-
-    it('omits thinking when request.config.thinkingConfig.includeThoughts is false', async () => {
-      const { AnthropicContentGenerator } = await importGenerator();
-      anthropicState.createImpl.mockResolvedValue({
-        id: 'anthropic-1',
-        model: 'claude-test',
-        content: [{ type: 'text', text: 'hi' }],
-      });
-
-      const generator = new AnthropicContentGenerator(
-        {
-          model: 'claude-test',
-          apiKey: 'test-key',
-          timeout: 10_000,
-          maxRetries: 2,
-          samplingParams: { max_tokens: 500 },
-          schemaCompliance: 'auto',
-          reasoning: { effort: 'high' },
-        },
-        mockConfig,
-      );
-
-      await generator.generateContent({
-        model: 'models/ignored',
-        contents: 'Hello',
-        config: { thinkingConfig: { includeThoughts: false } },
-      } as unknown as GenerateContentParameters);
-
-      const [anthropicRequest] =
-        anthropicState.lastCreateArgs as AnthropicCreateArgs;
-      expect(anthropicRequest).toEqual(
-        expect.not.objectContaining({ thinking: expect.anything() }),
-      );
-    });
-  });
-
-  describe('countTokens', () => {
-    it('counts tokens using the request tokenizer', async () => {
-      const { AnthropicContentGenerator } = await importGenerator();
-      const generator = new AnthropicContentGenerator(
-        {
-          model: 'claude-test',
-          apiKey: 'test-key',
-          timeout: 10_000,
-          maxRetries: 2,
-          samplingParams: {},
-          schemaCompliance: 'auto',
-        },
-        mockConfig,
-      );
-
-      const request: CountTokensParameters = {
-        contents: [{ role: 'user', parts: [{ text: 'Hello world' }] }],
-        model: 'claude-test',
-      };
-
-      const result = await generator.countTokens(request);
-      expect(mockTokenizer.calculateTokens).toHaveBeenCalledWith(request, {
-        textEncoding: 'cl100k_base',
-      });
-      expect(result.totalTokens).toBe(50);
-    });
-
-    it('falls back to character approximation when tokenizer throws', async () => {
-      const { AnthropicContentGenerator } = await importGenerator();
-      mockTokenizer.calculateTokens.mockRejectedValueOnce(new Error('boom'));
-      const generator = new AnthropicContentGenerator(
-        {
-          model: 'claude-test',
-          apiKey: 'test-key',
-          timeout: 10_000,
-          maxRetries: 2,
-          samplingParams: {},
-          schemaCompliance: 'auto',
-        },
-        mockConfig,
-      );
-
-      const request: CountTokensParameters = {
-        contents: [{ role: 'user', parts: [{ text: 'Hello' }] }],
-        model: 'claude-test',
-      };
-
-      const content = JSON.stringify(request.contents);
-      const expected = Math.ceil(content.length / 4);
-      const result = await generator.countTokens(request);
-      expect(result.totalTokens).toBe(expected);
-    });
-  });
-
-  describe('generateContentStream', () => {
-    it('requests stream=true and converts streamed events into Gemini chunks', async () => {
-      const { AnthropicContentGenerator } = await importGenerator();
-      anthropicState.createImpl.mockResolvedValue(
-        (async function* () {
-          yield {
-            type: 'message_start',
-            message: {
-              id: 'msg-1',
-              model: 'claude-test',
-              usage: { cache_read_input_tokens: 2, input_tokens: 3 },
-            },
-          };
-
-          yield {
-            type: 'content_block_start',
-            index: 0,
-            content_block: { type: 'text' },
-          };
-          yield {
-            type: 'content_block_delta',
-            index: 0,
-            delta: { type: 'text_delta', text: 'Hello' },
-          };
-          yield { type: 'content_block_stop', index: 0 };
-
-          yield {
-            type: 'content_block_start',
-            index: 1,
-            content_block: { type: 'thinking', signature: '' },
-          };
-          yield {
-            type: 'content_block_delta',
-            index: 1,
-            delta: { type: 'thinking_delta', thinking: 'Think' },
-          };
-          yield {
-            type: 'content_block_delta',
-            index: 1,
-            delta: { type: 'signature_delta', signature: 'abc' },
-          };
-          yield { type: 'content_block_stop', index: 1 };
-
-          yield {
-            type: 'content_block_start',
-            index: 2,
-            content_block: {
-              type: 'tool_use',
-              id: 't1',
-              name: 'tool',
-              input: {},
-            },
-          };
-          yield {
-            type: 'content_block_delta',
-            index: 2,
-            delta: { type: 'input_json_delta', partial_json: '{"x":' },
-          };
-          yield {
-            type: 'content_block_delta',
-            index: 2,
-            delta: { type: 'input_json_delta', partial_json: '1}' },
-          };
-          yield { type: 'content_block_stop', index: 2 };
-
-          yield {
-            type: 'message_delta',
-            delta: { stop_reason: 'end_turn' },
-            usage: {
-              output_tokens: 5,
-              input_tokens: 7,
-              cache_read_input_tokens: 2,
-            },
-          };
-          yield { type: 'message_stop' };
-        })(),
-      );
-
-      const generator = new AnthropicContentGenerator(
-        {
-          model: 'claude-test',
-          apiKey: 'test-key',
-          timeout: 10_000,
-          maxRetries: 2,
-          samplingParams: { max_tokens: 123 },
-          schemaCompliance: 'auto',
-        },
-        mockConfig,
-      );
-
-      const stream = await generator.generateContentStream({
-        model: 'models/ignored',
-        contents: 'Hello',
-      } as unknown as GenerateContentParameters);
-
-      const chunks: GenerateContentResponse[] = [];
-      for await (const chunk of stream) {
-        chunks.push(chunk);
-      }
-
-      const [anthropicRequest] =
-        anthropicState.lastCreateArgs as AnthropicCreateArgs;
-      expect(anthropicRequest).toEqual(
-        expect.objectContaining({ stream: true }),
-      );
-
-      // Text chunk.
-      expect(chunks[0]?.candidates?.[0]?.content?.parts?.[0]).toEqual({
-        text: 'Hello',
-      });
-
-      // Thinking chunk.
-      expect(chunks[1]?.candidates?.[0]?.content?.parts?.[0]).toEqual({
-        text: 'Think',
-        thought: true,
-      });
-
-      // Signature chunk.
-      expect(chunks[2]?.candidates?.[0]?.content?.parts?.[0]).toEqual({
-        thought: true,
-        thoughtSignature: 'abc',
-      });
-
-      // Tool call chunk.
-      expect(chunks[3]?.candidates?.[0]?.content?.parts?.[0]).toEqual({
-        functionCall: { id: 't1', name: 'tool', args: { x: 1 } },
-      });
-
-      // Usage/finish chunks exist; check the last one.
-      const last = chunks[chunks.length - 1]!;
-      expect(last.candidates?.[0]?.finishReason).toBe(FinishReason.STOP);
-      expect(last.usageMetadata).toEqual({
-        cachedContentTokenCount: 2,
-        promptTokenCount: 9, // cached(2) + input(7)
-        candidatesTokenCount: 5,
-        totalTokenCount: 14,
-      });
-    });
-  });
-});
--- a/packages/core/src/core/anthropicContentGenerator/anthropicContentGenerator.ts
+++ b/packages/core/src/core/anthropicContentGenerator/anthropicContentGenerator.ts
@@ -1,502 +0,0 @@
-/**
- * @license
- * Copyright 2025 Qwen
- * SPDX-License-Identifier: Apache-2.0
- */
-
-import Anthropic from '@anthropic-ai/sdk';
-import type {
-  CountTokensParameters,
-  CountTokensResponse,
-  EmbedContentParameters,
-  EmbedContentResponse,
-  GenerateContentParameters,
-  GenerateContentResponseUsageMetadata,
-  Part,
-} from '@google/genai';
-import { GenerateContentResponse } from '@google/genai';
-import type { Config } from '../../config/config.js';
-import type {
-  ContentGenerator,
-  ContentGeneratorConfig,
-} from '../contentGenerator.js';
-type Message = Anthropic.Message;
-type MessageCreateParamsNonStreaming =
-  Anthropic.MessageCreateParamsNonStreaming;
-type MessageCreateParamsStreaming = Anthropic.MessageCreateParamsStreaming;
-type RawMessageStreamEvent = Anthropic.RawMessageStreamEvent;
-import { getDefaultTokenizer } from '../../utils/request-tokenizer/index.js';
-import { safeJsonParse } from '../../utils/safeJsonParse.js';
-import { AnthropicContentConverter } from './converter.js';
-
-type StreamingBlockState = {
-  type: string;
-  id?: string;
-  name?: string;
-  inputJson: string;
-  signature: string;
-};
-
-type MessageCreateParamsWithThinking = MessageCreateParamsNonStreaming & {
-  thinking?: { type: 'enabled'; budget_tokens: number };
-  // Anthropic beta feature: output_config.effort (requires beta header effort-2025-11-24)
-  // This is not yet represented in the official SDK types we depend on.
-  output_config?: { effort: 'low' | 'medium' | 'high' };
-};
-
-export class AnthropicContentGenerator implements ContentGenerator {
-  private client: Anthropic;
-  private converter: AnthropicContentConverter;
-
-  constructor(
-    private contentGeneratorConfig: ContentGeneratorConfig,
-    private readonly cliConfig: Config,
-  ) {
-    const defaultHeaders = this.buildHeaders();
-    const baseURL = contentGeneratorConfig.baseUrl;
-
-    this.client = new Anthropic({
-      apiKey: contentGeneratorConfig.apiKey,
-      baseURL,
-      timeout: contentGeneratorConfig.timeout,
-      maxRetries: contentGeneratorConfig.maxRetries,
-      defaultHeaders,
-    });
-
-    this.converter = new AnthropicContentConverter(
-      contentGeneratorConfig.model,
-      contentGeneratorConfig.schemaCompliance,
-    );
-  }
-
-  async generateContent(
-    request: GenerateContentParameters,
-  ): Promise<GenerateContentResponse> {
-    const anthropicRequest = await this.buildRequest(request);
-    const response = (await this.client.messages.create(anthropicRequest, {
-      signal: request.config?.abortSignal,
-    })) as Message;
-
-    return this.converter.convertAnthropicResponseToGemini(response);
-  }
-
-  async generateContentStream(
-    request: GenerateContentParameters,
-  ): Promise<AsyncGenerator<GenerateContentResponse>> {
-    const anthropicRequest = await this.buildRequest(request);
-    const streamingRequest: MessageCreateParamsStreaming & {
-      thinking?: { type: 'enabled'; budget_tokens: number };
-    } = {
-      ...anthropicRequest,
-      stream: true,
-    };
-
-    const stream = (await this.client.messages.create(
-      streamingRequest as MessageCreateParamsStreaming,
-      {
-        signal: request.config?.abortSignal,
-      },
-    )) as AsyncIterable<RawMessageStreamEvent>;
-
-    return this.processStream(stream);
-  }
-
-  async countTokens(
-    request: CountTokensParameters,
-  ): Promise<CountTokensResponse> {
-    try {
-      const tokenizer = getDefaultTokenizer();
-      const result = await tokenizer.calculateTokens(request, {
-        textEncoding: 'cl100k_base',
-      });
-
-      return {
-        totalTokens: result.totalTokens,
-      };
-    } catch (error) {
-      console.warn(
-        'Failed to calculate tokens with tokenizer, ' +
-          'falling back to simple method:',
-        error,
-      );
-
-      const content = JSON.stringify(request.contents);
-      const totalTokens = Math.ceil(content.length / 4);
-      return {
-        totalTokens,
-      };
-    }
-  }
-
-  async embedContent(
-    _request: EmbedContentParameters,
-  ): Promise<EmbedContentResponse> {
-    throw new Error('Anthropic does not support embeddings.');
-  }
-
-  useSummarizedThinking(): boolean {
-    return false;
-  }
-
-  private buildHeaders(): Record<string, string> {
-    const version = this.cliConfig.getCliVersion() || 'unknown';
-    const userAgent = `QwenCode/${version} (${process.platform}; ${process.arch})`;
-
-    const betas: string[] = [];
-    const reasoning = this.contentGeneratorConfig.reasoning;
-
-    // Interleaved thinking is used when we send the `thinking` field.
-    if (reasoning !== false) {
-      betas.push('interleaved-thinking-2025-05-14');
-    }
-
-    // Effort (beta) is enabled when reasoning.effort is set.
-    if (reasoning !== false && reasoning?.effort !== undefined) {
-      betas.push('effort-2025-11-24');
-    }
-
-    const headers: Record<string, string> = {
-      'User-Agent': userAgent,
-    };
-
-    if (betas.length) {
-      headers['anthropic-beta'] = betas.join(',');
-    }
-
-    return headers;
-  }
-
-  private async buildRequest(
-    request: GenerateContentParameters,
-  ): Promise<MessageCreateParamsWithThinking> {
-    const { system, messages } =
-      this.converter.convertGeminiRequestToAnthropic(request);
-
-    const tools = request.config?.tools
-      ? await this.converter.convertGeminiToolsToAnthropic(request.config.tools)
-      : undefined;
-
-    const sampling = this.buildSamplingParameters(request);
-    const thinking = this.buildThinkingConfig(request);
-    const outputConfig = this.buildOutputConfig();
-
-    return {
-      model: this.contentGeneratorConfig.model,
-      system,
-      messages,
-      tools,
-      ...sampling,
-      ...(thinking ? { thinking } : {}),
-      ...(outputConfig ? { output_config: outputConfig } : {}),
-    };
-  }
-
-  private buildSamplingParameters(request: GenerateContentParameters): {
-    max_tokens: number;
-    temperature?: number;
-    top_p?: number;
-    top_k?: number;
-  } {
-    const configSamplingParams = this.contentGeneratorConfig.samplingParams;
-    const requestConfig = request.config || {};
-
-    const getParam = <T>(
-      configKey: keyof NonNullable<typeof configSamplingParams>,
-      requestKey?: keyof NonNullable<typeof requestConfig>,
-    ): T | undefined => {
-      const configValue = configSamplingParams?.[configKey] as T | undefined;
-      const requestValue = requestKey
-        ? (requestConfig[requestKey] as T | undefined)
-        : undefined;
-      return configValue !== undefined ? configValue : requestValue;
-    };
-
-    const maxTokens =
-      getParam<number>('max_tokens', 'maxOutputTokens') ?? 10_000;
-
-    return {
-      max_tokens: maxTokens,
-      temperature: getParam<number>('temperature', 'temperature') ?? 1,
-      top_p: getParam<number>('top_p', 'topP'),
-      top_k: getParam<number>('top_k', 'topK'),
-    };
-  }
-
-  private buildThinkingConfig(
-    request: GenerateContentParameters,
-  ): { type: 'enabled'; budget_tokens: number } | undefined {
-    if (request.config?.thinkingConfig?.includeThoughts === false) {
-      return undefined;
-    }
-
-    const reasoning = this.contentGeneratorConfig.reasoning;
-
-    if (reasoning === false) {
-      return undefined;
-    }
-
-    if (reasoning?.budget_tokens !== undefined) {
-      return {
-        type: 'enabled',
-        budget_tokens: reasoning.budget_tokens,
-      };
-    }
-
-    const effort = reasoning?.effort;
-    // When using interleaved thinking with tools, this budget token limit is the entire context window(200k tokens).
-    const budgetTokens =
-      effort === 'low' ? 16_000 : effort === 'high' ? 64_000 : 32_000;
-
-    return {
-      type: 'enabled',
-      budget_tokens: budgetTokens,
-    };
-  }
-
-  private buildOutputConfig():
-    | { effort: 'low' | 'medium' | 'high' }
-    | undefined {
-    const reasoning = this.contentGeneratorConfig.reasoning;
-    if (reasoning === false || reasoning === undefined) {
-      return undefined;
-    }
-
-    if (reasoning.effort === undefined) {
-      return undefined;
-    }
-
-    return { effort: reasoning.effort };
-  }
-
-  private async *processStream(
-    stream: AsyncIterable<RawMessageStreamEvent>,
-  ): AsyncGenerator<GenerateContentResponse> {
-    let messageId: string | undefined;
-    let model = this.contentGeneratorConfig.model;
-    let cachedTokens = 0;
-    let promptTokens = 0;
-    let completionTokens = 0;
-    let finishReason: string | undefined;
-
-    const blocks = new Map<number, StreamingBlockState>();
-    const collectedResponses: GenerateContentResponse[] = [];
-
-    for await (const event of stream) {
-      switch (event.type) {
-        case 'message_start': {
-          messageId = event.message.id ?? messageId;
-          model = event.message.model ?? model;
-          cachedTokens =
-            event.message.usage?.cache_read_input_tokens ?? cachedTokens;
-          promptTokens = event.message.usage?.input_tokens ?? promptTokens;
-          break;
-        }
-        case 'content_block_start': {
-          const index = event.index ?? 0;
-          const type = String(event.content_block.type || 'text');
-          const initialInput =
-            type === 'tool_use' && 'input' in event.content_block
-              ? JSON.stringify(event.content_block.input)
-              : '';
-          blocks.set(index, {
-            type,
-            id:
-              'id' in event.content_block ? event.content_block.id : undefined,
-            name:
-              'name' in event.content_block
-                ? event.content_block.name
-                : undefined,
-            inputJson: initialInput !== '{}' ? initialInput : '',
-            signature:
-              type === 'thinking' &&
-              'signature' in event.content_block &&
-              typeof event.content_block.signature === 'string'
-                ? event.content_block.signature
-                : '',
-          });
-          break;
-        }
-        case 'content_block_delta': {
-          const index = event.index ?? 0;
-          const deltaType = (event.delta as { type?: string }).type || '';
-          const blockState = blocks.get(index);
-
-          if (deltaType === 'text_delta') {
-            const text = 'text' in event.delta ? event.delta.text : '';
-            if (text) {
-              const chunk = this.buildGeminiChunk({ text }, messageId, model);
-              collectedResponses.push(chunk);
-              yield chunk;
-            }
-          } else if (deltaType === 'thinking_delta') {
-            const thinking =
-              (event.delta as { thinking?: string }).thinking || '';
-            if (thinking) {
-              const chunk = this.buildGeminiChunk(
-                { text: thinking, thought: true },
-                messageId,
-                model,
-              );
-              collectedResponses.push(chunk);
-              yield chunk;
-            }
-          } else if (deltaType === 'signature_delta' && blockState) {
-            const signature =
-              (event.delta as { signature?: string }).signature || '';
-            if (signature) {
-              blockState.signature += signature;
-              const chunk = this.buildGeminiChunk(
-                { thought: true, thoughtSignature: signature },
-                messageId,
-                model,
-              );
-              collectedResponses.push(chunk);
-              yield chunk;
-            }
-          } else if (deltaType === 'input_json_delta' && blockState) {
-            const jsonDelta =
-              (event.delta as { partial_json?: string }).partial_json || '';
-            if (jsonDelta) {
-              blockState.inputJson += jsonDelta;
-            }
-          }
-          break;
-        }
-        case 'content_block_stop': {
-          const index = event.index ?? 0;
-          const blockState = blocks.get(index);
-          if (blockState?.type === 'tool_use') {
-            const args = safeJsonParse(blockState.inputJson || '{}', {});
-            const chunk = this.buildGeminiChunk(
-              {
-                functionCall: {
-                  id: blockState.id,
-                  name: blockState.name,
-                  args,
-                },
-              },
-              messageId,
-              model,
-            );
-            collectedResponses.push(chunk);
-            yield chunk;
-          }
-          blocks.delete(index);
-          break;
-        }
-        case 'message_delta': {
-          const stopReasonValue = event.delta.stop_reason;
-          if (stopReasonValue) {
-            finishReason = stopReasonValue;
-          }
-
-          // Some Anthropic-compatible providers may include additional usage fields
-          // (e.g. `input_tokens`, `cache_read_input_tokens`) even though the official
-          // Anthropic SDK types only expose `output_tokens` here.
-          const usageUnknown = event.usage as unknown;
-          const usageRecord =
-            usageUnknown && typeof usageUnknown === 'object'
-              ? (usageUnknown as Record<string, unknown>)
-              : undefined;
-
-          if (event.usage?.output_tokens !== undefined) {
-            completionTokens = event.usage.output_tokens;
-          }
-          if (usageRecord?.['input_tokens'] !== undefined) {
-            const inputTokens = usageRecord['input_tokens'];
-            if (typeof inputTokens === 'number') {
-              promptTokens = inputTokens;
-            }
-          }
-          if (usageRecord?.['cache_read_input_tokens'] !== undefined) {
-            const cacheRead = usageRecord['cache_read_input_tokens'];
-            if (typeof cacheRead === 'number') {
-              cachedTokens = cacheRead;
-            }
-          }
-
-          if (finishReason || event.usage) {
-            const chunk = this.buildGeminiChunk(
-              undefined,
-              messageId,
-              model,
-              finishReason,
-              {
-                cachedContentTokenCount: cachedTokens,
-                promptTokenCount: cachedTokens + promptTokens,
-                candidatesTokenCount: completionTokens,
-                totalTokenCount: cachedTokens + promptTokens + completionTokens,
-              },
-            );
-            collectedResponses.push(chunk);
-            yield chunk;
-          }
-          break;
-        }
-        case 'message_stop': {
-          if (promptTokens || completionTokens) {
-            const chunk = this.buildGeminiChunk(
-              undefined,
-              messageId,
-              model,
-              finishReason,
-              {
-                cachedContentTokenCount: cachedTokens,
-                promptTokenCount: cachedTokens + promptTokens,
-                candidatesTokenCount: completionTokens,
-                totalTokenCount: cachedTokens + promptTokens + completionTokens,
-              },
-            );
-            collectedResponses.push(chunk);
-            yield chunk;
-          }
-          break;
-        }
-        default:
-          break;
-      }
-    }
-  }
-
-  private buildGeminiChunk(
-    part?: {
-      text?: string;
-      thought?: boolean;
-      thoughtSignature?: string;
-      functionCall?: unknown;
-    },
-    responseId?: string,
-    model?: string,
-    finishReason?: string,
-    usageMetadata?: GenerateContentResponseUsageMetadata,
-  ): GenerateContentResponse {
-    const response = new GenerateContentResponse();
-    response.responseId = responseId;
-    response.createTime = Date.now().toString();
-    response.modelVersion = model || this.contentGeneratorConfig.model;
-    response.promptFeedback = { safetyRatings: [] };
-
-    const candidateParts = part ? [part as unknown as Part] : [];
-    const mappedFinishReason =
-      finishReason !== undefined
-        ? this.converter.mapAnthropicFinishReasonToGemini(finishReason)
-        : undefined;
-    response.candidates = [
-      {
-        content: {
-          parts: candidateParts,
-          role: 'model' as const,
-        },
-        index: 0,
-        safetyRatings: [],
-        ...(mappedFinishReason ? { finishReason: mappedFinishReason } : {}),
-      },
-    ];
-
-    if (usageMetadata) {
-      response.usageMetadata = usageMetadata;
-    }
-
-    return response;
-  }
-}
--- a/packages/core/src/core/anthropicContentGenerator/converter.test.ts
+++ b/packages/core/src/core/anthropicContentGenerator/converter.test.ts
@@ -1,377 +0,0 @@
-/**
- * @license
- * Copyright 2025 Qwen
- * SPDX-License-Identifier: Apache-2.0
- */
-
-import { beforeEach, describe, expect, it, vi } from 'vitest';
-import type { CallableTool, Content, Tool } from '@google/genai';
-import { FinishReason } from '@google/genai';
-import type Anthropic from '@anthropic-ai/sdk';
-
-// Mock schema conversion so we can force edge-cases (e.g. missing `type`).
-vi.mock('../../utils/schemaConverter.js', () => ({
-  convertSchema: vi.fn((schema: unknown) => schema),
-}));
-
-import { convertSchema } from '../../utils/schemaConverter.js';
-import { AnthropicContentConverter } from './converter.js';
-
-describe('AnthropicContentConverter', () => {
-  let converter: AnthropicContentConverter;
-
-  beforeEach(() => {
-    vi.clearAllMocks();
-    converter = new AnthropicContentConverter('test-model', 'auto');
-  });
-
-  describe('convertGeminiRequestToAnthropic', () => {
-    it('extracts systemInstruction text from string', () => {
-      const { system } = converter.convertGeminiRequestToAnthropic({
-        model: 'models/test',
-        contents: 'hi',
-        config: { systemInstruction: 'sys' },
-      });
-
-      expect(system).toBe('sys');
-    });
-
-    it('extracts systemInstruction text from parts and joins with newlines', () => {
-      const { system } = converter.convertGeminiRequestToAnthropic({
-        model: 'models/test',
-        contents: 'hi',
-        config: {
-          systemInstruction: {
-            role: 'system',
-            parts: [{ text: 'a' }, { text: 'b' }],
-          } as unknown as Content,
-        },
-      });
-
-      expect(system).toBe('a\nb');
-    });
-
-    it('converts a plain string content into a user message', () => {
-      const { messages } = converter.convertGeminiRequestToAnthropic({
-        model: 'models/test',
-        contents: 'Hello',
-      });
-
-      expect(messages).toEqual([
-        { role: 'user', content: [{ type: 'text', text: 'Hello' }] },
-      ]);
-    });
-
-    it('converts user content parts into a user message with text blocks', () => {
-      const { messages } = converter.convertGeminiRequestToAnthropic({
-        model: 'models/test',
-        contents: [
-          {
-            role: 'user',
-            parts: [{ text: 'Hello' }, { text: 'World' }],
-          },
-        ],
-      });
-
-      expect(messages).toEqual([
-        {
-          role: 'user',
-          content: [
-            { type: 'text', text: 'Hello' },
-            { type: 'text', text: 'World' },
-          ],
-        },
-      ]);
-    });
-
-    it('converts assistant thought parts into Anthropic thinking blocks', () => {
-      const { messages } = converter.convertGeminiRequestToAnthropic({
-        model: 'models/test',
-        contents: [
-          {
-            role: 'model',
-            parts: [
-              { text: 'internal', thought: true, thoughtSignature: 'sig' },
-              { text: 'visible' },
-            ],
-          },
-        ],
-      });
-
-      expect(messages).toEqual([
-        {
-          role: 'assistant',
-          content: [
-            { type: 'thinking', thinking: 'internal', signature: 'sig' },
-            { type: 'text', text: 'visible' },
-          ],
-        },
-      ]);
-    });
-
-    it('converts functionCall parts from model role into tool_use blocks', () => {
-      const { messages } = converter.convertGeminiRequestToAnthropic({
-        model: 'models/test',
-        contents: [
-          {
-            role: 'model',
-            parts: [
-              { text: 'preface' },
-              {
-                functionCall: {
-                  id: 'call-1',
-                  name: 'tool_name',
-                  args: { a: 1 },
-                },
-              },
-            ],
-          },
-        ],
-      });
-
-      expect(messages).toEqual([
-        {
-          role: 'assistant',
-          content: [
-            { type: 'text', text: 'preface' },
-            {
-              type: 'tool_use',
-              id: 'call-1',
-              name: 'tool_name',
-              input: { a: 1 },
-            },
-          ],
-        },
-      ]);
-    });
-
-    it('converts functionResponse parts into user tool_result messages', () => {
-      const { messages } = converter.convertGeminiRequestToAnthropic({
-        model: 'models/test',
-        contents: [
-          {
-            role: 'user',
-            parts: [
-              {
-                functionResponse: {
-                  id: 'call-1',
-                  name: 'tool_name',
-                  response: { output: 'ok' },
-                },
-              },
-            ],
-          },
-        ],
-      });
-
-      expect(messages).toEqual([
-        {
-          role: 'user',
-          content: [
-            {
-              type: 'tool_result',
-              tool_use_id: 'call-1',
-              content: 'ok',
-            },
-          ],
-        },
-      ]);
-    });
-
-    it('extracts function response error field when present', () => {
-      const { messages } = converter.convertGeminiRequestToAnthropic({
-        model: 'models/test',
-        contents: [
-          {
-            role: 'user',
-            parts: [
-              {
-                functionResponse: {
-                  id: 'call-1',
-                  name: 'tool_name',
-                  response: { error: 'boom' },
-                },
-              },
-            ],
-          },
-        ],
-      });
-
-      expect(messages[0]).toEqual({
-        role: 'user',
-        content: [
-          {
-            type: 'tool_result',
-            tool_use_id: 'call-1',
-            content: 'boom',
-          },
-        ],
-      });
-    });
-  });
-
-  describe('convertGeminiToolsToAnthropic', () => {
-    it('converts Tool.functionDeclarations to Anthropic tools and runs schema conversion', async () => {
-      const tools = [
-        {
-          functionDeclarations: [
-            {
-              name: 'get_weather',
-              description: 'Get weather',
-              parametersJsonSchema: {
-                type: 'object',
-                properties: { location: { type: 'string' } },
-                required: ['location'],
-              },
-            },
-          ],
-        },
-      ] as Tool[];
-
-      const result = await converter.convertGeminiToolsToAnthropic(tools);
-
-      expect(result).toHaveLength(1);
-      expect(result[0]).toEqual({
-        name: 'get_weather',
-        description: 'Get weather',
-        input_schema: {
-          type: 'object',
-          properties: { location: { type: 'string' } },
-          required: ['location'],
-        },
-      });
-
-      expect(vi.mocked(convertSchema)).toHaveBeenCalledTimes(1);
-    });
-
-    it('resolves CallableTool.tool() and converts its functionDeclarations', async () => {
-      const callable = [
-        {
-          tool: async () =>
-            ({
-              functionDeclarations: [
-                {
-                  name: 'dynamic_tool',
-                  description: 'resolved tool',
-                  parametersJsonSchema: { type: 'object', properties: {} },
-                },
-              ],
-            }) as unknown as Tool,
-        },
-      ] as CallableTool[];
-
-      const result = await converter.convertGeminiToolsToAnthropic(callable);
-
-      expect(result).toHaveLength(1);
-      expect(result[0].name).toBe('dynamic_tool');
-    });
-
-    it('defaults missing parameters to an empty object schema', async () => {
-      const tools = [
-        {
-          functionDeclarations: [
-            { name: 'no_params', description: 'no params' },
-          ],
-        },
-      ] as Tool[];
-
-      const result = await converter.convertGeminiToolsToAnthropic(tools);
-
-      expect(result).toHaveLength(1);
-      expect(result[0]).toEqual({
-        name: 'no_params',
-        description: 'no params',
-        input_schema: { type: 'object', properties: {} },
-      });
-    });
-
-    it('forces input_schema.type to "object" when schema conversion yields no type', async () => {
-      vi.mocked(convertSchema).mockImplementationOnce(() => ({
-        properties: {},
-      }));
-      const tools = [
-        {
-          functionDeclarations: [
-            {
-              name: 'edge',
-              description: 'edge',
-              parametersJsonSchema: { type: 'object', properties: {} },
-            },
-          ],
-        },
-      ] as Tool[];
-
-      const result = await converter.convertGeminiToolsToAnthropic(tools);
-      expect(result[0]?.input_schema?.type).toBe('object');
-    });
-  });
-
-  describe('convertAnthropicResponseToGemini', () => {
-    it('converts text, tool_use, thinking, and redacted_thinking blocks', () => {
-      const response = converter.convertAnthropicResponseToGemini({
-        id: 'msg-1',
-        model: 'claude-test',
-        stop_reason: 'end_turn',
-        content: [
-          { type: 'thinking', thinking: 'thought', signature: 'sig' },
-          { type: 'text', text: 'hello' },
-          { type: 'tool_use', id: 't1', name: 'tool', input: { x: 1 } },
-          { type: 'redacted_thinking' },
-        ],
-        usage: { input_tokens: 3, output_tokens: 5 },
-      } as unknown as Anthropic.Message);
-
-      expect(response.responseId).toBe('msg-1');
-      expect(response.modelVersion).toBe('claude-test');
-      expect(response.candidates?.[0]?.finishReason).toBe(FinishReason.STOP);
-      expect(response.usageMetadata).toEqual({
-        promptTokenCount: 3,
-        candidatesTokenCount: 5,
-        totalTokenCount: 8,
-      });
-
-      const parts = response.candidates?.[0]?.content?.parts || [];
-      expect(parts).toEqual([
-        { text: 'thought', thought: true, thoughtSignature: 'sig' },
-        { text: 'hello' },
-        { functionCall: { id: 't1', name: 'tool', args: { x: 1 } } },
-        { text: '', thought: true },
-      ]);
-    });
-
-    it('handles tool_use input that is a JSON string', () => {
-      const response = converter.convertAnthropicResponseToGemini({
-        id: 'msg-1',
-        model: 'claude-test',
-        stop_reason: null,
-        content: [
-          { type: 'tool_use', id: 't1', name: 'tool', input: '{"x":1}' },
-        ],
-      } as unknown as Anthropic.Message);
-
-      const parts = response.candidates?.[0]?.content?.parts || [];
-      expect(parts).toEqual([
-        { functionCall: { id: 't1', name: 'tool', args: { x: 1 } } },
-      ]);
-    });
-  });
-
-  describe('mapAnthropicFinishReasonToGemini', () => {
-    it('maps known reasons', () => {
-      expect(converter.mapAnthropicFinishReasonToGemini('end_turn')).toBe(
-        FinishReason.STOP,
-      );
-      expect(converter.mapAnthropicFinishReasonToGemini('max_tokens')).toBe(
-        FinishReason.MAX_TOKENS,
-      );
-      expect(converter.mapAnthropicFinishReasonToGemini('content_filter')).toBe(
-        FinishReason.SAFETY,
-      );
-    });
-
-    it('returns undefined for null/empty', () => {
-      expect(converter.mapAnthropicFinishReasonToGemini(null)).toBeUndefined();
-      expect(converter.mapAnthropicFinishReasonToGemini('')).toBeUndefined();
-    });
-  });
-});
--- a/packages/core/src/core/anthropicContentGenerator/converter.ts
+++ b/packages/core/src/core/anthropicContentGenerator/converter.ts
@@ -1,448 +0,0 @@
-/**
- * @license
- * Copyright 2025 Qwen
- * SPDX-License-Identifier: Apache-2.0
- */
-
-import type {
-  Candidate,
-  CallableTool,
-  Content,
-  ContentListUnion,
-  ContentUnion,
-  FunctionCall,
-  FunctionResponse,
-  GenerateContentParameters,
-  Part,
-  PartUnion,
-  Tool,
-  ToolListUnion,
-} from '@google/genai';
-import { FinishReason, GenerateContentResponse } from '@google/genai';
-import type Anthropic from '@anthropic-ai/sdk';
-import { safeJsonParse } from '../../utils/safeJsonParse.js';
-import {
-  convertSchema,
-  type SchemaComplianceMode,
-} from '../../utils/schemaConverter.js';
-
-type AnthropicMessageParam = Anthropic.MessageParam;
-type AnthropicToolParam = Anthropic.Tool;
-type AnthropicContentBlockParam = Anthropic.ContentBlockParam;
-
-type ThoughtPart = { text: string; signature?: string };
-
-interface ParsedParts {
-  thoughtParts: ThoughtPart[];
-  contentParts: string[];
-  functionCalls: FunctionCall[];
-  functionResponses: FunctionResponse[];
-}
-
-export class AnthropicContentConverter {
-  private model: string;
-  private schemaCompliance: SchemaComplianceMode;
-
-  constructor(model: string, schemaCompliance: SchemaComplianceMode = 'auto') {
-    this.model = model;
-    this.schemaCompliance = schemaCompliance;
-  }
-
-  convertGeminiRequestToAnthropic(request: GenerateContentParameters): {
-    system?: string;
-    messages: AnthropicMessageParam[];
-  } {
-    const messages: AnthropicMessageParam[] = [];
-
-    const system = this.extractTextFromContentUnion(
-      request.config?.systemInstruction,
-    );
-
-    this.processContents(request.contents, messages);
-
-    return {
-      system: system || undefined,
-      messages,
-    };
-  }
-
-  async convertGeminiToolsToAnthropic(
-    geminiTools: ToolListUnion,
-  ): Promise<AnthropicToolParam[]> {
-    const tools: AnthropicToolParam[] = [];
-
-    for (const tool of geminiTools) {
-      let actualTool: Tool;
-
-      if ('tool' in tool) {
-        actualTool = await (tool as CallableTool).tool();
-      } else {
-        actualTool = tool as Tool;
-      }
-
-      if (!actualTool.functionDeclarations) {
-        continue;
-      }
-
-      for (const func of actualTool.functionDeclarations) {
-        if (!func.name) continue;
-
-        let inputSchema: Record<string, unknown> | undefined;
-        if (func.parametersJsonSchema) {
-          inputSchema = {
-            ...(func.parametersJsonSchema as Record<string, unknown>),
-          };
-        } else if (func.parameters) {
-          inputSchema = func.parameters as Record<string, unknown>;
-        }
-
-        if (!inputSchema) {
-          inputSchema = { type: 'object', properties: {} };
-        }
-
-        inputSchema = convertSchema(inputSchema, this.schemaCompliance);
-        if (typeof inputSchema['type'] !== 'string') {
-          inputSchema['type'] = 'object';
-        }
-
-        tools.push({
-          name: func.name,
-          description: func.description,
-          input_schema: inputSchema as Anthropic.Tool.InputSchema,
-        });
-      }
-    }
-
-    return tools;
-  }
-
-  convertAnthropicResponseToGemini(
-    response: Anthropic.Message,
-  ): GenerateContentResponse {
-    const geminiResponse = new GenerateContentResponse();
-    const parts: Part[] = [];
-
-    for (const block of response.content || []) {
-      const blockType = String((block as { type?: string })['type'] || '');
-      if (blockType === 'text') {
-        const text =
-          typeof (block as { text?: string }).text === 'string'
-            ? (block as { text?: string }).text
-            : '';
-        if (text) {
-          parts.push({ text });
-        }
-      } else if (blockType === 'tool_use') {
-        const toolUse = block as {
-          id?: string;
-          name?: string;
-          input?: unknown;
-        };
-        parts.push({
-          functionCall: {
-            id: typeof toolUse.id === 'string' ? toolUse.id : undefined,
-            name: typeof toolUse.name === 'string' ? toolUse.name : undefined,
-            args: this.safeInputToArgs(toolUse.input),
-          },
-        });
-      } else if (blockType === 'thinking') {
-        const thinking =
-          typeof (block as { thinking?: string }).thinking === 'string'
-            ? (block as { thinking?: string }).thinking
-            : '';
-        const signature =
-          typeof (block as { signature?: string }).signature === 'string'
-            ? (block as { signature?: string }).signature
-            : '';
-        if (thinking || signature) {
-          const thoughtPart: Part = {
-            text: thinking,
-            thought: true,
-            thoughtSignature: signature,
-          };
-          parts.push(thoughtPart);
-        }
-      } else if (blockType === 'redacted_thinking') {
-        parts.push({ text: '', thought: true });
-      }
-    }
-
-    const candidate: Candidate = {
-      content: {
-        parts,
-        role: 'model' as const,
-      },
-      index: 0,
-      safetyRatings: [],
-    };
-
-    const finishReason = this.mapAnthropicFinishReasonToGemini(
-      response.stop_reason,
-    );
-    if (finishReason) {
-      candidate.finishReason = finishReason;
-    }
-
-    geminiResponse.candidates = [candidate];
-    geminiResponse.responseId = response.id;
-    geminiResponse.createTime = Date.now().toString();
-    geminiResponse.modelVersion = response.model || this.model;
-    geminiResponse.promptFeedback = { safetyRatings: [] };
-
-    if (response.usage) {
-      const promptTokens = response.usage.input_tokens || 0;
-      const completionTokens = response.usage.output_tokens || 0;
-      geminiResponse.usageMetadata = {
-        promptTokenCount: promptTokens,
-        candidatesTokenCount: completionTokens,
-        totalTokenCount: promptTokens + completionTokens,
-      };
-    }
-
-    return geminiResponse;
-  }
-
-  private processContents(
-    contents: ContentListUnion,
-    messages: AnthropicMessageParam[],
-  ): void {
-    if (Array.isArray(contents)) {
-      for (const content of contents) {
-        this.processContent(content, messages);
-      }
-    } else if (contents) {
-      this.processContent(contents, messages);
-    }
-  }
-
-  private processContent(
-    content: ContentUnion | PartUnion,
-    messages: AnthropicMessageParam[],
-  ): void {
-    if (typeof content === 'string') {
-      messages.push({
-        role: 'user',
-        content: [{ type: 'text', text: content }],
-      });
-      return;
-    }
-
-    if (!this.isContentObject(content)) return;
-
-    const parsed = this.parseParts(content.parts || []);
-
-    if (parsed.functionResponses.length > 0) {
-      for (const response of parsed.functionResponses) {
-        messages.push({
-          role: 'user',
-          content: [
-            {
-              type: 'tool_result',
-              tool_use_id: response.id || '',
-              content: this.extractFunctionResponseContent(response.response),
-            },
-          ],
-        });
-      }
-      return;
-    }
-
-    if (content.role === 'model' && parsed.functionCalls.length > 0) {
-      const thinkingBlocks: AnthropicContentBlockParam[] =
-        parsed.thoughtParts.map((part) => {
-          const thinkingBlock: unknown = {
-            type: 'thinking',
-            thinking: part.text,
-          };
-          if (part.signature) {
-            (thinkingBlock as { signature?: string }).signature =
-              part.signature;
-          }
-          return thinkingBlock as AnthropicContentBlockParam;
-        });
-      const toolUses: AnthropicContentBlockParam[] = parsed.functionCalls.map(
-        (call, index) => ({
-          type: 'tool_use',
-          id: call.id || `tool_${index}`,
-          name: call.name || '',
-          input: (call.args as Record<string, unknown>) || {},
-        }),
-      );
-
-      const textBlocks: AnthropicContentBlockParam[] = parsed.contentParts.map(
-        (text) => ({
-          type: 'text' as const,
-          text,
-        }),
-      );
-
-      messages.push({
-        role: 'assistant',
-        content: [...thinkingBlocks, ...textBlocks, ...toolUses],
-      });
-      return;
-    }
-
-    const role = content.role === 'model' ? 'assistant' : 'user';
-    const thinkingBlocks: AnthropicContentBlockParam[] =
-      role === 'assistant'
-        ? parsed.thoughtParts.map((part) => {
-            const thinkingBlock: unknown = {
-              type: 'thinking',
-              thinking: part.text,
-            };
-            if (part.signature) {
-              (thinkingBlock as { signature?: string }).signature =
-                part.signature;
-            }
-            return thinkingBlock as AnthropicContentBlockParam;
-          })
-        : [];
-    const textBlocks: AnthropicContentBlockParam[] = [
-      ...thinkingBlocks,
-      ...parsed.contentParts.map((text) => ({
-        type: 'text' as const,
-        text,
-      })),
-    ];
-    if (textBlocks.length > 0) {
-      messages.push({ role, content: textBlocks });
-    }
-  }
-
-  private parseParts(parts: Part[]): ParsedParts {
-    const thoughtParts: ThoughtPart[] = [];
-    const contentParts: string[] = [];
-    const functionCalls: FunctionCall[] = [];
-    const functionResponses: FunctionResponse[] = [];
-
-    for (const part of parts) {
-      if (typeof part === 'string') {
-        contentParts.push(part);
-      } else if (
-        'text' in part &&
-        part.text &&
-        !('thought' in part && part.thought)
-      ) {
-        contentParts.push(part.text);
-      } else if ('text' in part && 'thought' in part && part.thought) {
-        thoughtParts.push({
-          text: part.text || '',
-          signature:
-            'thoughtSignature' in part &&
-            typeof part.thoughtSignature === 'string'
-              ? part.thoughtSignature
-              : undefined,
-        });
-      } else if ('functionCall' in part && part.functionCall) {
-        functionCalls.push(part.functionCall);
-      } else if ('functionResponse' in part && part.functionResponse) {
-        functionResponses.push(part.functionResponse);
-      }
-    }
-
-    return {
-      thoughtParts,
-      contentParts,
-      functionCalls,
-      functionResponses,
-    };
-  }
-
-  private extractTextFromContentUnion(contentUnion: unknown): string {
-    if (typeof contentUnion === 'string') {
-      return contentUnion;
-    }
-
-    if (Array.isArray(contentUnion)) {
-      return contentUnion
-        .map((item) => this.extractTextFromContentUnion(item))
-        .filter(Boolean)
-        .join('\n');
-    }
-
-    if (typeof contentUnion === 'object' && contentUnion !== null) {
-      if ('parts' in contentUnion) {
-        const content = contentUnion as Content;
-        return (
-          content.parts
-            ?.map((part: Part) => {
-              if (typeof part === 'string') return part;
-              if ('text' in part) return part.text || '';
-              return '';
-            })
-            .filter(Boolean)
-            .join('\n') || ''
-        );
-      }
-    }
-
-    return '';
-  }
-
-  private extractFunctionResponseContent(response: unknown): string {
-    if (response === null || response === undefined) {
-      return '';
-    }
-
-    if (typeof response === 'string') {
-      return response;
-    }
-
-    if (typeof response === 'object') {
-      const responseObject = response as Record<string, unknown>;
-      const output = responseObject['output'];
-      if (typeof output === 'string') {
-        return output;
-      }
-
-      const error = responseObject['error'];
-      if (typeof error === 'string') {
-        return error;
-      }
-    }
-
-    try {
-      const serialized = JSON.stringify(response);
-      return serialized ?? String(response);
-    } catch {
-      return String(response);
-    }
-  }
-
-  private safeInputToArgs(input: unknown): Record<string, unknown> {
-    if (input && typeof input === 'object') {
-      return input as Record<string, unknown>;
-    }
-    if (typeof input === 'string') {
-      return safeJsonParse(input, {});
-    }
-    return {};
-  }
-
-  mapAnthropicFinishReasonToGemini(
-    reason?: string | null,
-  ): FinishReason | undefined {
-    if (!reason) return undefined;
-    const mapping: Record<string, FinishReason> = {
-      end_turn: FinishReason.STOP,
-      stop_sequence: FinishReason.STOP,
-      tool_use: FinishReason.STOP,
-      max_tokens: FinishReason.MAX_TOKENS,
-      content_filter: FinishReason.SAFETY,
-    };
-    return mapping[reason] || FinishReason.FINISH_REASON_UNSPECIFIED;
-  }
-
-  private isContentObject(
-    content: unknown,
-  ): content is { role: string; parts: Part[] } {
-    return (
-      typeof content === 'object' &&
-      content !== null &&
-      'role' in content &&
-      'parts' in content &&
-      Array.isArray((content as Record<string, unknown>)['parts'])
-    );
-  }
-}
--- a/packages/core/src/core/anthropicContentGenerator/index.ts
+++ b/packages/core/src/core/anthropicContentGenerator/index.ts
@@ -1,21 +0,0 @@
-/**
- * @license
- * Copyright 2025 Qwen
- * SPDX-License-Identifier: Apache-2.0
- */
-
-import type {
-  ContentGenerator,
-  ContentGeneratorConfig,
-} from '../contentGenerator.js';
-import type { Config } from '../../config/config.js';
-import { AnthropicContentGenerator } from './anthropicContentGenerator.js';
-
-export { AnthropicContentGenerator } from './anthropicContentGenerator.js';
-
-export function createAnthropicContentGenerator(
-  contentGeneratorConfig: ContentGeneratorConfig,
-  cliConfig: Config,
-): ContentGenerator {
-  return new AnthropicContentGenerator(contentGeneratorConfig, cliConfig);
-}
--- a/packages/core/src/core/contentGenerator.test.ts
+++ b/packages/core/src/core/contentGenerator.test.ts
@@ -8,7 +8,7 @@ import { describe, it, expect, vi } from 'vitest';
 import { createContentGenerator, AuthType } from './contentGenerator.js';
 import { GoogleGenAI } from '@google/genai';
 import type { Config } from '../config/config.js';
-import { LoggingContentGenerator } from './loggingContentGenerator/index.js';
+import { LoggingContentGenerator } from './geminiContentGenerator/loggingContentGenerator.js';

 vi.mock('@google/genai');

--- a/packages/core/src/core/contentGenerator.ts
+++ b/packages/core/src/core/contentGenerator.ts
@@ -14,7 +14,6 @@ import type {
 } from '@google/genai';
 import { DEFAULT_QWEN_MODEL } from '../config/models.js';
 import type { Config } from '../config/config.js';
-import { LoggingContentGenerator } from './loggingContentGenerator/index.js';

 /**
 * Interface abstracting the core functionalities for generating content and counting tokens.
@@ -38,11 +37,10 @@ export interface ContentGenerator {
 }

 export enum AuthType {
+  USE_GEMINI = 'gemini-api-key',
+  USE_VERTEX_AI = 'vertex-ai',
  USE_OPENAI = 'openai',
  QWEN_OAUTH = 'qwen-oauth',
-  USE_GEMINI = 'gemini',
-  USE_VERTEX_AI = 'vertex-ai',
-  USE_ANTHROPIC = 'anthropic',
 }

 export type ContentGeneratorConfig = {
@@ -65,12 +63,9 @@ export type ContentGeneratorConfig = {
    temperature?: number;
    max_tokens?: number;
  };
-  reasoning?:
-    | false
-    | {
-        effort?: 'low' | 'medium' | 'high';
-        budget_tokens?: number;
-      };
+  reasoning?: {
+    effort?: 'low' | 'medium' | 'high';
+  };
  proxy?: string | undefined;
  userAgent?: string;
  // Schema compliance mode for tool definitions
@@ -82,7 +77,7 @@ export function createContentGeneratorConfig(
  authType: AuthType | undefined,
  generationConfig?: Partial<ContentGeneratorConfig>,
 ): ContentGeneratorConfig {
-  let newContentGeneratorConfig: Partial<ContentGeneratorConfig> = {
+  const newContentGeneratorConfig: Partial<ContentGeneratorConfig> = {
    ...(generationConfig || {}),
    authType,
    proxy: config?.getProxy(),
@@ -99,16 +94,8 @@ export function createContentGeneratorConfig(
  }

  if (authType === AuthType.USE_OPENAI) {
-    newContentGeneratorConfig = {
-      ...newContentGeneratorConfig,
-      apiKey: newContentGeneratorConfig.apiKey || process.env['OPENAI_API_KEY'],
-      baseUrl:
-        newContentGeneratorConfig.baseUrl || process.env['OPENAI_BASE_URL'],
-      model: newContentGeneratorConfig.model || process.env['OPENAI_MODEL'],
-    };
-
    if (!newContentGeneratorConfig.apiKey) {
-      throw new Error('OPENAI_API_KEY environment variable not found.');
+      throw new Error('OpenAI API key is required');
    }

    return {
@@ -117,62 +104,10 @@ export function createContentGeneratorConfig(
    } as ContentGeneratorConfig;
  }

-  if (authType === AuthType.USE_ANTHROPIC) {
-    newContentGeneratorConfig = {
-      ...newContentGeneratorConfig,
-      apiKey:
-        newContentGeneratorConfig.apiKey || process.env['ANTHROPIC_API_KEY'],
-      baseUrl:
-        newContentGeneratorConfig.baseUrl || process.env['ANTHROPIC_BASE_URL'],
-      model: newContentGeneratorConfig.model || process.env['ANTHROPIC_MODEL'],
-    };
-
-    if (!newContentGeneratorConfig.apiKey) {
-      throw new Error('ANTHROPIC_API_KEY environment variable not found.');
-    }
-
-    if (!newContentGeneratorConfig.baseUrl) {
-      throw new Error('ANTHROPIC_BASE_URL environment variable not found.');
-    }
-
-    if (!newContentGeneratorConfig.model) {
-      throw new Error('ANTHROPIC_MODEL environment variable not found.');
-    }
-  }
-
-  if (authType === AuthType.USE_GEMINI) {
-    newContentGeneratorConfig = {
-      ...newContentGeneratorConfig,
-      apiKey: newContentGeneratorConfig.apiKey || process.env['GEMINI_API_KEY'],
-      model: newContentGeneratorConfig.model || process.env['GEMINI_MODEL'],
-    };
-
-    if (!newContentGeneratorConfig.apiKey) {
-      throw new Error('GEMINI_API_KEY environment variable not found.');
-    }
-
-    if (!newContentGeneratorConfig.model) {
-      throw new Error('GEMINI_MODEL environment variable not found.');
-    }
-  }
-
-  if (authType === AuthType.USE_VERTEX_AI) {
-    newContentGeneratorConfig = {
-      ...newContentGeneratorConfig,
-      apiKey: newContentGeneratorConfig.apiKey || process.env['GOOGLE_API_KEY'],
-      model: newContentGeneratorConfig.model || process.env['GOOGLE_MODEL'],
-    };
-
-    if (!newContentGeneratorConfig.apiKey) {
-      throw new Error('GOOGLE_API_KEY environment variable not found.');
-    }
-
-    if (!newContentGeneratorConfig.model) {
-      throw new Error('GOOGLE_MODEL environment variable not found.');
-    }
-  }
-
-  return newContentGeneratorConfig as ContentGeneratorConfig;
+  return {
+    ...newContentGeneratorConfig,
+    model: newContentGeneratorConfig?.model || DEFAULT_QWEN_MODEL,
+  } as ContentGeneratorConfig;
 }

 export async function createContentGenerator(
@@ -180,9 +115,19 @@ export async function createContentGenerator(
  gcConfig: Config,
  isInitialAuth?: boolean,
 ): Promise<ContentGenerator> {
+  if (
+    config.authType === AuthType.USE_GEMINI ||
+    config.authType === AuthType.USE_VERTEX_AI
+  ) {
+    const { createGeminiContentGenerator } = await import(
+      './geminiContentGenerator/index.js'
+    );
+    return createGeminiContentGenerator(config, gcConfig);
+  }
+
  if (config.authType === AuthType.USE_OPENAI) {
    if (!config.apiKey) {
-      throw new Error('OPENAI_API_KEY environment variable not found.');
+      throw new Error('OpenAI API key is required');
    }

    // Import OpenAIContentGenerator dynamically to avoid circular dependencies
@@ -191,8 +136,7 @@ export async function createContentGenerator(
    );

    // Always use OpenAIContentGenerator, logging is controlled by enableOpenAILogging flag
-    const generator = createOpenAIContentGenerator(config, gcConfig);
-    return new LoggingContentGenerator(generator, gcConfig);
+    return createOpenAIContentGenerator(config, gcConfig);
  }

  if (config.authType === AuthType.QWEN_OAUTH) {
@@ -213,8 +157,7 @@ export async function createContentGenerator(
      );

      // Create the content generator with dynamic token management
-      const generator = new QwenContentGenerator(qwenClient, config, gcConfig);
-      return new LoggingContentGenerator(generator, gcConfig);
+      return new QwenContentGenerator(qwenClient, config, gcConfig);
    } catch (error) {
      throw new Error(
        `${error instanceof Error ? error.message : String(error)}`,
@@ -222,30 +165,6 @@ export async function createContentGenerator(
    }
  }

-  if (config.authType === AuthType.USE_ANTHROPIC) {
-    if (!config.apiKey) {
-      throw new Error('ANTHROPIC_API_KEY environment variable not found.');
-    }
-
-    const { createAnthropicContentGenerator } = await import(
-      './anthropicContentGenerator/index.js'
-    );
-
-    const generator = createAnthropicContentGenerator(config, gcConfig);
-    return new LoggingContentGenerator(generator, gcConfig);
-  }
-
-  if (
-    config.authType === AuthType.USE_GEMINI ||
-    config.authType === AuthType.USE_VERTEX_AI
-  ) {
-    const { createGeminiContentGenerator } = await import(
-      './geminiContentGenerator/index.js'
-    );
-    const generator = createGeminiContentGenerator(config, gcConfig);
-    return new LoggingContentGenerator(generator, gcConfig);
-  }
-
  throw new Error(
    `Error creating contentGenerator: Unsupported authType: ${config.authType}`,
  );
--- a/packages/core/src/core/geminiChat.test.ts
+++ b/packages/core/src/core/geminiChat.test.ts
@@ -720,6 +720,66 @@ describe('GeminiChat', () => {
      );
    });

+    it('should handle summarized thinking by conditionally including thoughts in history', async () => {
+      // Case 1: useSummarizedThinking is true -> thoughts NOT in history
+      vi.mocked(mockContentGenerator.useSummarizedThinking).mockReturnValue(
+        true,
+      );
+      const stream1 = (async function* () {
+        yield {
+          candidates: [
+            {
+              content: {
+                role: 'model',
+                parts: [{ thought: true, text: 'T1' }, { text: 'A1' }],
+              },
+              finishReason: 'STOP',
+            },
+          ],
+        } as unknown as GenerateContentResponse;
+      })();
+      vi.mocked(mockContentGenerator.generateContentStream).mockResolvedValue(
+        stream1,
+      );
+
+      const res1 = await chat.sendMessageStream('m1', { message: 'h1' }, 'p1');
+      for await (const _ of res1);
+
+      const history1 = chat.getHistory();
+      expect(history1[1].parts).toEqual([{ text: 'A1' }]);
+
+      // Case 2: useSummarizedThinking is false -> thoughts ARE in history
+      chat.clearHistory();
+      vi.mocked(mockContentGenerator.useSummarizedThinking).mockReturnValue(
+        false,
+      );
+      const stream2 = (async function* () {
+        yield {
+          candidates: [
+            {
+              content: {
+                role: 'model',
+                parts: [{ thought: true, text: 'T2' }, { text: 'A2' }],
+              },
+              finishReason: 'STOP',
+            },
+          ],
+        } as unknown as GenerateContentResponse;
+      })();
+      vi.mocked(mockContentGenerator.generateContentStream).mockResolvedValue(
+        stream2,
+      );
+
+      const res2 = await chat.sendMessageStream('m1', { message: 'h1' }, 'p2');
+      for await (const _ of res2);
+
+      const history2 = chat.getHistory();
+      expect(history2[1].parts).toEqual([
+        { text: 'T2', thought: true },
+        { text: 'A2' },
+      ]);
+    });
+
    it('should keep parts with thoughtSignature when consolidating history', async () => {
      const stream = (async function* () {
        yield {
--- a/packages/core/src/core/geminiChat.ts
+++ b/packages/core/src/core/geminiChat.ts
@@ -559,25 +559,14 @@ export class GeminiChat {
      yield chunk; // Yield every chunk to the UI immediately.
    }

-    let thoughtContentPart: Part | undefined;
-    const thoughtText = allModelParts
-      .filter((part) => part.thought)
-      .map((part) => part.text)
-      .join('')
-      .trim();
-
-    if (thoughtText !== '') {
-      thoughtContentPart = {
-        text: thoughtText,
-        thought: true,
-      };
-
-      const thoughtSignature = allModelParts.filter(
-        (part) => part.thoughtSignature && part.thought,
-      )?.[0]?.thoughtSignature;
-      if (thoughtContentPart && thoughtSignature) {
-        thoughtContentPart.thoughtSignature = thoughtSignature;
-      }
+    let thoughtText = '';
+    // Only include thoughts if not using summarized thinking.
+    if (!this.config.getContentGenerator().useSummarizedThinking()) {
+      thoughtText = allModelParts
+        .filter((part) => part.thought)
+        .map((part) => part.text)
+        .join('')
+        .trim();
    }

    const contentParts = allModelParts.filter((part) => !part.thought);
@@ -603,11 +592,11 @@ export class GeminiChat {
      .trim();

    // Record assistant turn with raw Content and metadata
-    if (thoughtContentPart || contentText || hasToolCall || usageMetadata) {
+    if (thoughtText || contentText || hasToolCall || usageMetadata) {
      this.chatRecordingService?.recordAssistantTurn({
        model,
        message: [
-          ...(thoughtContentPart ? [thoughtContentPart] : []),
+          ...(thoughtText ? [{ text: thoughtText, thought: true }] : []),
          ...(contentText ? [{ text: contentText }] : []),
          ...(hasToolCall
            ? contentParts
@@ -643,7 +632,7 @@ export class GeminiChat {
    this.history.push({
      role: 'model',
      parts: [
-        ...(thoughtContentPart ? [thoughtContentPart] : []),
+        ...(thoughtText ? [{ text: thoughtText, thought: true }] : []),
        ...consolidatedHistoryParts,
      ],
    });
--- a/packages/core/src/core/geminiContentGenerator/geminiContentGenerator.ts
+++ b/packages/core/src/core/geminiContentGenerator/geminiContentGenerator.ts
@@ -39,7 +39,7 @@ export class GeminiContentGenerator implements ContentGenerator {
    this.contentGeneratorConfig = contentGeneratorConfig;
  }

-  private buildGenerateContentConfig(
+  private buildSamplingParameters(
    request: GenerateContentParameters,
  ): GenerateContentConfig {
    const configSamplingParams = this.contentGeneratorConfig?.samplingParams;
@@ -84,7 +84,17 @@ export class GeminiContentGenerator implements ContentGenerator {
        'frequencyPenalty',
      ),
      thinkingConfig: getParameterValue(
-        this.buildThinkingConfig(),
+        this.contentGeneratorConfig?.reasoning
+          ? {
+              includeThoughts: true,
+              thinkingLevel: (this.contentGeneratorConfig.reasoning.effort ===
+              'low'
+                ? 'LOW'
+                : this.contentGeneratorConfig.reasoning.effort === 'high'
+                  ? 'HIGH'
+                  : 'THINKING_LEVEL_UNSPECIFIED') as ThinkingLevel,
+            }
+          : undefined,
        'thinkingConfig',
        {
          includeThoughts: true,
@@ -94,40 +104,13 @@ export class GeminiContentGenerator implements ContentGenerator {
    };
  }

-  private buildThinkingConfig():
-    | { includeThoughts: boolean; thinkingLevel?: ThinkingLevel }
-    | undefined {
-    const reasoning = this.contentGeneratorConfig?.reasoning;
-
-    if (reasoning === false) {
-      return { includeThoughts: false };
-    }
-
-    if (reasoning) {
-      const thinkingLevel = (
-        reasoning.effort === 'low'
-          ? 'LOW'
-          : reasoning.effort === 'high'
-            ? 'HIGH'
-            : 'THINKING_LEVEL_UNSPECIFIED'
-      ) as ThinkingLevel;
-
-      return {
-        includeThoughts: true,
-        thinkingLevel,
-      };
-    }
-
-    return undefined;
-  }
-
  async generateContent(
    request: GenerateContentParameters,
    _userPromptId: string,
  ): Promise<GenerateContentResponse> {
    const finalRequest = {
      ...request,
-      config: this.buildGenerateContentConfig(request),
+      config: this.buildSamplingParameters(request),
    };
    return this.googleGenAI.models.generateContent(finalRequest);
  }
@@ -138,7 +121,7 @@ export class GeminiContentGenerator implements ContentGenerator {
  ): Promise<AsyncGenerator<GenerateContentResponse>> {
    const finalRequest = {
      ...request,
-      config: this.buildGenerateContentConfig(request),
+      config: this.buildSamplingParameters(request),
    };
    return this.googleGenAI.models.generateContentStream(finalRequest);
  }
--- a/packages/core/src/core/geminiContentGenerator/index.test.ts
+++ b/packages/core/src/core/geminiContentGenerator/index.test.ts
@@ -7,6 +7,7 @@
 import { describe, it, expect, vi, beforeEach } from 'vitest';
 import { createGeminiContentGenerator } from './index.js';
 import { GeminiContentGenerator } from './geminiContentGenerator.js';
+import { LoggingContentGenerator } from './loggingContentGenerator.js';
 import type { Config } from '../../config/config.js';
 import { AuthType } from '../contentGenerator.js';

@@ -14,6 +15,10 @@ vi.mock('./geminiContentGenerator.js', () => ({
  GeminiContentGenerator: vi.fn().mockImplementation(() => ({})),
 }));

+vi.mock('./loggingContentGenerator.js', () => ({
+  LoggingContentGenerator: vi.fn().mockImplementation((wrapped) => wrapped),
+}));
+
 describe('createGeminiContentGenerator', () => {
  let mockConfig: Config;

@@ -26,7 +31,7 @@ describe('createGeminiContentGenerator', () => {
    } as unknown as Config;
  });

-  it('should create a GeminiContentGenerator', () => {
+  it('should create a GeminiContentGenerator wrapped in LoggingContentGenerator', () => {
    const config = {
      model: 'gemini-1.5-flash',
      apiKey: 'test-key',
@@ -36,6 +41,7 @@ describe('createGeminiContentGenerator', () => {
    const generator = createGeminiContentGenerator(config, mockConfig);

    expect(GeminiContentGenerator).toHaveBeenCalled();
+    expect(LoggingContentGenerator).toHaveBeenCalled();
    expect(generator).toBeDefined();
  });
 });
--- a/packages/core/src/core/geminiContentGenerator/index.ts
+++ b/packages/core/src/core/geminiContentGenerator/index.ts
@@ -11,8 +11,10 @@ import type {
 } from '../contentGenerator.js';
 import type { Config } from '../../config/config.js';
 import { InstallationManager } from '../../utils/installationManager.js';
+import { LoggingContentGenerator } from './loggingContentGenerator.js';

 export { GeminiContentGenerator } from './geminiContentGenerator.js';
+export { LoggingContentGenerator } from './loggingContentGenerator.js';

 /**
 * Create a Gemini content generator.
@@ -49,5 +51,5 @@ export function createGeminiContentGenerator(
    config,
  );

-  return geminiContentGenerator;
+  return new LoggingContentGenerator(geminiContentGenerator, gcConfig);
 }
--- a/packages/core/src/core/loggingContentGenerator/loggingContentGenerator.ts
+++ b/packages/core/src/core/loggingContentGenerator/loggingContentGenerator.ts
@@ -4,22 +4,20 @@
 * SPDX-License-Identifier: Apache-2.0
 */

-import {
+import type {
+  Content,
+  CountTokensParameters,
+  CountTokensResponse,
+  EmbedContentParameters,
+  EmbedContentResponse,
+  GenerateContentParameters,
+  GenerateContentResponseUsageMetadata,
  GenerateContentResponse,
-  type Content,
-  type CountTokensParameters,
-  type CountTokensResponse,
-  type EmbedContentParameters,
-  type EmbedContentResponse,
-  type GenerateContentParameters,
-  type GenerateContentResponseUsageMetadata,
-  type ContentListUnion,
-  type ContentUnion,
-  type Part,
-  type PartUnion,
-  type FinishReason,
+  ContentListUnion,
+  ContentUnion,
+  Part,
+  PartUnion,
 } from '@google/genai';
-import type OpenAI from 'openai';
 import {
  ApiRequestEvent,
  ApiResponseEvent,
@@ -33,8 +31,6 @@ import {
 } from '../../telemetry/loggers.js';
 import type { ContentGenerator } from '../contentGenerator.js';
 import { isStructuredError } from '../../utils/quotaErrorDetection.js';
-import { OpenAIContentConverter } from '../openaiContentGenerator/converter.js';
-import { OpenAILogger } from '../../utils/openaiLogger.js';

 interface StructuredError {
  status: number;
@@ -44,19 +40,10 @@ interface StructuredError {
 * A decorator that wraps a ContentGenerator to add logging to API calls.
 */
 export class LoggingContentGenerator implements ContentGenerator {
-  private openaiLogger?: OpenAILogger;
-  private schemaCompliance?: 'auto' | 'openapi_30';
-
  constructor(
    private readonly wrapped: ContentGenerator,
    private readonly config: Config,
-  ) {
-    const generatorConfig = this.config.getContentGeneratorConfig();
-    if (generatorConfig?.enableOpenAILogging) {
-      this.openaiLogger = new OpenAILogger(generatorConfig.openAILoggingDir);
-      this.schemaCompliance = generatorConfig.schemaCompliance;
-    }
-  }
+  ) {}

  getWrapped(): ContentGenerator {
    return this.wrapped;
@@ -104,31 +91,21 @@ export class LoggingContentGenerator implements ContentGenerator {
    prompt_id: string,
  ): void {
    const errorMessage = error instanceof Error ? error.message : String(error);
-    const errorType =
-      (error as { type?: string })?.type ||
-      (error instanceof Error ? error.name : 'unknown');
-    const errorResponseId =
-      (error as { requestID?: string; request_id?: string })?.requestID ||
-      (error as { requestID?: string; request_id?: string })?.request_id ||
-      responseId;
-    const errorStatus =
-      (error as { code?: string | number; status?: number })?.code ??
-      (error as { status?: number })?.status ??
-      (isStructuredError(error)
-        ? (error as StructuredError).status
-        : undefined);
+    const errorType = error instanceof Error ? error.name : 'unknown';

    logApiError(
      this.config,
      new ApiErrorEvent(
-        errorResponseId,
+        responseId,
        model,
        errorMessage,
        durationMs,
        prompt_id,
        this.config.getContentGeneratorConfig()?.authType,
        errorType,
-        errorStatus,
+        isStructuredError(error)
+          ? (error as StructuredError).status
+          : undefined,
      ),
    );
  }
@@ -139,7 +116,6 @@ export class LoggingContentGenerator implements ContentGenerator {
  ): Promise<GenerateContentResponse> {
    const startTime = Date.now();
    this.logApiRequest(this.toContents(req.contents), req.model, userPromptId);
-    const openaiRequest = await this.buildOpenAIRequestForLogging(req);
    try {
      const response = await this.wrapped.generateContent(req, userPromptId);
      const durationMs = Date.now() - startTime;
@@ -151,12 +127,10 @@ export class LoggingContentGenerator implements ContentGenerator {
        response.usageMetadata,
        JSON.stringify(response),
      );
-      await this.logOpenAIInteraction(openaiRequest, response);
      return response;
    } catch (error) {
      const durationMs = Date.now() - startTime;
      this._logApiError(undefined, durationMs, error, req.model, userPromptId);
-      await this.logOpenAIInteraction(openaiRequest, undefined, error);
      throw error;
    }
  }
@@ -167,7 +141,6 @@ export class LoggingContentGenerator implements ContentGenerator {
  ): Promise<AsyncGenerator<GenerateContentResponse>> {
    const startTime = Date.now();
    this.logApiRequest(this.toContents(req.contents), req.model, userPromptId);
-    const openaiRequest = await this.buildOpenAIRequestForLogging(req);

    let stream: AsyncGenerator<GenerateContentResponse>;
    try {
@@ -175,7 +148,6 @@ export class LoggingContentGenerator implements ContentGenerator {
    } catch (error) {
      const durationMs = Date.now() - startTime;
      this._logApiError(undefined, durationMs, error, req.model, userPromptId);
-      await this.logOpenAIInteraction(openaiRequest, undefined, error);
      throw error;
    }

@@ -184,7 +156,6 @@ export class LoggingContentGenerator implements ContentGenerator {
      startTime,
      userPromptId,
      req.model,
-      openaiRequest,
    );
  }

@@ -193,7 +164,6 @@ export class LoggingContentGenerator implements ContentGenerator {
    startTime: number,
    userPromptId: string,
    model: string,
-    openaiRequest?: OpenAI.Chat.ChatCompletionCreateParams,
  ): AsyncGenerator<GenerateContentResponse> {
    const responses: GenerateContentResponse[] = [];

@@ -216,9 +186,6 @@ export class LoggingContentGenerator implements ContentGenerator {
        lastUsageMetadata,
        JSON.stringify(responses),
      );
-      const consolidatedResponse =
-        this.consolidateGeminiResponsesForLogging(responses);
-      await this.logOpenAIInteraction(openaiRequest, consolidatedResponse);
    } catch (error) {
      const durationMs = Date.now() - startTime;
      this._logApiError(
@@ -228,182 +195,10 @@ export class LoggingContentGenerator implements ContentGenerator {
        responses[0]?.modelVersion || model,
        userPromptId,
      );
-      await this.logOpenAIInteraction(openaiRequest, undefined, error);
      throw error;
    }
  }

-  private async buildOpenAIRequestForLogging(
-    request: GenerateContentParameters,
-  ): Promise<OpenAI.Chat.ChatCompletionCreateParams | undefined> {
-    if (!this.openaiLogger) {
-      return undefined;
-    }
-
-    const converter = new OpenAIContentConverter(
-      request.model,
-      this.schemaCompliance,
-    );
-    const messages = converter.convertGeminiRequestToOpenAI(request, {
-      cleanOrphanToolCalls: false,
-    });
-
-    const openaiRequest: OpenAI.Chat.ChatCompletionCreateParams = {
-      model: request.model,
-      messages,
-    };
-
-    if (request.config?.tools) {
-      openaiRequest.tools = await converter.convertGeminiToolsToOpenAI(
-        request.config.tools,
-      );
-    }
-
-    if (request.config?.temperature !== undefined) {
-      openaiRequest.temperature = request.config.temperature;
-    }
-    if (request.config?.topP !== undefined) {
-      openaiRequest.top_p = request.config.topP;
-    }
-    if (request.config?.maxOutputTokens !== undefined) {
-      openaiRequest.max_tokens = request.config.maxOutputTokens;
-    }
-    if (request.config?.presencePenalty !== undefined) {
-      openaiRequest.presence_penalty = request.config.presencePenalty;
-    }
-    if (request.config?.frequencyPenalty !== undefined) {
-      openaiRequest.frequency_penalty = request.config.frequencyPenalty;
-    }
-
-    return openaiRequest;
-  }
-
-  private async logOpenAIInteraction(
-    openaiRequest: OpenAI.Chat.ChatCompletionCreateParams | undefined,
-    response?: GenerateContentResponse,
-    error?: unknown,
-  ): Promise<void> {
-    if (!this.openaiLogger || !openaiRequest) {
-      return;
-    }
-
-    const openaiResponse = response
-      ? this.convertGeminiResponseToOpenAIForLogging(response, openaiRequest)
-      : undefined;
-
-    await this.openaiLogger.logInteraction(
-      openaiRequest,
-      openaiResponse,
-      error instanceof Error
-        ? error
-        : error
-          ? new Error(String(error))
-          : undefined,
-    );
-  }
-
-  private convertGeminiResponseToOpenAIForLogging(
-    response: GenerateContentResponse,
-    openaiRequest: OpenAI.Chat.ChatCompletionCreateParams,
-  ): OpenAI.Chat.ChatCompletion {
-    const converter = new OpenAIContentConverter(
-      openaiRequest.model,
-      this.schemaCompliance,
-    );
-
-    return converter.convertGeminiResponseToOpenAI(response);
-  }
-
-  private consolidateGeminiResponsesForLogging(
-    responses: GenerateContentResponse[],
-  ): GenerateContentResponse | undefined {
-    if (responses.length === 0) {
-      return undefined;
-    }
-
-    const consolidated = new GenerateContentResponse();
-    const combinedParts: Part[] = [];
-    const functionCallIndex = new Map<string, number>();
-    let finishReason: FinishReason | undefined;
-    let usageMetadata: GenerateContentResponseUsageMetadata | undefined;
-
-    for (const response of responses) {
-      if (response.usageMetadata) {
-        usageMetadata = response.usageMetadata;
-      }
-
-      const candidate = response.candidates?.[0];
-      if (candidate?.finishReason) {
-        finishReason = candidate.finishReason;
-      }
-
-      const parts = candidate?.content?.parts ?? [];
-      for (const part of parts as Part[]) {
-        if (typeof part === 'string') {
-          combinedParts.push({ text: part });
-          continue;
-        }
-
-        if ('text' in part) {
-          if (part.text) {
-            combinedParts.push({
-              text: part.text,
-              ...(part.thought ? { thought: true } : {}),
-              ...(part.thoughtSignature
-                ? { thoughtSignature: part.thoughtSignature }
-                : {}),
-            });
-          }
-          continue;
-        }
-
-        if ('functionCall' in part && part.functionCall) {
-          const callKey =
-            part.functionCall.id || part.functionCall.name || 'tool_call';
-          const existingIndex = functionCallIndex.get(callKey);
-          const functionPart = { functionCall: part.functionCall };
-          if (existingIndex !== undefined) {
-            combinedParts[existingIndex] = functionPart;
-          } else {
-            functionCallIndex.set(callKey, combinedParts.length);
-            combinedParts.push(functionPart);
-          }
-          continue;
-        }
-
-        if ('functionResponse' in part && part.functionResponse) {
-          combinedParts.push({ functionResponse: part.functionResponse });
-          continue;
-        }
-
-        combinedParts.push(part);
-      }
-    }
-
-    const lastResponse = responses[responses.length - 1];
-    const lastCandidate = lastResponse.candidates?.[0];
-
-    consolidated.responseId = lastResponse.responseId;
-    consolidated.createTime = lastResponse.createTime;
-    consolidated.modelVersion = lastResponse.modelVersion;
-    consolidated.promptFeedback = lastResponse.promptFeedback;
-    consolidated.usageMetadata = usageMetadata;
-
-    consolidated.candidates = [
-      {
-        content: {
-          role: lastCandidate?.content?.role || 'model',
-          parts: combinedParts,
-        },
-        ...(finishReason ? { finishReason } : {}),
-        index: 0,
-        safetyRatings: lastCandidate?.safetyRatings || [],
-      },
-    ];
-
-    return consolidated;
-  }
-
  async countTokens(req: CountTokensParameters): Promise<CountTokensResponse> {
    return this.wrapped.countTokens(req);
  }
--- a/packages/core/src/core/loggingContentGenerator/index.ts
+++ b/packages/core/src/core/loggingContentGenerator/index.ts
@@ -1,7 +0,0 @@
-/**
- * @license
- * Copyright 2025 Qwen
- * SPDX-License-Identifier: Apache-2.0
- */
-
-export { LoggingContentGenerator } from './loggingContentGenerator.js';
--- a/packages/core/src/core/loggingContentGenerator/loggingContentGenerator.test.ts
+++ b/packages/core/src/core/loggingContentGenerator/loggingContentGenerator.test.ts
@@ -1,371 +0,0 @@
-/**
- * @license
- * Copyright 2025 Qwen
- * SPDX-License-Identifier: Apache-2.0
- */
-
-import { describe, it, expect, vi, beforeEach, afterEach } from 'vitest';
-import type {
-  GenerateContentParameters,
-  GenerateContentResponseUsageMetadata,
-} from '@google/genai';
-import { GenerateContentResponse } from '@google/genai';
-import type { Config } from '../../config/config.js';
-import type { ContentGenerator } from '../contentGenerator.js';
-import { LoggingContentGenerator } from './index.js';
-import { OpenAIContentConverter } from '../openaiContentGenerator/converter.js';
-import {
-  logApiRequest,
-  logApiResponse,
-  logApiError,
-} from '../../telemetry/loggers.js';
-import { OpenAILogger } from '../../utils/openaiLogger.js';
-import type OpenAI from 'openai';
-
-vi.mock('../../telemetry/loggers.js', () => ({
-  logApiRequest: vi.fn(),
-  logApiResponse: vi.fn(),
-  logApiError: vi.fn(),
-}));
-
-vi.mock('../../utils/openaiLogger.js', () => ({
-  OpenAILogger: vi.fn().mockImplementation(() => ({
-    logInteraction: vi.fn().mockResolvedValue(undefined),
-  })),
-}));
-
-const convertGeminiRequestToOpenAISpy = vi
-  .spyOn(OpenAIContentConverter.prototype, 'convertGeminiRequestToOpenAI')
-  .mockReturnValue([{ role: 'user', content: 'converted' }]);
-const convertGeminiToolsToOpenAISpy = vi
-  .spyOn(OpenAIContentConverter.prototype, 'convertGeminiToolsToOpenAI')
-  .mockResolvedValue([{ type: 'function', function: { name: 'tool' } }]);
-const convertGeminiResponseToOpenAISpy = vi
-  .spyOn(OpenAIContentConverter.prototype, 'convertGeminiResponseToOpenAI')
-  .mockReturnValue({
-    id: 'openai-response',
-    object: 'chat.completion',
-    created: 123456789,
-    model: 'test-model',
-    choices: [],
-  } as OpenAI.Chat.ChatCompletion);
-
-const createConfig = (overrides: Record<string, unknown> = {}): Config =>
-  ({
-    getContentGeneratorConfig: () => ({
-      authType: 'openai',
-      enableOpenAILogging: false,
-      ...overrides,
-    }),
-  }) as Config;
-
-const createWrappedGenerator = (
-  generateContent: ContentGenerator['generateContent'],
-  generateContentStream: ContentGenerator['generateContentStream'],
-): ContentGenerator =>
-  ({
-    generateContent,
-    generateContentStream,
-    countTokens: vi.fn(),
-    embedContent: vi.fn(),
-    useSummarizedThinking: vi.fn().mockReturnValue(false),
-  }) as ContentGenerator;
-
-const createResponse = (
-  responseId: string,
-  modelVersion: string,
-  parts: Array<Record<string, unknown>>,
-  usageMetadata?: GenerateContentResponseUsageMetadata,
-  finishReason?: string,
-): GenerateContentResponse => {
-  const response = new GenerateContentResponse();
-  response.responseId = responseId;
-  response.modelVersion = modelVersion;
-  response.usageMetadata = usageMetadata;
-  response.candidates = [
-    {
-      content: {
-        role: 'model',
-        parts: parts as never[],
-      },
-      finishReason: finishReason as never,
-      index: 0,
-      safetyRatings: [],
-    },
-  ];
-  return response;
-};
-
-describe('LoggingContentGenerator', () => {
-  beforeEach(() => {
-    vi.clearAllMocks();
-  });
-
-  afterEach(() => {
-    convertGeminiRequestToOpenAISpy.mockClear();
-    convertGeminiToolsToOpenAISpy.mockClear();
-    convertGeminiResponseToOpenAISpy.mockClear();
-  });
-
-  it('logs request/response, normalizes thought parts, and logs OpenAI interaction', async () => {
-    const wrapped = createWrappedGenerator(
-      vi.fn().mockResolvedValue(
-        createResponse(
-          'resp-1',
-          'model-v2',
-          [{ text: 'ok' }],
-          {
-            promptTokenCount: 3,
-            candidatesTokenCount: 5,
-            totalTokenCount: 8,
-          },
-          'STOP',
-        ),
-      ),
-      vi.fn(),
-    );
-    const generator = new LoggingContentGenerator(
-      wrapped,
-      createConfig({
-        enableOpenAILogging: true,
-        openAILoggingDir: 'logs',
-        schemaCompliance: 'openapi_30',
-      }),
-    );
-
-    const request = {
-      model: 'test-model',
-      contents: [
-        {
-          role: 'user',
-          parts: [
-            { text: 'Hello', thought: 'internal' },
-            {
-              functionCall: { id: 'call-1', name: 'tool', args: '{}' },
-              thought: 'strip-me',
-            },
-            null,
-          ],
-        },
-      ],
-      config: {
-        temperature: 0.3,
-        topP: 0.9,
-        maxOutputTokens: 256,
-        presencePenalty: 0.2,
-        frequencyPenalty: 0.1,
-        tools: [
-          {
-            functionDeclarations: [
-              { name: 'tool', description: 'desc', parameters: {} },
-            ],
-          },
-        ],
-      },
-    } as unknown as GenerateContentParameters;
-
-    const response = await generator.generateContent(request, 'prompt-1');
-
-    expect(response.responseId).toBe('resp-1');
-    expect(logApiRequest).toHaveBeenCalledTimes(1);
-    const [, requestEvent] = vi.mocked(logApiRequest).mock.calls[0];
-    const loggedContents = JSON.parse(requestEvent.request_text || '[]');
-    expect(loggedContents[0].parts[0]).toEqual({
-      text: 'Hello\n[Thought: internal]',
-    });
-    expect(loggedContents[0].parts[1]).toEqual({
-      functionCall: { id: 'call-1', name: 'tool', args: '{}' },
-    });
-
-    expect(logApiResponse).toHaveBeenCalledTimes(1);
-    const [, responseEvent] = vi.mocked(logApiResponse).mock.calls[0];
-    expect(responseEvent.response_id).toBe('resp-1');
-    expect(responseEvent.model).toBe('model-v2');
-    expect(responseEvent.prompt_id).toBe('prompt-1');
-    expect(responseEvent.input_token_count).toBe(3);
-
-    expect(convertGeminiRequestToOpenAISpy).toHaveBeenCalledTimes(1);
-    expect(convertGeminiToolsToOpenAISpy).toHaveBeenCalledTimes(1);
-    expect(convertGeminiResponseToOpenAISpy).toHaveBeenCalledTimes(1);
-
-    const openaiLoggerInstance = vi.mocked(OpenAILogger).mock.results[0]
-      ?.value as { logInteraction: ReturnType<typeof vi.fn> };
-    expect(openaiLoggerInstance.logInteraction).toHaveBeenCalledTimes(1);
-    const [openaiRequest, openaiResponse, openaiError] =
-      openaiLoggerInstance.logInteraction.mock.calls[0];
-    expect(openaiRequest).toEqual(
-      expect.objectContaining({
-        model: 'test-model',
-        messages: [{ role: 'user', content: 'converted' }],
-        tools: [{ type: 'function', function: { name: 'tool' } }],
-        temperature: 0.3,
-        top_p: 0.9,
-        max_tokens: 256,
-        presence_penalty: 0.2,
-        frequency_penalty: 0.1,
-      }),
-    );
-    expect(openaiResponse).toEqual({
-      id: 'openai-response',
-      object: 'chat.completion',
-      created: 123456789,
-      model: 'test-model',
-      choices: [],
-    });
-    expect(openaiError).toBeUndefined();
-  });
-
-  it('logs errors with status code and request id, then rethrows', async () => {
-    const error = Object.assign(new Error('boom'), {
-      code: 429,
-      request_id: 'req-99',
-      type: 'rate_limit',
-    });
-    const wrapped = createWrappedGenerator(
-      vi.fn().mockRejectedValue(error),
-      vi.fn(),
-    );
-    const generator = new LoggingContentGenerator(
-      wrapped,
-      createConfig({ enableOpenAILogging: true }),
-    );
-
-    const request = {
-      model: 'test-model',
-      contents: 'Hello',
-    } as unknown as GenerateContentParameters;
-
-    await expect(
-      generator.generateContent(request, 'prompt-2'),
-    ).rejects.toThrow('boom');
-
-    expect(logApiError).toHaveBeenCalledTimes(1);
-    const [, errorEvent] = vi.mocked(logApiError).mock.calls[0];
-    expect(errorEvent.response_id).toBe('req-99');
-    expect(errorEvent.status_code).toBe(429);
-    expect(errorEvent.error_type).toBe('rate_limit');
-    expect(errorEvent.prompt_id).toBe('prompt-2');
-
-    const openaiLoggerInstance = vi.mocked(OpenAILogger).mock.results[0]
-      ?.value as { logInteraction: ReturnType<typeof vi.fn> };
-    const [, , loggedError] = openaiLoggerInstance.logInteraction.mock.calls[0];
-    expect(loggedError).toBeInstanceOf(Error);
-    expect((loggedError as Error).message).toBe('boom');
-  });
-
-  it('logs streaming responses and consolidates tool calls', async () => {
-    const usage1 = {
-      promptTokenCount: 1,
-    } as GenerateContentResponseUsageMetadata;
-    const usage2 = {
-      promptTokenCount: 2,
-      candidatesTokenCount: 4,
-      totalTokenCount: 6,
-    } as GenerateContentResponseUsageMetadata;
-
-    const response1 = createResponse(
-      'resp-1',
-      'model-stream',
-      [
-        { text: 'Hello' },
-        { functionCall: { id: 'call-1', name: 'tool', args: '{}' } },
-      ],
-      usage1,
-    );
-    const response2 = createResponse(
-      'resp-2',
-      'model-stream',
-      [
-        { text: ' world' },
-        { functionCall: { id: 'call-1', name: 'tool', args: '{"x":1}' } },
-        { functionResponse: { name: 'tool', response: { output: 'ok' } } },
-      ],
-      usage2,
-      'STOP',
-    );
-
-    const wrapped = createWrappedGenerator(
-      vi.fn(),
-      vi.fn().mockResolvedValue(
-        (async function* () {
-          yield response1;
-          yield response2;
-        })(),
-      ),
-    );
-    const generator = new LoggingContentGenerator(
-      wrapped,
-      createConfig({ enableOpenAILogging: true }),
-    );
-
-    const request = {
-      model: 'test-model',
-      contents: 'Hello',
-    } as unknown as GenerateContentParameters;
-
-    const stream = await generator.generateContentStream(request, 'prompt-3');
-    const seen: GenerateContentResponse[] = [];
-    for await (const item of stream) {
-      seen.push(item);
-    }
-    expect(seen).toHaveLength(2);
-
-    expect(logApiResponse).toHaveBeenCalledTimes(1);
-    const [, responseEvent] = vi.mocked(logApiResponse).mock.calls[0];
-    expect(responseEvent.response_id).toBe('resp-1');
-    expect(responseEvent.input_token_count).toBe(2);
-
-    expect(convertGeminiResponseToOpenAISpy).toHaveBeenCalledTimes(1);
-    const [consolidatedResponse] =
-      convertGeminiResponseToOpenAISpy.mock.calls[0];
-    const consolidatedParts =
-      consolidatedResponse.candidates?.[0]?.content?.parts || [];
-    expect(consolidatedParts).toEqual([
-      { text: 'Hello' },
-      { functionCall: { id: 'call-1', name: 'tool', args: '{"x":1}' } },
-      { text: ' world' },
-      { functionResponse: { name: 'tool', response: { output: 'ok' } } },
-    ]);
-    expect(consolidatedResponse.usageMetadata).toBe(usage2);
-    expect(consolidatedResponse.responseId).toBe('resp-2');
-    expect(consolidatedResponse.candidates?.[0]?.finishReason).toBe('STOP');
-  });
-
-  it('logs stream errors and skips response logging', async () => {
-    const response1 = createResponse('resp-1', 'model-stream', [
-      { text: 'partial' },
-    ]);
-    const streamError = new Error('stream-fail');
-    const wrapped = createWrappedGenerator(
-      vi.fn(),
-      vi.fn().mockResolvedValue(
-        (async function* () {
-          yield response1;
-          throw streamError;
-        })(),
-      ),
-    );
-    const generator = new LoggingContentGenerator(
-      wrapped,
-      createConfig({ enableOpenAILogging: true }),
-    );
-
-    const request = {
-      model: 'test-model',
-      contents: 'Hello',
-    } as unknown as GenerateContentParameters;
-
-    const stream = await generator.generateContentStream(request, 'prompt-4');
-    await expect(async () => {
-      for await (const _item of stream) {
-        // Consume stream to trigger error.
-      }
-    }).rejects.toThrow('stream-fail');
-
-    expect(logApiResponse).not.toHaveBeenCalled();
-    expect(logApiError).toHaveBeenCalledTimes(1);
-    const openaiLoggerInstance = vi.mocked(OpenAILogger).mock.results[0]
-      ?.value as { logInteraction: ReturnType<typeof vi.fn> };
-    expect(openaiLoggerInstance.logInteraction).toHaveBeenCalledTimes(1);
-  });
-});
--- a/packages/core/src/core/openaiContentGenerator/converter.ts
+++ b/packages/core/src/core/openaiContentGenerator/converter.ts
@@ -236,9 +236,8 @@ export class OpenAIContentConverter {
   */
  convertGeminiRequestToOpenAI(
    request: GenerateContentParameters,
-    options: { cleanOrphanToolCalls: boolean } = { cleanOrphanToolCalls: true },
  ): OpenAI.Chat.ChatCompletionMessageParam[] {
-    let messages: OpenAI.Chat.ChatCompletionMessageParam[] = [];
+    const messages: OpenAI.Chat.ChatCompletionMessageParam[] = [];

    // Handle system instruction from config
    this.addSystemInstructionMessage(request, messages);
@@ -247,89 +246,11 @@ export class OpenAIContentConverter {
    this.processContents(request.contents, messages);

    // Clean up orphaned tool calls and merge consecutive assistant messages
-    if (options.cleanOrphanToolCalls) {
-      messages = this.cleanOrphanedToolCalls(messages);
-    }
-    messages = this.mergeConsecutiveAssistantMessages(messages);
+    const cleanedMessages = this.cleanOrphanedToolCalls(messages);
+    const mergedMessages =
+      this.mergeConsecutiveAssistantMessages(cleanedMessages);

-    return messages;
-  }
-
-  /**
-   * Convert Gemini response to OpenAI completion format (for logging).
-   */
-  convertGeminiResponseToOpenAI(
-    response: GenerateContentResponse,
-  ): OpenAI.Chat.ChatCompletion {
-    const candidate = response.candidates?.[0];
-    const parts = (candidate?.content?.parts || []) as Part[];
-    const parsedParts = this.parseParts(parts);
-
-    const message: ExtendedCompletionMessage = {
-      role: 'assistant',
-      content: parsedParts.contentParts.join('') || null,
-      refusal: null,
-    };
-
-    const reasoningContent = parsedParts.thoughtParts.join('');
-    if (reasoningContent) {
-      message.reasoning_content = reasoningContent;
-    }
-
-    if (parsedParts.functionCalls.length > 0) {
-      message.tool_calls = parsedParts.functionCalls.map((call, index) => ({
-        id: call.id || `call_${index}`,
-        type: 'function' as const,
-        function: {
-          name: call.name || '',
-          arguments: JSON.stringify(call.args || {}),
-        },
-      }));
-    }
-
-    const finishReason = this.mapGeminiFinishReasonToOpenAI(
-      candidate?.finishReason,
-    );
-
-    const usageMetadata = response.usageMetadata;
-    const usage: OpenAI.CompletionUsage = {
-      prompt_tokens: usageMetadata?.promptTokenCount || 0,
-      completion_tokens: usageMetadata?.candidatesTokenCount || 0,
-      total_tokens: usageMetadata?.totalTokenCount || 0,
-    };
-
-    if (usageMetadata?.cachedContentTokenCount !== undefined) {
-      (
-        usage as OpenAI.CompletionUsage & {
-          prompt_tokens_details?: { cached_tokens?: number };
-        }
-      ).prompt_tokens_details = {
-        cached_tokens: usageMetadata.cachedContentTokenCount,
-      };
-    }
-
-    const createdMs = response.createTime
-      ? Number(response.createTime)
-      : Date.now();
-    const createdSeconds = Number.isFinite(createdMs)
-      ? Math.floor(createdMs / 1000)
-      : Math.floor(Date.now() / 1000);
-
-    return {
-      id: response.responseId || `gemini-${Date.now()}`,
-      object: 'chat.completion',
-      created: createdSeconds,
-      model: response.modelVersion || this.model,
-      choices: [
-        {
-          index: 0,
-          message,
-          finish_reason: finishReason,
-          logprobs: null,
-        },
-      ],
-      usage,
-    };
+    return mergedMessages;
  }

  /**
@@ -915,6 +836,84 @@ export class OpenAIContentConverter {
    return response;
  }

+  /**
+   * Convert Gemini response format to OpenAI chat completion format for logging
+   */
+  convertGeminiResponseToOpenAI(
+    response: GenerateContentResponse,
+  ): OpenAI.Chat.ChatCompletion {
+    const candidate = response.candidates?.[0];
+    const content = candidate?.content;
+
+    let messageContent: string | null = null;
+    const toolCalls: OpenAI.Chat.ChatCompletionMessageToolCall[] = [];
+
+    if (content?.parts) {
+      const textParts: string[] = [];
+
+      for (const part of content.parts) {
+        if ('text' in part && part.text) {
+          textParts.push(part.text);
+        } else if ('functionCall' in part && part.functionCall) {
+          toolCalls.push({
+            id: part.functionCall.id || `call_${toolCalls.length}`,
+            type: 'function' as const,
+            function: {
+              name: part.functionCall.name || '',
+              arguments: JSON.stringify(part.functionCall.args || {}),
+            },
+          });
+        }
+      }
+
+      messageContent = textParts.join('').trimEnd();
+    }
+
+    const choice: OpenAI.Chat.ChatCompletion.Choice = {
+      index: 0,
+      message: {
+        role: 'assistant',
+        content: messageContent,
+        refusal: null,
+      },
+      finish_reason: this.mapGeminiFinishReasonToOpenAI(
+        candidate?.finishReason,
+      ) as OpenAI.Chat.ChatCompletion.Choice['finish_reason'],
+      logprobs: null,
+    };
+
+    if (toolCalls.length > 0) {
+      choice.message.tool_calls = toolCalls;
+    }
+
+    const openaiResponse: OpenAI.Chat.ChatCompletion = {
+      id: response.responseId || `chatcmpl-${Date.now()}`,
+      object: 'chat.completion',
+      created: response.createTime
+        ? Number(response.createTime)
+        : Math.floor(Date.now() / 1000),
+      model: this.model,
+      choices: [choice],
+    };
+
+    // Add usage metadata if available
+    if (response.usageMetadata) {
+      openaiResponse.usage = {
+        prompt_tokens: response.usageMetadata.promptTokenCount || 0,
+        completion_tokens: response.usageMetadata.candidatesTokenCount || 0,
+        total_tokens: response.usageMetadata.totalTokenCount || 0,
+      };
+
+      if (response.usageMetadata.cachedContentTokenCount) {
+        openaiResponse.usage.prompt_tokens_details = {
+          cached_tokens: response.usageMetadata.cachedContentTokenCount,
+        };
+      }
+    }
+
+    return openaiResponse;
+  }
+
  /**
   * Map OpenAI finish reasons to Gemini finish reasons
   */
@@ -932,24 +931,29 @@ export class OpenAIContentConverter {
    return mapping[openaiReason] || FinishReason.FINISH_REASON_UNSPECIFIED;
  }

-  private mapGeminiFinishReasonToOpenAI(
-    geminiReason?: FinishReason,
-  ): 'stop' | 'length' | 'tool_calls' | 'content_filter' | 'function_call' {
-    if (!geminiReason) {
-      return 'stop';
-    }
+  /**
+   * Map Gemini finish reasons to OpenAI finish reasons
+   */
+  private mapGeminiFinishReasonToOpenAI(geminiReason?: unknown): string {
+    if (!geminiReason) return 'stop';

    switch (geminiReason) {
-      case FinishReason.STOP:
+      case 'STOP':
+      case 1: // FinishReason.STOP
        return 'stop';
-      case FinishReason.MAX_TOKENS:
+      case 'MAX_TOKENS':
+      case 2: // FinishReason.MAX_TOKENS
        return 'length';
-      case FinishReason.SAFETY:
+      case 'SAFETY':
+      case 3: // FinishReason.SAFETY
        return 'content_filter';
+      case 'RECITATION':
+      case 4: // FinishReason.RECITATION
+        return 'content_filter';
+      case 'OTHER':
+      case 5: // FinishReason.OTHER
+        return 'stop';
      default:
-        if (geminiReason === ('RECITATION' as FinishReason)) {
-          return 'content_filter';
-        }
        return 'stop';
    }
  }
--- a/packages/core/src/core/openaiContentGenerator/errorHandler.test.ts
+++ b/packages/core/src/core/openaiContentGenerator/errorHandler.test.ts
@@ -7,7 +7,7 @@
 import { describe, it, expect, vi, beforeEach, afterEach } from 'vitest';
 import type { GenerateContentParameters } from '@google/genai';
 import { EnhancedErrorHandler } from './errorHandler.js';
-import type { RequestContext } from './errorHandler.js';
+import type { RequestContext } from './telemetryService.js';

 describe('EnhancedErrorHandler', () => {
  let errorHandler: EnhancedErrorHandler;
--- a/packages/core/src/core/openaiContentGenerator/errorHandler.ts
+++ b/packages/core/src/core/openaiContentGenerator/errorHandler.ts
@@ -5,15 +5,7 @@
 */

 import type { GenerateContentParameters } from '@google/genai';
-
-export interface RequestContext {
-  userPromptId: string;
-  model: string;
-  authType: string;
-  startTime: number;
-  duration: number;
-  isStreaming: boolean;
-}
+import type { RequestContext } from './telemetryService.js';

 export interface ErrorHandler {
  handle(
--- a/packages/core/src/core/openaiContentGenerator/index.ts
+++ b/packages/core/src/core/openaiContentGenerator/index.ts
@@ -91,4 +91,11 @@ export function determineProvider(
  return new DefaultOpenAICompatibleProvider(contentGeneratorConfig, cliConfig);
 }

+// Services
+export {
+  type TelemetryService,
+  type RequestContext,
+  DefaultTelemetryService,
+} from './telemetryService.js';
+
 export { type ErrorHandler, EnhancedErrorHandler } from './errorHandler.js';
--- a/packages/core/src/core/openaiContentGenerator/openaiContentGenerator.ts
+++ b/packages/core/src/core/openaiContentGenerator/openaiContentGenerator.ts
@@ -11,6 +11,7 @@ import type {
 } from '@google/genai';
 import type { PipelineConfig } from './pipeline.js';
 import { ContentGenerationPipeline } from './pipeline.js';
+import { DefaultTelemetryService } from './telemetryService.js';
 import { EnhancedErrorHandler } from './errorHandler.js';
 import { getDefaultTokenizer } from '../../utils/request-tokenizer/index.js';
 import type { ContentGeneratorConfig } from '../contentGenerator.js';
@@ -28,6 +29,11 @@ export class OpenAIContentGenerator implements ContentGenerator {
      cliConfig,
      provider,
      contentGeneratorConfig,
+      telemetryService: new DefaultTelemetryService(
+        cliConfig,
+        contentGeneratorConfig.enableOpenAILogging,
+        contentGeneratorConfig.openAILoggingDir,
+      ),
      errorHandler: new EnhancedErrorHandler(
        (error: unknown, request: GenerateContentParameters) =>
          this.shouldSuppressErrorLogging(error, request),
--- a/packages/core/src/core/openaiContentGenerator/pipeline.test.ts
+++ b/packages/core/src/core/openaiContentGenerator/pipeline.test.ts
@@ -15,6 +15,7 @@ import { OpenAIContentConverter } from './converter.js';
 import type { Config } from '../../config/config.js';
 import type { ContentGeneratorConfig, AuthType } from '../contentGenerator.js';
 import type { OpenAICompatibleProvider } from './provider/index.js';
+import type { TelemetryService } from './telemetryService.js';
 import type { ErrorHandler } from './errorHandler.js';

 // Mock dependencies
@@ -27,6 +28,7 @@ describe('ContentGenerationPipeline', () => {
  let mockProvider: OpenAICompatibleProvider;
  let mockClient: OpenAI;
  let mockConverter: OpenAIContentConverter;
+  let mockTelemetryService: TelemetryService;
  let mockErrorHandler: ErrorHandler;
  let mockContentGeneratorConfig: ContentGeneratorConfig;
  let mockCliConfig: Config;
@@ -61,6 +63,13 @@ describe('ContentGenerationPipeline', () => {
      getDefaultGenerationConfig: vi.fn().mockReturnValue({}),
    };

+    // Mock telemetry service
+    mockTelemetryService = {
+      logSuccess: vi.fn().mockResolvedValue(undefined),
+      logError: vi.fn().mockResolvedValue(undefined),
+      logStreamingSuccess: vi.fn().mockResolvedValue(undefined),
+    };
+
    // Mock error handler
    mockErrorHandler = {
      handle: vi.fn().mockImplementation((error: unknown) => {
@@ -90,6 +99,7 @@ describe('ContentGenerationPipeline', () => {
      cliConfig: mockCliConfig,
      provider: mockProvider,
      contentGeneratorConfig: mockContentGeneratorConfig,
+      telemetryService: mockTelemetryService,
      errorHandler: mockErrorHandler,
    };

@@ -162,6 +172,17 @@ describe('ContentGenerationPipeline', () => {
      expect(mockConverter.convertOpenAIResponseToGemini).toHaveBeenCalledWith(
        mockOpenAIResponse,
      );
+      expect(mockTelemetryService.logSuccess).toHaveBeenCalledWith(
+        expect.objectContaining({
+          userPromptId,
+          model: 'test-model',
+          authType: 'openai',
+          isStreaming: false,
+        }),
+        mockGeminiResponse,
+        expect.any(Object),
+        mockOpenAIResponse,
+      );
    });

    it('should handle tools in request', async () => {
@@ -247,6 +268,16 @@ describe('ContentGenerationPipeline', () => {
        'API Error',
      );

+      expect(mockTelemetryService.logError).toHaveBeenCalledWith(
+        expect.objectContaining({
+          userPromptId,
+          model: 'test-model',
+          authType: 'openai',
+          isStreaming: false,
+        }),
+        testError,
+        expect.any(Object),
+      );
      expect(mockErrorHandler.handle).toHaveBeenCalledWith(
        testError,
        expect.any(Object),
@@ -345,6 +376,17 @@ describe('ContentGenerationPipeline', () => {
          signal: undefined,
        }),
      );
+      expect(mockTelemetryService.logStreamingSuccess).toHaveBeenCalledWith(
+        expect.objectContaining({
+          userPromptId,
+          model: 'test-model',
+          authType: 'openai',
+          isStreaming: true,
+        }),
+        [mockGeminiResponse1, mockGeminiResponse2],
+        expect.any(Object),
+        [mockChunk1, mockChunk2],
+      );
    });

    it('should filter empty responses', async () => {
@@ -448,6 +490,16 @@ describe('ContentGenerationPipeline', () => {

      expect(results).toHaveLength(0); // No results due to error
      expect(mockConverter.resetStreamingToolCalls).toHaveBeenCalledTimes(2); // Once at start, once on error
+      expect(mockTelemetryService.logError).toHaveBeenCalledWith(
+        expect.objectContaining({
+          userPromptId,
+          model: 'test-model',
+          authType: 'openai',
+          isStreaming: true,
+        }),
+        testError,
+        expect.any(Object),
+      );
      expect(mockErrorHandler.handle).toHaveBeenCalledWith(
        testError,
        expect.any(Object),
@@ -598,6 +650,18 @@ describe('ContentGenerationPipeline', () => {
        candidatesTokenCount: 20,
        totalTokenCount: 30,
      });
+
+      expect(mockTelemetryService.logStreamingSuccess).toHaveBeenCalledWith(
+        expect.objectContaining({
+          userPromptId,
+          model: 'test-model',
+          authType: 'openai',
+          isStreaming: true,
+        }),
+        results,
+        expect.any(Object),
+        [mockChunk1, mockChunk2, mockChunk3],
+      );
    });

    it('should handle ideal case where last chunk has both finishReason and usageMetadata', async () => {
@@ -789,6 +853,18 @@ describe('ContentGenerationPipeline', () => {
        candidatesTokenCount: 20,
        totalTokenCount: 30,
      });
+
+      expect(mockTelemetryService.logStreamingSuccess).toHaveBeenCalledWith(
+        expect.objectContaining({
+          userPromptId,
+          model: 'test-model',
+          authType: 'openai',
+          isStreaming: true,
+        }),
+        results,
+        expect.any(Object),
+        [mockChunk1, mockChunk2, mockChunk3],
+      );
    });

    it('should handle providers that send finishReason and valid usage in same chunk', async () => {
@@ -1042,6 +1118,19 @@ describe('ContentGenerationPipeline', () => {
      await pipeline.execute(request, userPromptId);

      // Assert
+      expect(mockTelemetryService.logSuccess).toHaveBeenCalledWith(
+        expect.objectContaining({
+          userPromptId,
+          model: 'test-model',
+          authType: 'openai',
+          isStreaming: false,
+          startTime: expect.any(Number),
+          duration: expect.any(Number),
+        }),
+        expect.any(Object),
+        expect.any(Object),
+        expect.any(Object),
+      );
    });

    it('should create context with correct properties for streaming request', async () => {
@@ -1084,6 +1173,19 @@ describe('ContentGenerationPipeline', () => {
      }

      // Assert
+      expect(mockTelemetryService.logStreamingSuccess).toHaveBeenCalledWith(
+        expect.objectContaining({
+          userPromptId,
+          model: 'test-model',
+          authType: 'openai',
+          isStreaming: true,
+          startTime: expect.any(Number),
+          duration: expect.any(Number),
+        }),
+        expect.any(Array),
+        expect.any(Object),
+        expect.any(Array),
+      );
    });

    it('should collect all OpenAI chunks for logging even when Gemini responses are filtered', async () => {
@@ -1227,6 +1329,22 @@ describe('ContentGenerationPipeline', () => {
      // Should only yield the final response (empty ones are filtered)
      expect(responses).toHaveLength(1);
      expect(responses[0]).toBe(finalGeminiResponse);
+
+      // Verify telemetry was called with ALL OpenAI chunks, including the filtered ones
+      expect(mockTelemetryService.logStreamingSuccess).toHaveBeenCalledWith(
+        expect.objectContaining({
+          model: 'test-model',
+          duration: expect.any(Number),
+          userPromptId: 'test-prompt-id',
+          authType: 'openai',
+        }),
+        [finalGeminiResponse], // Only the non-empty Gemini response
+        expect.objectContaining({
+          model: 'test-model',
+          messages: [{ role: 'user', content: 'test' }],
+        }),
+        [partialToolCallChunk1, partialToolCallChunk2, finishChunk], // ALL OpenAI chunks
+      );
    });
  });
 });
--- a/packages/core/src/core/openaiContentGenerator/pipeline.ts
+++ b/packages/core/src/core/openaiContentGenerator/pipeline.ts
@@ -13,12 +13,14 @@ import type { Config } from '../../config/config.js';
 import type { ContentGeneratorConfig } from '../contentGenerator.js';
 import type { OpenAICompatibleProvider } from './provider/index.js';
 import { OpenAIContentConverter } from './converter.js';
-import type { ErrorHandler, RequestContext } from './errorHandler.js';
+import type { TelemetryService, RequestContext } from './telemetryService.js';
+import type { ErrorHandler } from './errorHandler.js';

 export interface PipelineConfig {
  cliConfig: Config;
  provider: OpenAICompatibleProvider;
  contentGeneratorConfig: ContentGeneratorConfig;
+  telemetryService: TelemetryService;
  errorHandler: ErrorHandler;
 }

@@ -44,7 +46,7 @@ export class ContentGenerationPipeline {
      request,
      userPromptId,
      false,
-      async (openaiRequest) => {
+      async (openaiRequest, context) => {
        const openaiResponse = (await this.client.chat.completions.create(
          openaiRequest,
          {
@@ -55,6 +57,14 @@ export class ContentGenerationPipeline {
        const geminiResponse =
          this.converter.convertOpenAIResponseToGemini(openaiResponse);

+        // Log success
+        await this.config.telemetryService.logSuccess(
+          context,
+          geminiResponse,
+          openaiRequest,
+          openaiResponse,
+        );
+
        return geminiResponse;
      },
    );
@@ -78,7 +88,12 @@ export class ContentGenerationPipeline {
        )) as AsyncIterable<OpenAI.Chat.ChatCompletionChunk>;

        // Stage 2: Process stream with conversion and logging
-        return this.processStreamWithLogging(stream, context, request);
+        return this.processStreamWithLogging(
+          stream,
+          context,
+          openaiRequest,
+          request,
+        );
      },
    );
  }
@@ -95,9 +110,11 @@ export class ContentGenerationPipeline {
  private async *processStreamWithLogging(
    stream: AsyncIterable<OpenAI.Chat.ChatCompletionChunk>,
    context: RequestContext,
+    openaiRequest: OpenAI.Chat.ChatCompletionCreateParams,
    request: GenerateContentParameters,
  ): AsyncGenerator<GenerateContentResponse> {
    const collectedGeminiResponses: GenerateContentResponse[] = [];
+    const collectedOpenAIChunks: OpenAI.Chat.ChatCompletionChunk[] = [];

    // Reset streaming tool calls to prevent data pollution from previous streams
    this.converter.resetStreamingToolCalls();
@@ -108,6 +125,9 @@ export class ContentGenerationPipeline {
    try {
      // Stage 2a: Convert and yield each chunk while preserving original
      for await (const chunk of stream) {
+        // Always collect OpenAI chunks for logging, regardless of Gemini conversion result
+        collectedOpenAIChunks.push(chunk);
+
        const response = this.converter.convertOpenAIChunkToGemini(chunk);

        // Stage 2b: Filter empty responses to avoid downstream issues
@@ -144,8 +164,15 @@ export class ContentGenerationPipeline {
        yield pendingFinishResponse;
      }

-      // Stage 2e: Stream completed successfully
+      // Stage 2e: Stream completed successfully - perform logging with original OpenAI chunks
      context.duration = Date.now() - context.startTime;
+
+      await this.config.telemetryService.logStreamingSuccess(
+        context,
+        collectedGeminiResponses,
+        openaiRequest,
+        collectedOpenAIChunks,
+      );
    } catch (error) {
      // Clear streaming tool calls on error to prevent data pollution
      this.converter.resetStreamingToolCalls();
@@ -231,7 +258,7 @@ export class ContentGenerationPipeline {
    const baseRequest: OpenAI.Chat.ChatCompletionCreateParams = {
      model: this.contentGeneratorConfig.model,
      messages,
-      ...this.buildGenerateContentConfig(request),
+      ...this.buildSamplingParameters(request),
    };

    // Add streaming options if present
@@ -253,7 +280,7 @@ export class ContentGenerationPipeline {
    return this.config.provider.buildRequest(baseRequest, userPromptId);
  }

-  private buildGenerateContentConfig(
+  private buildSamplingParameters(
    request: GenerateContentParameters,
  ): Record<string, unknown> {
    const defaultSamplingParams =
@@ -289,7 +316,7 @@ export class ContentGenerationPipeline {
      return value !== undefined ? { [key]: value } : {};
    };

-    const params: Record<string, unknown> = {
+    const params = {
      // Parameters with request fallback but no defaults
      ...addParameterIfDefined('temperature', 'temperature', 'temperature'),
      ...addParameterIfDefined('top_p', 'top_p', 'topP'),
@@ -310,24 +337,11 @@ export class ContentGenerationPipeline {
        'frequency_penalty',
        'frequencyPenalty',
      ),
-      ...this.buildReasoningConfig(),
    };

    return params;
  }

-  private buildReasoningConfig(): Record<string, unknown> {
-    const reasoning = this.contentGeneratorConfig.reasoning;
-
-    if (reasoning === false) {
-      return {};
-    }
-
-    return {
-      reasoning_effort: reasoning?.effort ?? 'medium',
-    };
-  }
-
  /**
   * Common error handling wrapper for execute methods
   */
@@ -355,7 +369,13 @@ export class ContentGenerationPipeline {
      return result;
    } catch (error) {
      // Use shared error handling logic
-      return await this.handleError(error, context, request);
+      return await this.handleError(
+        error,
+        context,
+        request,
+        userPromptId,
+        isStreaming,
+      );
    }
  }

@@ -367,8 +387,37 @@ export class ContentGenerationPipeline {
    error: unknown,
    context: RequestContext,
    request: GenerateContentParameters,
+    userPromptId?: string,
+    isStreaming?: boolean,
  ): Promise<never> {
    context.duration = Date.now() - context.startTime;
+
+    // Build request for logging (may fail, but we still want to log the error)
+    let openaiRequest: OpenAI.Chat.ChatCompletionCreateParams;
+    try {
+      if (userPromptId !== undefined && isStreaming !== undefined) {
+        openaiRequest = await this.buildRequest(
+          request,
+          userPromptId,
+          isStreaming,
+        );
+      } else {
+        // For processStreamWithLogging, we don't have userPromptId/isStreaming,
+        // so create a minimal request
+        openaiRequest = {
+          model: this.contentGeneratorConfig.model,
+          messages: [],
+        };
+      }
+    } catch (_buildError) {
+      // If we can't build the request, create a minimal one for logging
+      openaiRequest = {
+        model: this.contentGeneratorConfig.model,
+        messages: [],
+      };
+    }
+
+    await this.config.telemetryService.logError(context, error, openaiRequest);
    this.config.errorHandler.handle(error, context, request);
  }

--- a/packages/core/src/core/openaiContentGenerator/provider/dashscope.ts
+++ b/packages/core/src/core/openaiContentGenerator/provider/dashscope.ts
@@ -39,8 +39,7 @@ export class DashScopeOpenAICompatibleProvider
    return (
      authType === AuthType.QWEN_OAUTH ||
      baseUrl === 'https://dashscope.aliyuncs.com/compatible-mode/v1' ||
-      baseUrl === 'https://dashscope-intl.aliyuncs.com/compatible-mode/v1' ||
-      !baseUrl
+      baseUrl === 'https://dashscope-intl.aliyuncs.com/compatible-mode/v1'
    );
  }

@@ -145,7 +144,9 @@ export class DashScopeOpenAICompatibleProvider

  getDefaultGenerationConfig(): GenerateContentConfig {
    return {
-      temperature: 0.3,
+      temperature: 0.7,
+      topP: 0.8,
+      topK: 20,
    };
  }

--- a/packages/core/src/core/openaiContentGenerator/telemetryService.test.ts
+++ b/packages/core/src/core/openaiContentGenerator/telemetryService.test.ts
--- a/packages/core/src/core/openaiContentGenerator/telemetryService.ts
+++ b/packages/core/src/core/openaiContentGenerator/telemetryService.ts
@@ -0,0 +1,275 @@
+/**
+ * @license
+ * Copyright 2025 Qwen
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+import type { Config } from '../../config/config.js';
+import { logApiError, logApiResponse } from '../../telemetry/loggers.js';
+import { ApiErrorEvent, ApiResponseEvent } from '../../telemetry/types.js';
+import { OpenAILogger } from '../../utils/openaiLogger.js';
+import type { GenerateContentResponse } from '@google/genai';
+import type OpenAI from 'openai';
+import type { ExtendedCompletionChunkDelta } from './converter.js';
+
+export interface RequestContext {
+  userPromptId: string;
+  model: string;
+  authType: string;
+  startTime: number;
+  duration: number;
+  isStreaming: boolean;
+}
+
+export interface TelemetryService {
+  logSuccess(
+    context: RequestContext,
+    response: GenerateContentResponse,
+    openaiRequest?: OpenAI.Chat.ChatCompletionCreateParams,
+    openaiResponse?: OpenAI.Chat.ChatCompletion,
+  ): Promise<void>;
+
+  logError(
+    context: RequestContext,
+    error: unknown,
+    openaiRequest?: OpenAI.Chat.ChatCompletionCreateParams,
+  ): Promise<void>;
+
+  logStreamingSuccess(
+    context: RequestContext,
+    responses: GenerateContentResponse[],
+    openaiRequest?: OpenAI.Chat.ChatCompletionCreateParams,
+    openaiChunks?: OpenAI.Chat.ChatCompletionChunk[],
+  ): Promise<void>;
+}
+
+export class DefaultTelemetryService implements TelemetryService {
+  private logger: OpenAILogger;
+
+  constructor(
+    private config: Config,
+    private enableOpenAILogging: boolean = false,
+    openAILoggingDir?: string,
+  ) {
+    // Always create a new logger instance to ensure correct working directory
+    // If no custom directory is provided, undefined will use the default path
+    this.logger = new OpenAILogger(openAILoggingDir);
+  }
+
+  async logSuccess(
+    context: RequestContext,
+    response: GenerateContentResponse,
+    openaiRequest?: OpenAI.Chat.ChatCompletionCreateParams,
+    openaiResponse?: OpenAI.Chat.ChatCompletion,
+  ): Promise<void> {
+    // Log API response event for UI telemetry
+    const responseEvent = new ApiResponseEvent(
+      response.responseId || 'unknown',
+      context.model,
+      context.duration,
+      context.userPromptId,
+      context.authType,
+      response.usageMetadata,
+    );
+
+    logApiResponse(this.config, responseEvent);
+
+    // Log interaction if enabled
+    if (this.enableOpenAILogging && openaiRequest && openaiResponse) {
+      await this.logger.logInteraction(openaiRequest, openaiResponse);
+    }
+  }
+
+  async logError(
+    context: RequestContext,
+    error: unknown,
+    openaiRequest?: OpenAI.Chat.ChatCompletionCreateParams,
+  ): Promise<void> {
+    const errorMessage = error instanceof Error ? error.message : String(error);
+
+    // Log API error event for UI telemetry
+    const errorEvent = new ApiErrorEvent(
+      // eslint-disable-next-line @typescript-eslint/no-explicit-any
+      (error as any)?.requestID || 'unknown',
+      context.model,
+      errorMessage,
+      context.duration,
+      context.userPromptId,
+      context.authType,
+      // eslint-disable-next-line @typescript-eslint/no-explicit-any
+      (error as any)?.type,
+      // eslint-disable-next-line @typescript-eslint/no-explicit-any
+      (error as any)?.code,
+    );
+    logApiError(this.config, errorEvent);
+
+    // Log error interaction if enabled
+    if (this.enableOpenAILogging && openaiRequest) {
+      await this.logger.logInteraction(
+        openaiRequest,
+        undefined,
+        error as Error,
+      );
+    }
+  }
+
+  async logStreamingSuccess(
+    context: RequestContext,
+    responses: GenerateContentResponse[],
+    openaiRequest?: OpenAI.Chat.ChatCompletionCreateParams,
+    openaiChunks?: OpenAI.Chat.ChatCompletionChunk[],
+  ): Promise<void> {
+    // Get final usage metadata from the last response that has it
+    const finalUsageMetadata = responses
+      .slice()
+      .reverse()
+      .find((r) => r.usageMetadata)?.usageMetadata;
+
+    // Log API response event for UI telemetry
+    const responseEvent = new ApiResponseEvent(
+      responses[responses.length - 1]?.responseId || 'unknown',
+      context.model,
+      context.duration,
+      context.userPromptId,
+      context.authType,
+      finalUsageMetadata,
+    );
+
+    logApiResponse(this.config, responseEvent);
+
+    // Log interaction if enabled - combine chunks only when needed
+    if (
+      this.enableOpenAILogging &&
+      openaiRequest &&
+      openaiChunks &&
+      openaiChunks.length > 0
+    ) {
+      const combinedResponse = this.combineOpenAIChunksForLogging(openaiChunks);
+      await this.logger.logInteraction(openaiRequest, combinedResponse);
+    }
+  }
+
+  /**
+   * Combine OpenAI chunks for logging purposes
+   * This method consolidates all OpenAI stream chunks into a single ChatCompletion response
+   * for telemetry and logging purposes, avoiding unnecessary format conversions
+   */
+  private combineOpenAIChunksForLogging(
+    chunks: OpenAI.Chat.ChatCompletionChunk[],
+  ): OpenAI.Chat.ChatCompletion {
+    if (chunks.length === 0) {
+      throw new Error('No chunks to combine');
+    }
+
+    const firstChunk = chunks[0];
+
+    // Combine all content from chunks
+    let combinedContent = '';
+    const toolCalls: OpenAI.Chat.ChatCompletionMessageToolCall[] = [];
+    let finishReason:
+      | 'stop'
+      | 'length'
+      | 'tool_calls'
+      | 'content_filter'
+      | 'function_call'
+      | null = null;
+    let combinedReasoning = '';
+    let usage:
+      | {
+          prompt_tokens: number;
+          completion_tokens: number;
+          total_tokens: number;
+        }
+      | undefined;
+
+    for (const chunk of chunks) {
+      const choice = chunk.choices?.[0];
+      if (choice) {
+        // Combine reasoning content
+        const reasoningContent = (choice.delta as ExtendedCompletionChunkDelta)
+          ?.reasoning_content;
+        if (reasoningContent) {
+          combinedReasoning += reasoningContent;
+        }
+        // Combine text content
+        if (choice.delta?.content) {
+          combinedContent += choice.delta.content;
+        }
+
+        // Collect tool calls
+        if (choice.delta?.tool_calls) {
+          for (const toolCall of choice.delta.tool_calls) {
+            if (toolCall.index !== undefined) {
+              if (!toolCalls[toolCall.index]) {
+                toolCalls[toolCall.index] = {
+                  id: toolCall.id || '',
+                  type: toolCall.type || 'function',
+                  function: { name: '', arguments: '' },
+                };
+              }
+
+              if (toolCall.function?.name) {
+                toolCalls[toolCall.index].function.name +=
+                  toolCall.function.name;
+              }
+              if (toolCall.function?.arguments) {
+                toolCalls[toolCall.index].function.arguments +=
+                  toolCall.function.arguments;
+              }
+            }
+          }
+        }
+
+        // Get finish reason from the last chunk
+        if (choice.finish_reason) {
+          finishReason = choice.finish_reason;
+        }
+      }
+
+      // Get usage from the last chunk that has it
+      if (chunk.usage) {
+        usage = chunk.usage;
+      }
+    }
+
+    // Create the combined ChatCompletion response
+    const message: OpenAI.Chat.ChatCompletionMessage = {
+      role: 'assistant',
+      content: combinedContent || null,
+      refusal: null,
+    };
+    if (combinedReasoning) {
+      // Attach reasoning content if any thought tokens were streamed
+      (message as { reasoning_content?: string }).reasoning_content =
+        combinedReasoning;
+    }
+
+    // Add tool calls if any
+    if (toolCalls.length > 0) {
+      message.tool_calls = toolCalls.filter((tc) => tc.id); // Filter out empty tool calls
+    }
+
+    const combinedResponse: OpenAI.Chat.ChatCompletion = {
+      id: firstChunk.id,
+      object: 'chat.completion',
+      created: firstChunk.created,
+      model: firstChunk.model,
+      choices: [
+        {
+          index: 0,
+          message,
+          finish_reason: finishReason || 'stop',
+          logprobs: null,
+        },
+      ],
+      usage: usage || {
+        prompt_tokens: 0,
+        completion_tokens: 0,
+        total_tokens: 0,
+      },
+      system_fingerprint: firstChunk.system_fingerprint,
+    };
+
+    return combinedResponse;
+  }
+}
--- a/packages/core/src/telemetry/loggers.test.ts
+++ b/packages/core/src/telemetry/loggers.test.ts
@@ -264,7 +264,7 @@ describe('loggers', () => {
          'event.timestamp': '2025-01-01T00:00:00.000Z',
          prompt_length: 11,
          prompt_id: 'prompt-id-9',
-          auth_type: 'gemini',
+          auth_type: 'gemini-api-key',
        },
      });
    });
@@ -333,7 +333,7 @@ describe('loggers', () => {
          total_token_count: 0,
          response_text: 'test-response',
          prompt_id: 'prompt-id-1',
-          auth_type: 'gemini',
+          auth_type: 'gemini-api-key',
        },
      });

--- a/packages/sdk-typescript/package.json
+++ b/packages/sdk-typescript/package.json
@@ -1,6 +1,6 @@
 {
  "name": "@qwen-code/sdk",
-  "version": "0.1.0",
+  "version": "0.6.0-nightly.20251225.9f65bd3b",
  "description": "TypeScript SDK for programmatic access to qwen-code CLI",
  "main": "./dist/index.cjs",
  "module": "./dist/index.mjs",
--- a/packages/sdk-typescript/src/query/Query.ts
+++ b/packages/sdk-typescript/src/query/Query.ts
@@ -272,6 +272,8 @@ export class Query implements AsyncIterable<SDKMessage> {
      // Get only successfully connected SDK servers for CLI
      const sdkMcpServersForCli = this.getSdkMcpServersForCli();
      const mcpServersForCli = this.getMcpServersForCli();
+      logger.debug('SDK MCP servers for CLI:', sdkMcpServersForCli);
+      logger.debug('External MCP servers for CLI:', mcpServersForCli);

      await this.sendControlRequest(ControlRequestType.INITIALIZE, {
        hooks: null,
@@ -627,11 +629,6 @@ export class Query implements AsyncIterable<SDKMessage> {
      return Promise.reject(new Error('Query is closed'));
    }

-    if (subtype !== ControlRequestType.INITIALIZE) {
-      // Ensure all other control requests get processed after initialization
-      await this.initialized;
-    }
-
    const requestId = randomUUID();

    const request: CLIControlRequest = {
--- a/packages/test-utils/package.json
+++ b/packages/test-utils/package.json
@@ -1,6 +1,6 @@
 {
  "name": "@qwen-code/qwen-code-test-utils",
-  "version": "0.6.0-nightly.20251226.17eb20c1",
+  "version": "0.6.0-nightly.20251225.9f65bd3b",
  "private": true,
  "main": "src/index.ts",
  "license": "Apache-2.0",
--- a/packages/vscode-ide-companion/package.json
+++ b/packages/vscode-ide-companion/package.json
@@ -2,7 +2,7 @@
  "name": "qwen-code-vscode-ide-companion",
  "displayName": "Qwen Code Companion",
  "description": "Enable Qwen Code with direct access to your VS Code workspace.",
-  "version": "0.6.0-nightly.20251226.17eb20c1",
+  "version": "0.6.0-nightly.20251225.9f65bd3b",
  "publisher": "qwenlm",
  "icon": "assets/icon.png",
  "repository": {
--- a/packages/vscode-ide-companion/src/extension.test.ts
+++ b/packages/vscode-ide-companion/src/extension.test.ts
@@ -50,9 +50,6 @@ vi.mock('vscode', () => ({
    registerTextDocumentContentProvider: vi.fn(),
    onDidChangeWorkspaceFolders: vi.fn(),
    onDidGrantWorkspaceTrust: vi.fn(),
-    registerFileSystemProvider: vi.fn(() => ({
-      dispose: vi.fn(),
-    })),
  },
  commands: {
    registerCommand: vi.fn(),
--- a/packages/vscode-ide-companion/src/extension.ts
+++ b/packages/vscode-ide-companion/src/extension.ts
@@ -16,7 +16,6 @@ import {
 } from '@qwen-code/qwen-code-core/src/ide/detect-ide.js';
 import { WebViewProvider } from './webview/WebViewProvider.js';
 import { registerNewCommands } from './commands/index.js';
-import { ReadonlyFileSystemProvider } from './services/readonlyFileSystemProvider.js';

 const CLI_IDE_COMPANION_IDENTIFIER = 'qwenlm.qwen-code-vscode-ide-companion';
 const INFO_MESSAGE_SHOWN_KEY = 'qwenCodeInfoMessageShown';
@@ -111,19 +110,6 @@ export async function activate(context: vscode.ExtensionContext) {

  checkForUpdates(context, log);

-  // Create and register readonly file system provider
-  // The provider registers itself as a singleton in the constructor
-  const readonlyProvider = new ReadonlyFileSystemProvider();
-  context.subscriptions.push(
-    vscode.workspace.registerFileSystemProvider(
-      ReadonlyFileSystemProvider.getScheme(),
-      readonlyProvider,
-      { isCaseSensitive: true, isReadonly: true },
-    ),
-    readonlyProvider,
-  );
-  log('Readonly file system provider registered');
-
  const diffContentProvider = new DiffContentProvider();
  const diffManager = new DiffManager(
    log,
--- a/packages/vscode-ide-companion/src/ide-server.test.ts
+++ b/packages/vscode-ide-companion/src/ide-server.test.ts
@@ -38,10 +38,6 @@ vi.mock('node:os', async (importOriginal) => {
  };
 });

-vi.mock('@qwen-code/qwen-code-core/src/ide/detect-ide.js', () => ({
-  detectIdeFromEnv: vi.fn(() => ({ name: 'vscode', displayName: 'VS Code' })),
-}));
-
 const vscodeMock = vi.hoisted(() => ({
  workspace: {
    workspaceFolders: [
--- a/packages/vscode-ide-companion/src/services/acpConnection.ts
+++ b/packages/vscode-ide-companion/src/services/acpConnection.ts
@@ -146,8 +146,6 @@ export class AcpConnection {
      console.error(
        `[ACP qwen] Process exited with code: ${code}, signal: ${signal}`,
      );
-      // Clear pending requests when process exits
-      this.pendingRequests.clear();
    });

    // Wait for process to start
--- a/packages/vscode-ide-companion/src/services/qwenAgentManager.ts
+++ b/packages/vscode-ide-companion/src/services/qwenAgentManager.ts
@@ -8,7 +8,6 @@ import type {
  AcpSessionUpdate,
  AcpPermissionRequest,
  AuthenticateUpdateNotification,
-  ModelInfo,
 } from '../types/acpTypes.js';
 import type { ApprovalModeValue } from '../types/approvalModeValueTypes.js';
 import { QwenSessionReader, type QwenSession } from './qwenSessionReader.js';
@@ -18,7 +17,6 @@ import type {
  PlanEntry,
  ToolCallUpdateData,
  QwenAgentCallbacks,
-  UsageStatsPayload,
 } from '../types/chatTypes.js';
 import {
  QwenConnectionHandler,
@@ -26,7 +24,6 @@ import {
 } from '../services/qwenConnectionHandler.js';
 import { QwenSessionUpdateHandler } from './qwenSessionUpdateHandler.js';
 import { authMethod } from '../types/acpTypes.js';
-import { extractModelInfoFromNewSessionResult } from '../utils/acpModelInfo.js';
 import { isAuthenticationRequiredError } from '../utils/authErrors.js';
 import { handleAuthenticateUpdate } from '../utils/authNotificationHandler.js';

@@ -198,16 +195,12 @@ export class QwenAgentManager {
    options?: AgentConnectOptions,
  ): Promise<QwenConnectionResult> {
    this.currentWorkingDir = workingDir;
-    const res = await this.connectionHandler.connect(
+    return this.connectionHandler.connect(
      this.connection,
      workingDir,
      cliEntryPath,
      options,
    );
-    if (res.modelInfo && this.callbacks.onModelInfo) {
-      this.callbacks.onModelInfo(res.modelInfo);
-    }
-    return res;
  }

  /**
@@ -1098,10 +1091,9 @@ export class QwenAgentManager {

    this.sessionCreateInFlight = (async () => {
      try {
-        let newSessionResult: unknown;
        // Try to create a new ACP session. If Qwen asks for auth, let it handle authentication.
        try {
-          newSessionResult = await this.connection.newSession(workingDir);
+          await this.connection.newSession(workingDir);
        } catch (err) {
          const requiresAuth = isAuthenticationRequiredError(err);

@@ -1123,7 +1115,7 @@ export class QwenAgentManager {
              );
              // Add a slight delay to ensure auth state is settled
              await new Promise((resolve) => setTimeout(resolve, 300));
-              newSessionResult = await this.connection.newSession(workingDir);
+              await this.connection.newSession(workingDir);
            } catch (reauthErr) {
              console.error(
                '[QwenAgentManager] Re-authentication failed:',
@@ -1135,13 +1127,6 @@ export class QwenAgentManager {
            throw err;
          }
        }
-
-        const modelInfo =
-          extractModelInfoFromNewSessionResult(newSessionResult);
-        if (modelInfo && this.callbacks.onModelInfo) {
-          this.callbacks.onModelInfo(modelInfo);
-        }
-
        const newSessionId = this.connection.currentSessionId;
        console.log(
          '[QwenAgentManager] New session created with ID:',
@@ -1272,22 +1257,6 @@ export class QwenAgentManager {
    this.sessionUpdateHandler.updateCallbacks(this.callbacks);
  }

-  /**
-   * Register callback for usage metadata updates
-   */
-  onUsageUpdate(callback: (stats: UsageStatsPayload) => void): void {
-    this.callbacks.onUsageUpdate = callback;
-    this.sessionUpdateHandler.updateCallbacks(this.callbacks);
-  }
-
-  /**
-   * Register callback for model info updates
-   */
-  onModelInfo(callback: (info: ModelInfo) => void): void {
-    this.callbacks.onModelInfo = callback;
-    this.sessionUpdateHandler.updateCallbacks(this.callbacks);
-  }
-
  /**
   * Disconnect
   */
--- a/packages/vscode-ide-companion/src/services/qwenConnectionHandler.ts
+++ b/packages/vscode-ide-companion/src/services/qwenConnectionHandler.ts
@@ -13,13 +13,10 @@
 import type { AcpConnection } from './acpConnection.js';
 import { isAuthenticationRequiredError } from '../utils/authErrors.js';
 import { authMethod } from '../types/acpTypes.js';
-import { extractModelInfoFromNewSessionResult } from '../utils/acpModelInfo.js';
-import type { ModelInfo } from '../types/acpTypes.js';

 export interface QwenConnectionResult {
  sessionCreated: boolean;
  requiresAuth: boolean;
-  modelInfo?: ModelInfo;
 }

 /**
@@ -47,7 +44,6 @@ export class QwenConnectionHandler {
    const autoAuthenticate = options?.autoAuthenticate ?? true;
    let sessionCreated = false;
    let requiresAuth = false;
-    let modelInfo: ModelInfo | undefined;

    // Build extra CLI arguments (only essential parameters)
    const extraArgs: string[] = [];
@@ -70,15 +66,13 @@ export class QwenConnectionHandler {
        console.log(
          '[QwenAgentManager] Creating new session (letting CLI handle authentication)...',
        );
-        const newSessionResult = await this.newSessionWithRetry(
+        await this.newSessionWithRetry(
          connection,
          workingDir,
          3,
          authMethod,
          autoAuthenticate,
        );
-        modelInfo =
-          extractModelInfoFromNewSessionResult(newSessionResult) || undefined;
        console.log('[QwenAgentManager] New session created successfully');
        sessionCreated = true;
      } catch (sessionError) {
@@ -105,7 +99,7 @@ export class QwenConnectionHandler {
    console.log(`\n========================================`);
    console.log(`[QwenAgentManager] ✅ CONNECT() COMPLETED SUCCESSFULLY`);
    console.log(`========================================\n`);
-    return { sessionCreated, requiresAuth, modelInfo };
+    return { sessionCreated, requiresAuth };
  }

  /**
@@ -121,15 +115,15 @@ export class QwenConnectionHandler {
    maxRetries: number,
    authMethod: string,
    autoAuthenticate: boolean,
-  ): Promise<unknown> {
+  ): Promise<void> {
    for (let attempt = 1; attempt <= maxRetries; attempt++) {
      try {
        console.log(
          `[QwenAgentManager] Creating session (attempt ${attempt}/${maxRetries})...`,
        );
-        const res = await connection.newSession(workingDir);
+        await connection.newSession(workingDir);
        console.log('[QwenAgentManager] Session created successfully');
-        return res;
+        return;
      } catch (error) {
        const errorMessage =
          error instanceof Error ? error.message : String(error);
@@ -161,11 +155,11 @@ export class QwenConnectionHandler {
              '[QwenAgentManager] newSessionWithRetry Authentication successful',
            );
            // Retry immediately after successful auth
-            const res = await connection.newSession(workingDir);
+            await connection.newSession(workingDir);
            console.log(
              '[QwenAgentManager] Session created successfully after auth',
            );
-            return res;
+            return;
          } catch (authErr) {
            console.error(
              '[QwenAgentManager] Re-authentication failed:',
@@ -186,7 +180,5 @@ export class QwenConnectionHandler {
        await new Promise((resolve) => setTimeout(resolve, delay));
      }
    }
-
-    throw new Error('Session creation failed unexpectedly');
  }
 }
--- a/packages/vscode-ide-companion/src/services/qwenSessionUpdateHandler.ts
+++ b/packages/vscode-ide-companion/src/services/qwenSessionUpdateHandler.ts
@@ -10,12 +10,9 @@
 * Handles session updates from ACP and dispatches them to appropriate callbacks
 */

-import type { AcpSessionUpdate, SessionUpdateMeta } from '../types/acpTypes.js';
+import type { AcpSessionUpdate } from '../types/acpTypes.js';
 import type { ApprovalModeValue } from '../types/approvalModeValueTypes.js';
-import type {
-  QwenAgentCallbacks,
-  UsageStatsPayload,
-} from '../types/chatTypes.js';
+import type { QwenAgentCallbacks } from '../types/chatTypes.js';

 /**
 * Qwen Session Update Handler class
@@ -60,7 +57,6 @@ export class QwenSessionUpdateHandler {
        if (update.content?.text && this.callbacks.onStreamChunk) {
          this.callbacks.onStreamChunk(update.content.text);
        }
-        this.emitUsageMeta(update._meta);
        break;

      case 'agent_thought_chunk':
@@ -75,7 +71,6 @@ export class QwenSessionUpdateHandler {
            this.callbacks.onStreamChunk(update.content.text);
          }
        }
-        this.emitUsageMeta(update._meta);
        break;

      case 'tool_call': {
@@ -165,17 +160,4 @@ export class QwenSessionUpdateHandler {
        break;
    }
  }
-
-  private emitUsageMeta(meta?: SessionUpdateMeta): void {
-    if (!meta || !this.callbacks.onUsageUpdate) {
-      return;
-    }
-
-    const payload: UsageStatsPayload = {
-      usage: meta.usage || undefined,
-      durationMs: meta.durationMs ?? undefined,
-    };
-
-    this.callbacks.onUsageUpdate(payload);
-  }
 }
--- a/packages/vscode-ide-companion/src/services/readonlyFileSystemProvider.ts
+++ b/packages/vscode-ide-companion/src/services/readonlyFileSystemProvider.ts
@@ -1,204 +0,0 @@
-/**
- * @license
- * Copyright 2025 Qwen Team
- * SPDX-License-Identifier: Apache-2.0
- */
-
-import * as vscode from 'vscode';
-
-/**
- * Readonly file system provider for temporary files
- * Uses custom URI scheme to create readonly documents in VS Code
- */
-export class ReadonlyFileSystemProvider
-  implements vscode.FileSystemProvider, vscode.Disposable
-{
-  private static readonly scheme = 'qwen-readonly';
-  private static instance: ReadonlyFileSystemProvider | null = null;
-
-  private readonly files = new Map<string, Uint8Array>();
-  private readonly emitter = new vscode.EventEmitter<
-    vscode.FileChangeEvent[]
-  >();
-  private readonly disposables: vscode.Disposable[] = [];
-
-  readonly onDidChangeFile = this.emitter.event;
-
-  constructor() {
-    // Ensure only one instance exists
-    if (ReadonlyFileSystemProvider.instance !== null) {
-      console.warn(
-        '[ReadonlyFileSystemProvider] Instance already exists, replacing with new instance',
-      );
-    }
-    this.disposables.push(this.emitter);
-    // Register as global singleton
-    ReadonlyFileSystemProvider.instance = this;
-  }
-
-  static getScheme(): string {
-    return ReadonlyFileSystemProvider.scheme;
-  }
-
-  /**
-   * Get the global singleton instance
-   * Returns null if not initialized yet
-   */
-  static getInstance(): ReadonlyFileSystemProvider | null {
-    return ReadonlyFileSystemProvider.instance;
-  }
-
-  /**
-   * Create a URI for a readonly temporary file (static version)
-   */
-  static createUri(fileName: string, content: string): vscode.Uri {
-    // For tool-call related filenames, keep the URI stable so repeated clicks focus the same document.
-    // Note: toolCallId can include underscores (e.g. "call_..."), so match everything after the prefix.
-    const isToolCallFile =
-      /^(bash-input|bash-output|execute-input|execute-output)-.+$/.test(
-        fileName,
-      );
-
-    if (isToolCallFile) {
-      return vscode.Uri.from({
-        scheme: ReadonlyFileSystemProvider.scheme,
-        path: `/${fileName}`,
-      });
-    }
-
-    // For other cases, keep the original approach with timestamp to avoid collisions.
-    const timestamp = Date.now();
-    const hash = Buffer.from(content.substring(0, 100)).toString('base64url');
-    const uniqueId = `${timestamp}-${hash.substring(0, 8)}`;
-    return vscode.Uri.from({
-      scheme: ReadonlyFileSystemProvider.scheme,
-      path: `/${fileName}-${uniqueId}`,
-    });
-  }
-
-  /**
-   * Create a URI for a readonly temporary file (instance method)
-   */
-  createUri(fileName: string, content: string): vscode.Uri {
-    return ReadonlyFileSystemProvider.createUri(fileName, content);
-  }
-
-  /**
-   * Set content for a URI
-   */
-  setContent(uri: vscode.Uri, content: string): void {
-    const buffer = Buffer.from(content, 'utf8');
-    const key = uri.toString();
-    const existed = this.files.has(key);
-    this.files.set(key, buffer);
-    this.emitter.fire([
-      {
-        type: existed
-          ? vscode.FileChangeType.Changed
-          : vscode.FileChangeType.Created,
-        uri,
-      },
-    ]);
-  }
-
-  /**
-   * Get content for a URI
-   */
-  getContent(uri: vscode.Uri): string | undefined {
-    const buffer = this.files.get(uri.toString());
-    return buffer ? Buffer.from(buffer).toString('utf8') : undefined;
-  }
-
-  // FileSystemProvider implementation
-
-  watch(): vscode.Disposable {
-    // No watching needed for readonly files
-    return new vscode.Disposable(() => {});
-  }
-
-  stat(uri: vscode.Uri): vscode.FileStat {
-    const buffer = this.files.get(uri.toString());
-    if (!buffer) {
-      throw vscode.FileSystemError.FileNotFound(uri);
-    }
-
-    return {
-      type: vscode.FileType.File,
-      ctime: Date.now(),
-      mtime: Date.now(),
-      size: buffer.byteLength,
-    };
-  }
-
-  readDirectory(): Array<[string, vscode.FileType]> {
-    // Not needed for our use case
-    return [];
-  }
-
-  createDirectory(): void {
-    throw vscode.FileSystemError.NoPermissions('Readonly file system');
-  }
-
-  readFile(uri: vscode.Uri): Uint8Array {
-    const buffer = this.files.get(uri.toString());
-    if (!buffer) {
-      throw vscode.FileSystemError.FileNotFound(uri);
-    }
-    return buffer;
-  }
-
-  writeFile(
-    uri: vscode.Uri,
-    content: Uint8Array,
-    options: { create: boolean; overwrite: boolean },
-  ): void {
-    // Check if file exists
-    const exists = this.files.has(uri.toString());
-
-    // For readonly files, only allow creation, not modification
-    if (exists && !options.overwrite) {
-      throw vscode.FileSystemError.FileExists(uri);
-    }
-    if (!exists && !options.create) {
-      throw vscode.FileSystemError.FileNotFound(uri);
-    }
-
-    this.files.set(uri.toString(), content);
-    this.emitter.fire([
-      {
-        type: exists
-          ? vscode.FileChangeType.Changed
-          : vscode.FileChangeType.Created,
-        uri,
-      },
-    ]);
-  }
-
-  delete(uri: vscode.Uri): void {
-    if (!this.files.has(uri.toString())) {
-      throw vscode.FileSystemError.FileNotFound(uri);
-    }
-    this.files.delete(uri.toString());
-    this.emitter.fire([{ type: vscode.FileChangeType.Deleted, uri }]);
-  }
-
-  rename(): void {
-    throw vscode.FileSystemError.NoPermissions('Readonly file system');
-  }
-
-  /**
-   * Clear all cached files
-   */
-  clear(): void {
-    this.files.clear();
-  }
-
-  dispose(): void {
-    this.clear();
-    this.disposables.forEach((d) => d.dispose());
-    // Clear global instance on dispose
-    if (ReadonlyFileSystemProvider.instance === this) {
-      ReadonlyFileSystemProvider.instance = null;
-    }
-  }
-}
--- a/packages/vscode-ide-companion/src/types/acpTypes.ts
+++ b/packages/vscode-ide-companion/src/types/acpTypes.ts
@@ -48,35 +48,6 @@ export interface ContentBlock {
  uri?: string;
 }

-export interface UsageMetadata {
-  promptTokens?: number | null;
-  completionTokens?: number | null;
-  thoughtsTokens?: number | null;
-  totalTokens?: number | null;
-  cachedTokens?: number | null;
-}
-
-export interface SessionUpdateMeta {
-  usage?: UsageMetadata | null;
-  durationMs?: number | null;
-}
-
-export type AcpMeta = Record<string, unknown>;
-export type ModelId = string;
-
-export interface ModelInfo {
-  _meta?: AcpMeta | null;
-  description?: string | null;
-  modelId: ModelId;
-  name: string;
-}
-
-export interface SessionModelState {
-  _meta?: AcpMeta | null;
-  availableModels: ModelInfo[];
-  currentModelId: ModelId;
-}
-
 export interface UserMessageChunkUpdate extends BaseSessionUpdate {
  update: {
    sessionUpdate: 'user_message_chunk';
@@ -88,7 +59,6 @@ export interface AgentMessageChunkUpdate extends BaseSessionUpdate {
  update: {
    sessionUpdate: 'agent_message_chunk';
    content: ContentBlock;
-    _meta?: SessionUpdateMeta;
  };
 }

@@ -96,7 +66,6 @@ export interface AgentThoughtChunkUpdate extends BaseSessionUpdate {
  update: {
    sessionUpdate: 'agent_thought_chunk';
    content: ContentBlock;
-    _meta?: SessionUpdateMeta;
  };
 }

--- a/packages/vscode-ide-companion/src/types/chatTypes.ts
+++ b/packages/vscode-ide-companion/src/types/chatTypes.ts
@@ -3,7 +3,7 @@
 * Copyright 2025 Qwen Team
 * SPDX-License-Identifier: Apache-2.0
 */
-import type { AcpPermissionRequest, ModelInfo } from './acpTypes.js';
+import type { AcpPermissionRequest } from './acpTypes.js';
 import type { ApprovalModeValue } from './approvalModeValueTypes.js';

 export interface ChatMessage {
@@ -28,18 +28,6 @@ export interface ToolCallUpdateData {
  locations?: Array<{ path: string; line?: number | null }>;
 }

-export interface UsageStatsPayload {
-  usage?: {
-    promptTokens?: number | null;
-    completionTokens?: number | null;
-    thoughtsTokens?: number | null;
-    totalTokens?: number | null;
-    cachedTokens?: number | null;
-  } | null;
-  durationMs?: number | null;
-  tokenLimit?: number | null;
-}
-
 export interface QwenAgentCallbacks {
  onMessage?: (message: ChatMessage) => void;
  onStreamChunk?: (chunk: string) => void;
@@ -57,8 +45,6 @@ export interface QwenAgentCallbacks {
    }>;
  }) => void;
  onModeChanged?: (modeId: ApprovalModeValue) => void;
-  onUsageUpdate?: (stats: UsageStatsPayload) => void;
-  onModelInfo?: (info: ModelInfo) => void;
 }

 export interface ToolCallUpdate {
--- a/packages/vscode-ide-companion/src/utils/acpModelInfo.test.ts
+++ b/packages/vscode-ide-companion/src/utils/acpModelInfo.test.ts
@@ -1,77 +0,0 @@
-/**
- * @license
- * Copyright 2025 Qwen Team
- * SPDX-License-Identifier: Apache-2.0
- */
-
-import { describe, expect, it } from 'vitest';
-import { extractModelInfoFromNewSessionResult } from './acpModelInfo.js';
-
-describe('extractModelInfoFromNewSessionResult', () => {
-  it('extracts from NewSessionResponse.models (SessionModelState)', () => {
-    expect(
-      extractModelInfoFromNewSessionResult({
-        sessionId: 's',
-        models: {
-          currentModelId: 'qwen3-coder-plus',
-          availableModels: [
-            {
-              modelId: 'qwen3-coder-plus',
-              name: 'Qwen3 Coder Plus',
-              description: null,
-              _meta: { contextLimit: 123 },
-            },
-          ],
-        },
-      }),
-    ).toEqual({
-      modelId: 'qwen3-coder-plus',
-      name: 'Qwen3 Coder Plus',
-      description: null,
-      _meta: { contextLimit: 123 },
-    });
-  });
-
-  it('skips invalid model entries and returns first valid one', () => {
-    expect(
-      extractModelInfoFromNewSessionResult({
-        models: {
-          currentModelId: 'ok',
-          availableModels: [
-            { name: '', modelId: '' },
-            { name: 'Ok', modelId: 'ok', _meta: { contextLimit: null } },
-          ],
-        },
-      }),
-    ).toEqual({ name: 'Ok', modelId: 'ok', _meta: { contextLimit: null } });
-  });
-
-  it('falls back to single `model` object', () => {
-    expect(
-      extractModelInfoFromNewSessionResult({
-        model: {
-          name: 'Single',
-          modelId: 'single',
-          _meta: { contextLimit: 999 },
-        },
-      }),
-    ).toEqual({
-      name: 'Single',
-      modelId: 'single',
-      _meta: { contextLimit: 999 },
-    });
-  });
-
-  it('falls back to legacy `modelInfo`', () => {
-    expect(
-      extractModelInfoFromNewSessionResult({
-        modelInfo: { name: 'legacy' },
-      }),
-    ).toEqual({ name: 'legacy', modelId: 'legacy' });
-  });
-
-  it('returns null when missing', () => {
-    expect(extractModelInfoFromNewSessionResult({})).toBeNull();
-    expect(extractModelInfoFromNewSessionResult(null)).toBeNull();
-  });
-});
--- a/packages/vscode-ide-companion/src/utils/acpModelInfo.ts
+++ b/packages/vscode-ide-companion/src/utils/acpModelInfo.ts
@@ -1,135 +0,0 @@
-/**
- * @license
- * Copyright 2025 Qwen Team
- * SPDX-License-Identifier: Apache-2.0
- */
-
-import type { AcpMeta, ModelInfo } from '../types/acpTypes.js';
-
-const asMeta = (value: unknown): AcpMeta | null | undefined => {
-  if (value === null) {
-    return null;
-  }
-  if (value && typeof value === 'object' && !Array.isArray(value)) {
-    return value as AcpMeta;
-  }
-  return undefined;
-};
-
-const normalizeModelInfo = (value: unknown): ModelInfo | null => {
-  if (!value || typeof value !== 'object') {
-    return null;
-  }
-
-  const obj = value as Record<string, unknown>;
-  const nameRaw = obj['name'];
-  const modelIdRaw = obj['modelId'];
-  const descriptionRaw = obj['description'];
-
-  const name = typeof nameRaw === 'string' ? nameRaw.trim() : '';
-  const modelId =
-    typeof modelIdRaw === 'string' && modelIdRaw.trim().length > 0
-      ? modelIdRaw.trim()
-      : name;
-
-  if (!modelId || modelId.trim().length === 0 || !name) {
-    return null;
-  }
-
-  const description =
-    typeof descriptionRaw === 'string' || descriptionRaw === null
-      ? descriptionRaw
-      : undefined;
-
-  const metaFromWire = asMeta(obj['_meta']);
-
-  // Back-compat: older implementations used `contextLimit` at the top-level.
-  const legacyContextLimit = obj['contextLimit'];
-  const contextLimit =
-    typeof legacyContextLimit === 'number' || legacyContextLimit === null
-      ? legacyContextLimit
-      : undefined;
-
-  let mergedMeta: AcpMeta | null | undefined = metaFromWire;
-  if (typeof contextLimit !== 'undefined') {
-    if (mergedMeta === null) {
-      mergedMeta = { contextLimit };
-    } else if (typeof mergedMeta === 'undefined') {
-      mergedMeta = { contextLimit };
-    } else {
-      mergedMeta = { ...mergedMeta, contextLimit };
-    }
-  }
-
-  return {
-    modelId,
-    name,
-    ...(typeof description !== 'undefined' ? { description } : {}),
-    ...(typeof mergedMeta !== 'undefined' ? { _meta: mergedMeta } : {}),
-  };
-};
-
-/**
- * Extract model info from ACP `session/new` result.
- *
- * Per Agent Client Protocol draft schema, NewSessionResponse includes `models`.
- * We also accept legacy shapes for compatibility.
- */
-export const extractModelInfoFromNewSessionResult = (
-  result: unknown,
-): ModelInfo | null => {
-  if (!result || typeof result !== 'object') {
-    return null;
-  }
-
-  const obj = result as Record<string, unknown>;
-
-  const models = obj['models'];
-
-  // ACP draft: NewSessionResponse.models is a SessionModelState object.
-  if (models && typeof models === 'object' && !Array.isArray(models)) {
-    const state = models as Record<string, unknown>;
-    const availableModels = state['availableModels'];
-    const currentModelId = state['currentModelId'];
-    if (Array.isArray(availableModels)) {
-      const normalizedModels = availableModels
-        .map(normalizeModelInfo)
-        .filter((m): m is ModelInfo => Boolean(m));
-      if (normalizedModels.length > 0) {
-        if (typeof currentModelId === 'string' && currentModelId.length > 0) {
-          const selected = normalizedModels.find(
-            (m) => m.modelId === currentModelId,
-          );
-          if (selected) {
-            return selected;
-          }
-        }
-        return normalizedModels[0];
-      }
-    }
-  }
-
-  // Legacy: some implementations returned `models` as a raw array.
-  if (Array.isArray(models)) {
-    for (const entry of models) {
-      const normalized = normalizeModelInfo(entry);
-      if (normalized) {
-        return normalized;
-      }
-    }
-  }
-
-  // Some implementations may return a single model object.
-  const model = normalizeModelInfo(obj['model']);
-  if (model) {
-    return model;
-  }
-
-  // Legacy: modelInfo on initialize; allow as a fallback.
-  const legacy = normalizeModelInfo(obj['modelInfo']);
-  if (legacy) {
-    return legacy;
-  }
-
-  return null;
-};
--- a/packages/vscode-ide-companion/src/utils/editorGroupUtils.ts
+++ b/packages/vscode-ide-companion/src/utils/editorGroupUtils.ts
@@ -53,40 +53,11 @@ export function findLeftGroupOfChatWebview(): vscode.ViewColumn | undefined {
  }
 }

-/**
- * Wait for a condition to become true, driven by tab-group change events.
- * Falls back to a timeout to avoid hanging forever.
- */
-function waitForTabGroupsCondition(
-  condition: () => boolean,
-  timeout: number = 2000,
-): Promise<boolean> {
-  if (condition()) {
-    return Promise.resolve(true);
-  }
-
-  return new Promise<boolean>((resolve) => {
-    const subscription = vscode.window.tabGroups.onDidChangeTabGroups(() => {
-      if (!condition()) {
-        return;
-      }
-      clearTimeout(timeoutHandle);
-      subscription.dispose();
-      resolve(true);
-    });
-
-    const timeoutHandle = setTimeout(() => {
-      subscription.dispose();
-      resolve(false);
-    }, timeout);
-  });
-}
-
 /**
 * Ensure there is an editor group directly to the left of the Qwen chat webview.
 * - If one exists, return its ViewColumn.
 * - If none exists, focus the chat panel and create a new group on its left,
- *   then return the new group's ViewColumn.
+ *   then return the new group's ViewColumn (which equals the chat's previous column).
 * - If the chat webview cannot be located, returns undefined.
 */
 export async function ensureLeftGroupOfChatWebview(): Promise<
@@ -116,7 +87,7 @@ export async function ensureLeftGroupOfChatWebview(): Promise<
    return undefined;
  }

-  const initialGroupCount = vscode.window.tabGroups.all.length;
+  const previousChatColumn = webviewGroup.viewColumn;

  // Make the chat group active by revealing the panel
  try {
@@ -133,22 +104,6 @@ export async function ensureLeftGroupOfChatWebview(): Promise<
    return undefined;
  }

-  // Wait for the new group to actually be created (check that group count increased)
-  const groupCreated = await waitForTabGroupsCondition(
-    () => vscode.window.tabGroups.all.length > initialGroupCount,
-    1000, // 1 second timeout
-  );
-
-  if (!groupCreated) {
-    // Fallback if group creation didn't complete in time
-    return vscode.ViewColumn.One;
-  }
-
-  // After creating a new group to the left, the new group takes ViewColumn.One
-  // and all existing groups shift right. So the new left group is always ViewColumn.One.
-  // However, to be safe, let's query for it again.
-  const newLeftGroup = findLeftGroupOfChatWebview();
-
  // Restore focus to chat (optional), so we don't disturb user focus
  try {
    await vscode.commands.executeCommand(openChatCommand);
@@ -156,7 +111,6 @@ export async function ensureLeftGroupOfChatWebview(): Promise<
    // Ignore
  }

-  // If we successfully found the new left group, return it
-  // Otherwise, fallback to ViewColumn.One (the newly created group should be first)
-  return newLeftGroup ?? vscode.ViewColumn.One;
+  // The new left group's column equals the chat's previous column
+  return previousChatColumn;
 }
--- a/packages/vscode-ide-companion/src/webview/App.tsx
+++ b/packages/vscode-ide-companion/src/webview/App.tsx
@@ -27,7 +27,7 @@ import type { TextMessage } from './hooks/message/useMessageHandling.js';
 import type { ToolCallData } from './components/messages/toolcalls/ToolCall.js';
 import { PermissionDrawer } from './components/PermissionDrawer/PermissionDrawer.js';
 import { ToolCall } from './components/messages/toolcalls/ToolCall.js';
-import { hasToolCallOutput } from './utils/utils.js';
+import { hasToolCallOutput } from './components/messages/toolcalls/shared/utils.js';
 import { EmptyState } from './components/layout/EmptyState.js';
 import { Onboarding } from './components/layout/Onboarding.js';
 import { type CompletionItem } from '../types/completionItemTypes.js';
@@ -45,12 +45,7 @@ import { SessionSelector } from './components/layout/SessionSelector.js';
 import { FileIcon, UserIcon } from './components/icons/index.js';
 import { ApprovalMode, NEXT_APPROVAL_MODE } from '../types/acpTypes.js';
 import type { ApprovalModeValue } from '../types/approvalModeValueTypes.js';
-import type { PlanEntry, UsageStatsPayload } from '../types/chatTypes.js';
-import type { ModelInfo } from '../types/acpTypes.js';
-import {
-  DEFAULT_TOKEN_LIMIT,
-  tokenLimit,
-} from '@qwen-code/qwen-code-core/src/core/tokenLimits.js';
+import type { PlanEntry } from '../types/chatTypes.js';

 export const App: React.FC = () => {
  const vscode = useVSCode();
@@ -75,8 +70,6 @@ export const App: React.FC = () => {
  const [planEntries, setPlanEntries] = useState<PlanEntry[]>([]);
  const [isAuthenticated, setIsAuthenticated] = useState<boolean | null>(null);
  const [isLoading, setIsLoading] = useState<boolean>(true); // Track if we're still initializing/loading
-  const [modelInfo, setModelInfo] = useState<ModelInfo | null>(null);
-  const [usageStats, setUsageStats] = useState<UsageStatsPayload | null>(null);
  const messagesEndRef = useRef<HTMLDivElement>(
    null,
  ) as React.RefObject<HTMLDivElement>;
@@ -167,48 +160,6 @@ export const App: React.FC = () => {

  const completion = useCompletionTrigger(inputFieldRef, getCompletionItems);

-  const contextUsage = useMemo(() => {
-    if (!usageStats && !modelInfo) {
-      return null;
-    }
-
-    const modelName =
-      modelInfo?.modelId && typeof modelInfo.modelId === 'string'
-        ? modelInfo.modelId
-        : modelInfo?.name && typeof modelInfo.name === 'string'
-          ? modelInfo.name
-          : undefined;
-
-    const derivedLimit =
-      modelName && modelName.length > 0 ? tokenLimit(modelName) : undefined;
-
-    const metaLimitRaw = modelInfo?._meta?.['contextLimit'];
-    const metaLimit =
-      typeof metaLimitRaw === 'number' || metaLimitRaw === null
-        ? metaLimitRaw
-        : undefined;
-
-    const limit =
-      usageStats?.tokenLimit ??
-      metaLimit ??
-      derivedLimit ??
-      DEFAULT_TOKEN_LIMIT;
-
-    const used = usageStats?.usage?.promptTokens ?? 0;
-    if (typeof limit !== 'number' || limit <= 0 || used < 0) {
-      return null;
-    }
-    const percentLeft = Math.max(
-      0,
-      Math.min(100, Math.round(((limit - used) / limit) * 100)),
-    );
-    return {
-      percentLeft,
-      usedTokens: used,
-      tokenLimit: limit,
-    };
-  }, [usageStats, modelInfo]);
-
  // Track a lightweight signature of workspace files to detect content changes even when length is unchanged
  const workspaceFilesSignature = useMemo(
    () =>
@@ -297,10 +248,6 @@ export const App: React.FC = () => {
    setInputText,
    setEditMode,
    setIsAuthenticated,
-    setUsageStats: (stats) => setUsageStats(stats ?? null),
-    setModelInfo: (info) => {
-      setModelInfo(info);
-    },
  });

  // Auto-scroll handling: keep the view pinned to bottom when new content arrives,
@@ -813,7 +760,6 @@ export const App: React.FC = () => {
          activeFileName={fileContext.activeFileName}
          activeSelection={fileContext.activeSelection}
          skipAutoActiveContext={skipAutoActiveContext}
-          contextUsage={contextUsage}
          onInputChange={setInputText}
          onCompositionStart={() => setIsComposing(true)}
          onCompositionEnd={() => setIsComposing(false)}
--- a/packages/vscode-ide-companion/src/webview/WebViewProvider.ts
+++ b/packages/vscode-ide-companion/src/webview/WebViewProvider.ts
@@ -118,20 +118,6 @@ export class WebViewProvider {
      });
    });

-    this.agentManager.onUsageUpdate((stats) => {
-      this.sendMessageToWebView({
-        type: 'usageStats',
-        data: stats,
-      });
-    });
-
-    this.agentManager.onModelInfo((info) => {
-      this.sendMessageToWebView({
-        type: 'modelInfo',
-        data: info,
-      });
-    });
-
    // Setup end-turn handler from ACP stopReason notifications
    this.agentManager.onEndTurn((reason) => {
      // Ensure WebView exits streaming state even if no explicit streamEnd was emitted elsewhere
--- a/packages/vscode-ide-companion/src/webview/components/Tooltip.tsx
+++ b/packages/vscode-ide-companion/src/webview/components/Tooltip.tsx
@@ -1,61 +0,0 @@
-/**
- * @license
- * Copyright 2025 Qwen Team
- * SPDX-License-Identifier: Apache-2.0
- */
-
-import type React from 'react';
-
-interface TooltipProps {
-  children: React.ReactNode;
-  content: React.ReactNode;
-  position?: 'top' | 'bottom' | 'left' | 'right';
-}
-
-export const Tooltip: React.FC<TooltipProps> = ({
-  children,
-  content,
-  position = 'top',
-}) => (
-  <div className="relative inline-block">
-    <div className="group relative">
-      {children}
-      <div
-        className={`
-          absolute z-50 px-2 py-1 text-xs rounded-md shadow-lg
-          bg-[var(--app-primary-background)] border border-[var(--app-input-border)]
-          text-[var(--app-primary-foreground)] whitespace-nowrap
-          opacity-0 group-hover:opacity-100 transition-opacity duration-150
-          -translate-x-1/2 left-1/2
-          ${
-            position === 'top'
-              ? '-translate-y-1 bottom-full mb-1'
-              : position === 'bottom'
-                ? 'translate-y-1 top-full mt-1'
-                : position === 'left'
-                  ? '-translate-x-full left-0 translate-y-[-50%] top-1/2'
-                  : 'translate-x-0 right-0 translate-y-[-50%] top-1/2'
-          }
-          pointer-events-none
-        `}
-      >
-        {content}
-        <div
-          className={`
-            absolute w-2 h-2 bg-[var(--app-primary-background)] border-l border-b border-[var(--app-input-border)]
-            -rotate-45
-            ${
-              position === 'top'
-                ? 'top-full left-1/2 -translate-x-1/2 -translate-y-1/2'
-                : position === 'bottom'
-                  ? 'bottom-full left-1/2 -translate-x-1/2 translate-y-1/2'
-                  : position === 'left'
-                    ? 'right-full top-1/2 translate-x-1/2 -translate-y-1/2'
-                    : 'left-full top-1/2 -translate-x-1/2 -translate-y-1/2'
-            }
-          `}
-        />
-      </div>
-    </div>
-  </div>
-);
--- a/packages/vscode-ide-companion/src/webview/components/layout/ContextIndicator.tsx
+++ b/packages/vscode-ide-companion/src/webview/components/layout/ContextIndicator.tsx
@@ -1,88 +0,0 @@
-/**
- * @license
- * Copyright 2025 Qwen Team
- * SPDX-License-Identifier: Apache-2.0
- */
-
-import type React from 'react';
-import { Tooltip } from '../Tooltip.js';
-
-interface ContextUsage {
-  percentLeft: number;
-  usedTokens: number;
-  tokenLimit: number;
-}
-
-interface ContextIndicatorProps {
-  contextUsage: ContextUsage | null;
-}
-
-export const ContextIndicator: React.FC<ContextIndicatorProps> = ({
-  contextUsage,
-}) => {
-  if (!contextUsage) {
-    return null;
-  }
-
-  // Calculate used percentage for the progress indicator
-  // contextUsage.percentLeft is the percentage remaining, so 100 - percentLeft = percent used
-  const percentUsed = 100 - contextUsage.percentLeft;
-  const percentFormatted = Math.max(0, Math.min(100, Math.round(percentUsed)));
-  const radius = 9;
-  const circumference = 2 * Math.PI * radius;
-  // To show the used portion, we need to offset the unused portion
-  // If 20% is used, we want to show 20% filled, so offset the remaining 80%
-  const dashOffset = ((100 - percentUsed) / 100) * circumference;
-  const formatNumber = (value: number) => {
-    if (value >= 1000) {
-      return `${(Math.round((value / 1000) * 10) / 10).toFixed(1)}k`;
-    }
-    return Math.round(value).toLocaleString();
-  };
-
-  // Create tooltip content with proper formatting
-  const tooltipContent = (
-    <div className="flex flex-col gap-1">
-      <div className="font-medium">
-        {percentFormatted}% • {formatNumber(contextUsage.usedTokens)} /{' '}
-        {formatNumber(contextUsage.tokenLimit)} context used
-      </div>
-    </div>
-  );
-
-  return (
-    <Tooltip content={tooltipContent} position="top">
-      <button
-        className="btn-icon-compact"
-        aria-label={`${percentFormatted}% • ${formatNumber(contextUsage.usedTokens)} / ${formatNumber(contextUsage.tokenLimit)} context used`}
-      >
-        <svg viewBox="0 0 24 24" aria-hidden="true" role="presentation">
-          <circle
-            className="context-indicator__track"
-            cx="12"
-            cy="12"
-            r={radius}
-            fill="none"
-            stroke="currentColor"
-            opacity="0.2"
-          />
-          <circle
-            className="context-indicator__progress"
-            cx="12"
-            cy="12"
-            r={radius}
-            fill="none"
-            stroke="currentColor"
-            strokeWidth="2"
-            strokeDasharray={circumference}
-            strokeDashoffset={dashOffset}
-            style={{
-              transform: 'rotate(-90deg)',
-              transformOrigin: '50% 50%',
-            }}
-          />
-        </svg>
-      </button>
-    </Tooltip>
-  );
-};
--- a/packages/vscode-ide-companion/src/webview/components/layout/InputForm.tsx
+++ b/packages/vscode-ide-companion/src/webview/components/layout/InputForm.tsx
@@ -21,7 +21,6 @@ import { CompletionMenu } from '../layout/CompletionMenu.js';
 import type { CompletionItem } from '../../../types/completionItemTypes.js';
 import { getApprovalModeInfoFromString } from '../../../types/acpTypes.js';
 import type { ApprovalModeValue } from '../../../types/approvalModeValueTypes.js';
-import { ContextIndicator } from './ContextIndicator.js';

 interface InputFormProps {
  inputText: string;
@@ -37,11 +36,6 @@ interface InputFormProps {
  activeSelection: { startLine: number; endLine: number } | null;
  // Whether to auto-load the active editor selection/path into context
  skipAutoActiveContext: boolean;
-  contextUsage: {
-    percentLeft: number;
-    usedTokens: number;
-    tokenLimit: number;
-  } | null;
  onInputChange: (text: string) => void;
  onCompositionStart: () => void;
  onCompositionEnd: () => void;
@@ -102,7 +96,6 @@ export const InputForm: React.FC<InputFormProps> = ({
  activeFileName,
  activeSelection,
  skipAutoActiveContext,
-  contextUsage,
  onInputChange,
  onCompositionStart,
  onCompositionEnd,
@@ -247,9 +240,6 @@ export const InputForm: React.FC<InputFormProps> = ({
            {/* Spacer */}
            <div className="flex-1 min-w-0" />

-            {/* Context usage indicator */}
-            <ContextIndicator contextUsage={contextUsage} />
-
            {/* @yiliang114. closed temporarily */}
            {/* Thinking button */}
            {/* <button
--- a/packages/vscode-ide-companion/src/webview/components/messages/toolcalls/Bash/Bash.css
+++ b/packages/vscode-ide-companion/src/webview/components/messages/toolcalls/Bash/Bash.css
@@ -3,10 +3,10 @@
 * Copyright 2025 Qwen Team
 * SPDX-License-Identifier: Apache-2.0
 *
- * Bash tool call styles - Enhanced styling with semantic class names
+ * Execute tool call styles - Enhanced styling with semantic class names
 */

-/* Root container for bash tool call output */
+/* Root container for execute tool call output */
 .bash-toolcall-card {
  border: 0.5px solid var(--app-input-border);
  border-radius: 5px;
@@ -100,9 +100,3 @@
 .bash-toolcall-error-content {
  color: #c74e39;
 }
-
-/* Row with copy button */
-.bash-toolcall-row-with-copy {
-  position: relative;
-  grid-template-columns: max-content 1fr max-content;
-}
--- a/packages/vscode-ide-companion/src/webview/components/messages/toolcalls/Bash/Bash.tsx
+++ b/packages/vscode-ide-companion/src/webview/components/messages/toolcalls/Bash/Bash.tsx
@@ -9,10 +9,9 @@
 import type React from 'react';
 import type { BaseToolCallProps } from '../shared/types.js';
 import { ToolCallContainer } from '../shared/LayoutComponents.js';
-import { safeTitle, groupContent } from '../../../../utils/utils.js';
+import { safeTitle, groupContent } from '../shared/utils.js';
 import { useVSCode } from '../../../../hooks/useVSCode.js';
-import { createAndOpenTempFile } from '../../../../utils/diffUtils.js';
-import { CopyButton } from '../shared/copyUtils.js';
+import { createAndOpenTempFile } from '../../../../utils/tempFileManager.js';
 import './Bash.css';

 /**
@@ -38,14 +37,19 @@ export const ExecuteToolCall: React.FC<BaseToolCallProps> = ({ toolCall }) => {

  // Handle click on IN section
  const handleInClick = () => {
-    createAndOpenTempFile(vscode, inputCommand, `bash-input-${toolCallId}`);
+    createAndOpenTempFile(
+      vscode.postMessage,
+      inputCommand,
+      'bash-input',
+      '.sh',
+    );
  };

  // Handle click on OUT section
  const handleOutClick = () => {
    if (textOutputs.length > 0) {
      const output = textOutputs.join('\n');
-      createAndOpenTempFile(vscode, output, `bash-output-${toolCallId}`);
+      createAndOpenTempFile(vscode.postMessage, output, 'bash-output', '.txt');
    }
  };

@@ -80,7 +84,7 @@ export const ExecuteToolCall: React.FC<BaseToolCallProps> = ({ toolCall }) => {
          <div className="bash-toolcall-content">
            {/* IN row */}
            <div
-              className="bash-toolcall-row bash-toolcall-row-with-copy group"
+              className="bash-toolcall-row"
              onClick={handleInClick}
              style={{ cursor: 'pointer' }}
            >
@@ -88,7 +92,6 @@ export const ExecuteToolCall: React.FC<BaseToolCallProps> = ({ toolCall }) => {
              <div className="bash-toolcall-row-content">
                <pre className="bash-toolcall-pre">{inputCommand}</pre>
              </div>
-              <CopyButton text={inputCommand} />
            </div>

            {/* ERROR row */}
@@ -128,7 +131,7 @@ export const ExecuteToolCall: React.FC<BaseToolCallProps> = ({ toolCall }) => {
          <div className="bash-toolcall-content">
            {/* IN row */}
            <div
-              className="bash-toolcall-row bash-toolcall-row-with-copy group"
+              className="bash-toolcall-row"
              onClick={handleInClick}
              style={{ cursor: 'pointer' }}
            >
@@ -136,7 +139,6 @@ export const ExecuteToolCall: React.FC<BaseToolCallProps> = ({ toolCall }) => {
              <div className="bash-toolcall-row-content">
                <pre className="bash-toolcall-pre">{inputCommand}</pre>
              </div>
-              <CopyButton text={inputCommand} />
            </div>

            {/* OUT row */}
--- a/packages/vscode-ide-companion/src/webview/components/messages/toolcalls/Edit/EditToolCall.tsx
+++ b/packages/vscode-ide-companion/src/webview/components/messages/toolcalls/Edit/EditToolCall.tsx
@@ -11,7 +11,7 @@ import type { BaseToolCallProps } from '../shared/types.js';
 import {
  groupContent,
  mapToolStatusToContainerStatus,
-} from '../../../../utils/utils.js';
+} from '../shared/utils.js';
 import { FileLink } from '../../../layout/FileLink.js';
 import type { ToolCallContainerProps } from '../shared/LayoutComponents.js';

--- a/packages/vscode-ide-companion/src/webview/components/messages/toolcalls/Execute/Execute.css
+++ b/packages/vscode-ide-companion/src/webview/components/messages/toolcalls/Execute/Execute.css
@@ -61,7 +61,11 @@
 /* Truncated content styling */
 .execute-toolcall-row-content:not(.execute-toolcall-full) {
  max-height: 60px;
-  mask-image: linear-gradient(to bottom, var(--app-primary-background) 40px, transparent 60px);
+  mask-image: linear-gradient(
+    to bottom,
+    var(--app-primary-background) 40px,
+    transparent 60px
+  );
  overflow: hidden;
 }

@@ -83,6 +87,7 @@

 /* Output content with subtle styling */
 .execute-toolcall-output-subtle {
+  background-color: var(--app-code-background);
  white-space: pre;
  overflow-x: auto;
  max-width: 100%;
@@ -95,9 +100,3 @@
 .execute-toolcall-error-content {
  color: #c74e39;
 }
-
-/* Row with copy button */
-.execute-toolcall-row-with-copy {
-  position: relative;
-  grid-template-columns: max-content 1fr max-content;
-}
--- a/packages/vscode-ide-companion/src/webview/components/messages/toolcalls/Execute/Execute.tsx
+++ b/packages/vscode-ide-companion/src/webview/components/messages/toolcalls/Execute/Execute.tsx
@@ -8,12 +8,9 @@

 import type React from 'react';
 import type { BaseToolCallProps } from '../shared/types.js';
-import { safeTitle, groupContent } from '../../../../utils/utils.js';
+import { safeTitle, groupContent } from '../shared/utils.js';
 import './Execute.css';
 import type { ToolCallContainerProps } from '../shared/LayoutComponents.js';
-import { useVSCode } from '../../../../hooks/useVSCode.js';
-import { createAndOpenTempFile } from '../../../../utils/diffUtils.js';
-import { CopyButton } from '../shared/copyUtils.js';

 export const ToolCallContainer: React.FC<ToolCallContainerProps> = ({
  label,
@@ -51,7 +48,6 @@ export const ExecuteToolCall: React.FC<BaseToolCallProps> = ({ toolCall }) => {
  const commandText = safeTitle(
    (rawInput as Record<string, unknown>)?.description || title,
  );
-  const vscode = useVSCode();

  // Group content by type
  const { textOutputs, errors } = groupContent(content);
@@ -65,19 +61,6 @@ export const ExecuteToolCall: React.FC<BaseToolCallProps> = ({ toolCall }) => {
    inputCommand = rawInput;
  }

-  // Handle click on IN section
-  const handleInClick = () => {
-    createAndOpenTempFile(vscode, inputCommand, `execute-input-${toolCallId}`);
-  };
-
-  // Handle click on OUT section
-  const handleOutClick = () => {
-    if (textOutputs.length > 0) {
-      const output = textOutputs.join('\n');
-      createAndOpenTempFile(vscode, output, `execute-output-${toolCallId}`);
-    }
-  };
-
  // Map tool status to container status for proper bullet coloring
  const containerStatus:
    | 'success'
@@ -109,16 +92,11 @@ export const ExecuteToolCall: React.FC<BaseToolCallProps> = ({ toolCall }) => {
        <div className="execute-toolcall-card">
          <div className="execute-toolcall-content">
            {/* IN row */}
-            <div
-              className="execute-toolcall-row execute-toolcall-row-with-copy group"
-              onClick={handleInClick}
-              style={{ cursor: 'pointer' }}
-            >
+            <div className="execute-toolcall-row">
              <div className="execute-toolcall-label">IN</div>
              <div className="execute-toolcall-row-content">
                <pre className="execute-toolcall-pre">{inputCommand}</pre>
              </div>
-              <CopyButton text={inputCommand} />
            </div>

            {/* ERROR row */}
@@ -157,24 +135,15 @@ export const ExecuteToolCall: React.FC<BaseToolCallProps> = ({ toolCall }) => {
        <div className="execute-toolcall-card">
          <div className="execute-toolcall-content">
            {/* IN row */}
-            <div
-              className="execute-toolcall-row execute-toolcall-row-with-copy group"
-              onClick={handleInClick}
-              style={{ cursor: 'pointer' }}
-            >
+            <div className="execute-toolcall-row">
              <div className="execute-toolcall-label">IN</div>
              <div className="execute-toolcall-row-content">
                <pre className="execute-toolcall-pre">{inputCommand}</pre>
              </div>
-              <CopyButton text={inputCommand} />
            </div>

            {/* OUT row */}
-            <div
-              className="execute-toolcall-row"
-              onClick={handleOutClick}
-              style={{ cursor: 'pointer' }}
-            >
+            <div className="execute-toolcall-row">
              <div className="execute-toolcall-label">OUT</div>
              <div className="execute-toolcall-row-content">
                <div className="execute-toolcall-output-subtle">
@@ -195,11 +164,7 @@ export const ExecuteToolCall: React.FC<BaseToolCallProps> = ({ toolCall }) => {
      status={containerStatus}
      toolCallId={toolCallId}
    >
-      <div
-        className="inline-flex text-[var(--app-secondary-foreground)] text-[0.85em] opacity-70 mt-[2px] mb-[2px] flex-row items-start w-full gap-1"
-        onClick={handleInClick}
-        style={{ cursor: 'pointer' }}
-      >
+      <div className="inline-flex text-[var(--app-secondary-foreground)] text-[0.85em] opacity-70 mt-[2px] mb-[2px] flex-row items-start w-full gap-1">
        <span className="flex-shrink-0 relative top-[-0.1em]">⎿</span>
        <span className="flex-shrink-0 w-full">{commandText}</span>
      </div>
--- a/packages/vscode-ide-companion/src/webview/components/messages/toolcalls/GenericToolCall.tsx
+++ b/packages/vscode-ide-companion/src/webview/components/messages/toolcalls/GenericToolCall.tsx
@@ -14,7 +14,7 @@ import {
  ToolCallRow,
  LocationsList,
 } from './shared/LayoutComponents.js';
-import { safeTitle, groupContent } from '../../../utils/utils.js';
+import { safeTitle, groupContent } from './shared/utils.js';

 /**
 * Generic tool call component that can display any tool call type
--- a/packages/vscode-ide-companion/src/webview/components/messages/toolcalls/Read/ReadToolCall.tsx
+++ b/packages/vscode-ide-companion/src/webview/components/messages/toolcalls/Read/ReadToolCall.tsx
@@ -12,7 +12,7 @@ import type { BaseToolCallProps } from '../shared/types.js';
 import {
  groupContent,
  mapToolStatusToContainerStatus,
-} from '../../../../utils/utils.js';
+} from '../shared/utils.js';
 import { FileLink } from '../../../layout/FileLink.js';
 import { useVSCode } from '../../../../hooks/useVSCode.js';
 import { handleOpenDiff } from '../../../../utils/diffUtils.js';
--- a/packages/vscode-ide-companion/src/webview/components/messages/toolcalls/Search/SearchToolCall.tsx
+++ b/packages/vscode-ide-companion/src/webview/components/messages/toolcalls/Search/SearchToolCall.tsx
@@ -13,7 +13,7 @@ import {
  safeTitle,
  groupContent,
  mapToolStatusToContainerStatus,
-} from '../../../../utils/utils.js';
+} from '../shared/utils.js';

 /**
 * Specialized component for Search tool calls
@@ -195,7 +195,7 @@ export const SearchToolCall: React.FC<BaseToolCallProps> = ({
        isLast={isLast}
      >
        <div className="flex flex-col">
-          {textOutputs.map((text: string, index: number) => (
+          {textOutputs.map((text, index) => (
            <div
              key={index}
              className="inline-flex text-[var(--app-secondary-foreground)] text-[0.85em] opacity-70 mt-[2px] mb-[2px] flex-row items-start w-full gap-1"
--- a/packages/vscode-ide-companion/src/webview/components/messages/toolcalls/Think/ThinkToolCall.tsx
+++ b/packages/vscode-ide-companion/src/webview/components/messages/toolcalls/Think/ThinkToolCall.tsx
@@ -13,7 +13,7 @@ import {
  ToolCallCard,
  ToolCallRow,
 } from '../shared/LayoutComponents.js';
-import { groupContent } from '../../../../utils/utils.js';
+import { groupContent } from '../shared/utils.js';

 /**
 * Specialized component for Think tool calls
--- a/packages/vscode-ide-companion/src/webview/components/messages/toolcalls/UpdatedPlan/UpdatedPlanToolCall.tsx
+++ b/packages/vscode-ide-companion/src/webview/components/messages/toolcalls/UpdatedPlan/UpdatedPlanToolCall.tsx
@@ -9,7 +9,7 @@
 import type React from 'react';
 import type { BaseToolCallProps } from '../shared/types.js';
 import type { ToolCallContainerProps } from '../shared/LayoutComponents.js';
-import { groupContent, safeTitle } from '../../../../utils/utils.js';
+import { groupContent, safeTitle } from '../shared/utils.js';
 import { CheckboxDisplay } from './CheckboxDisplay.js';
 import type { PlanEntry } from '../../../../../types/chatTypes.js';

--- a/packages/vscode-ide-companion/src/webview/components/messages/toolcalls/Write/WriteToolCall.tsx
+++ b/packages/vscode-ide-companion/src/webview/components/messages/toolcalls/Write/WriteToolCall.tsx
@@ -12,7 +12,7 @@ import { ToolCallContainer } from '../shared/LayoutComponents.js';
 import {
  groupContent,
  mapToolStatusToContainerStatus,
-} from '../../../../utils/utils.js';
+} from '../shared/utils.js';
 import { FileLink } from '../../../layout/FileLink.js';

 /**
--- a/packages/vscode-ide-companion/src/webview/components/messages/toolcalls/index.tsx
+++ b/packages/vscode-ide-companion/src/webview/components/messages/toolcalls/index.tsx
@@ -8,7 +8,7 @@

 import type React from 'react';
 import type { BaseToolCallProps } from './shared/types.js';
-import { shouldShowToolCall } from '../../../utils/utils.js';
+import { shouldShowToolCall } from './shared/utils.js';
 import { GenericToolCall } from './GenericToolCall.js';
 import { ReadToolCall } from './Read/ReadToolCall.js';
 import { WriteToolCall } from './Write/WriteToolCall.js';
--- a/Show More
+++ b/Show More