From ea287fae98c59867bccc28045979c083400428aa Mon Sep 17 00:00:00 2001
From: "mingholy.lmh" <mingholy.lmh@alibaba-inc.com>
Date: Mon, 29 Dec 2025 15:37:08 +0800
Subject: [PATCH] docs: add reasoning configuration and sampling parameters
 instructions

---
 docs/users/configuration/settings.md      | 26 +++++++++++++++
 packages/cli/src/config/settingsSchema.ts | 39 +++++++++++++++++++++++
 2 files changed, 65 insertions(+)

diff --git a/docs/users/configuration/settings.md b/docs/users/configuration/settings.md
index 3e87985b8..f2b156be0 100644
--- a/docs/users/configuration/settings.md
+++ b/docs/users/configuration/settings.md
@@ -124,12 +124,38 @@ Settings are organized into categories. All settings should be placed within the
         "temperature": 0.2,
         "top_p": 0.8,
         "max_tokens": 1024
+      },
+      "reasoning": {
+        "effort": "medium"
       }
     }
   }
 }
 ```
 
+**Reasoning Configuration:**
+
+The `reasoning` field controls reasoning behavior for models that support it:
+
+- Set to `false` to disable reasoning entirely
+- Set to an object with `effort` field to enable reasoning with a specific effort level:
+  - `"low"`: Minimal reasoning effort
+  - `"medium"`: Balanced reasoning effort (default)
+  - `"high"`: Maximum reasoning effort
+- Optionally include `budget_tokens` to limit reasoning token usage
+
+Example to disable reasoning:
+
+```json
+{
+  "model": {
+    "generationConfig": {
+      "reasoning": false
+    }
+  }
+}
+```
+
 **model.openAILoggingDir examples:**
 
 - `"~/qwen-logs"` - Logs to `~/qwen-logs` directory
diff --git a/packages/cli/src/config/settingsSchema.ts b/packages/cli/src/config/settingsSchema.ts
index 2fe467ba9..ab151b67d 100644
--- a/packages/cli/src/config/settingsSchema.ts
+++ b/packages/cli/src/config/settingsSchema.ts
@@ -675,6 +675,45 @@ const SETTINGS_SCHEMA = {
               { value: 'openapi_30', label: 'OpenAPI 3.0 Strict' },
             ],
           },
+          samplingParams: {
+            type: 'object',
+            label: 'Sampling Parameters',
+            category: 'Generation Configuration',
+            requiresRestart: false,
+            default: undefined as
+              | {
+                  top_p?: number;
+                  top_k?: number;
+                  repetition_penalty?: number;
+                  presence_penalty?: number;
+                  frequency_penalty?: number;
+                  temperature?: number;
+                  max_tokens?: number;
+                }
+              | undefined,
+            description: 'Sampling parameters for content generation.',
+            parentKey: 'generationConfig',
+            childKey: 'samplingParams',
+            showInDialog: false,
+          },
+          reasoning: {
+            type: 'object',
+            label: 'Reasoning Configuration',
+            category: 'Generation Configuration',
+            requiresRestart: false,
+            default: undefined as
+              | false
+              | {
+                  effort?: 'low' | 'medium' | 'high';
+                  budget_tokens?: number;
+                }
+              | undefined,
+            description:
+              'Reasoning configuration for models that support reasoning. Set to false to disable reasoning, or provide an object with effort level and optional token budget.',
+            parentKey: 'generationConfig',
+            childKey: 'reasoning',
+            showInDialog: false,
+          },
         },
       },
     },