mirror of
https://github.com/QwenLM/qwen-code.git
synced 2025-12-20 16:57:46 +00:00
Clearcut logging - initial implementation (#1274)
Flag-guarded initial implementation of a clearcut logger to collect telemetry data and send it to Concord for dashboards, etc.
This commit is contained in:
owenofbrien
GitHub
338
packages/core/src/telemetry/clearcut-logger/clearcut-logger.ts
Normal file
338
packages/core/src/telemetry/clearcut-logger/clearcut-logger.ts
Normal file
@@ -0,0 +1,338 @@
|
||||
/**
|
||||
* @license
|
||||
* Copyright 2025 Google LLC
|
||||
* SPDX-License-Identifier: Apache-2.0
|
||||
*/
|
||||
|
||||
import { Buffer } from 'buffer';
|
||||
import * as https from 'https';
|
||||
import {
|
||||
StartSessionEvent,
|
||||
EndSessionEvent,
|
||||
UserPromptEvent,
|
||||
ToolCallEvent,
|
||||
ApiRequestEvent,
|
||||
ApiResponseEvent,
|
||||
ApiErrorEvent,
|
||||
} from '../types.js';
|
||||
import { EventMetadataKey } from './event-metadata-key.js';
|
||||
import { Config } from '../../config/config.js';
|
||||
import { getPersistentUserId } from '../../utils/user_id.js';
|
||||
|
||||
const start_session_event_name = 'start_session';
|
||||
const new_prompt_event_name = 'new_prompt';
|
||||
const tool_call_event_name = 'tool_call';
|
||||
const api_request_event_name = 'api_request';
|
||||
const api_response_event_name = 'api_response';
|
||||
const api_error_event_name = 'api_error';
|
||||
const end_session_event_name = 'end_session';
|
||||
|
||||
export interface LogResponse {
|
||||
nextRequestWaitMs?: number;
|
||||
}
|
||||
|
||||
// Singleton class for batch posting log events to Clearcut. When a new event comes in, the elapsed time
|
||||
// is checked and events are flushed to Clearcut if at least a minute has passed since the last flush.
|
||||
export class ClearcutLogger {
|
||||
private static instance: ClearcutLogger;
|
||||
private config?: Config;
|
||||
// eslint-disable-next-line @typescript-eslint/no-explicit-any -- Clearcut expects this format.
|
||||
private readonly events: any = [];
|
||||
private last_flush_time: number = Date.now();
|
||||
private flush_interval_ms: number = 1000 * 60; // Wait at least a minute before flushing events.
|
||||
|
||||
private constructor(config?: Config) {
|
||||
this.config = config;
|
||||
}
|
||||
|
||||
static getInstance(config?: Config): ClearcutLogger | undefined {
|
||||
if (config === undefined || config?.getDisableDataCollection())
|
||||
return undefined;
|
||||
if (!ClearcutLogger.instance) {
|
||||
ClearcutLogger.instance = new ClearcutLogger(config);
|
||||
}
|
||||
return ClearcutLogger.instance;
|
||||
}
|
||||
|
||||
// eslint-disable-next-line @typescript-eslint/no-explicit-any -- Clearcut expects this format.
|
||||
enqueueLogEvent(event: any): void {
|
||||
this.events.push([
|
||||
{
|
||||
event_time_ms: Date.now(),
|
||||
source_extension_json: JSON.stringify(event),
|
||||
},
|
||||
]);
|
||||
}
|
||||
|
||||
createLogEvent(name: string, data: Map<EventMetadataKey, string>): object {
|
||||
return {
|
||||
Application: 'GEMINI_CLI',
|
||||
event_name: name,
|
||||
client_install_id: getPersistentUserId(),
|
||||
event_metadata: [data] as object[],
|
||||
};
|
||||
}
|
||||
|
||||
flushIfNeeded(): void {
|
||||
if (Date.now() - this.last_flush_time < this.flush_interval_ms) {
|
||||
return;
|
||||
}
|
||||
|
||||
this.flushToClearcut();
|
||||
this.last_flush_time = Date.now();
|
||||
}
|
||||
|
||||
flushToClearcut(): Promise<LogResponse> {
|
||||
return new Promise<Buffer>((resolve, reject) => {
|
||||
const request = [
|
||||
{
|
||||
log_source_name: 'CONCORD',
|
||||
request_time_ms: Date.now(),
|
||||
log_event: this.events,
|
||||
},
|
||||
];
|
||||
const body = JSON.stringify(request);
|
||||
const options = {
|
||||
hostname: 'play.googleapis.com',
|
||||
path: '/log',
|
||||
method: 'POST',
|
||||
headers: { 'Content-Length': Buffer.byteLength(body) },
|
||||
};
|
||||
const bufs: Buffer[] = [];
|
||||
const req = https.request(options, (res) => {
|
||||
res.on('data', (buf) => bufs.push(buf));
|
||||
res.on('end', () => {
|
||||
resolve(Buffer.concat(bufs));
|
||||
});
|
||||
});
|
||||
req.on('error', (e) => {
|
||||
reject(e);
|
||||
});
|
||||
req.end(body);
|
||||
}).then((buf: Buffer) => {
|
||||
try {
|
||||
this.events.length = 0;
|
||||
return this.decodeLogResponse(buf) || {};
|
||||
} catch (error: unknown) {
|
||||
console.error('Error flushing log events:', error);
|
||||
return {};
|
||||
}
|
||||
});
|
||||
}
|
||||
|
||||
// Visible for testing. Decodes protobuf-encoded response from Clearcut server.
|
||||
decodeLogResponse(buf: Buffer): LogResponse | undefined {
|
||||
// TODO(obrienowen): return specific errors to facilitate debugging.
|
||||
if (buf.length < 1) {
|
||||
return undefined;
|
||||
}
|
||||
|
||||
// The first byte of the buffer is `field<<3 | type`. We're looking for field
|
||||
// 1, with type varint, represented by type=0. If the first byte isn't 8, that
|
||||
// means field 1 is missing or the message is corrupted. Either way, we return
|
||||
// undefined.
|
||||
if (buf.readUInt8(0) !== 8) {
|
||||
return undefined;
|
||||
}
|
||||
|
||||
let ms = BigInt(0);
|
||||
let cont = true;
|
||||
|
||||
// In each byte, the most significant bit is the continuation bit. If it's
|
||||
// set, we keep going. The lowest 7 bits, are data bits. They are concatenated
|
||||
// in reverse order to form the final number.
|
||||
for (let i = 1; cont && i < buf.length; i++) {
|
||||
const byte = buf.readUInt8(i);
|
||||
ms |= BigInt(byte & 0x7f) << BigInt(7 * (i - 1));
|
||||
cont = (byte & 0x80) !== 0;
|
||||
}
|
||||
|
||||
if (cont) {
|
||||
// We have fallen off the buffer without seeing a terminating byte. The
|
||||
// message is corrupted.
|
||||
return undefined;
|
||||
}
|
||||
return {
|
||||
nextRequestWaitMs: Number(ms),
|
||||
};
|
||||
}
|
||||
|
||||
logStartSessionEvent(event: StartSessionEvent): void {
|
||||
const data: Map<EventMetadataKey, string> = new Map();
|
||||
|
||||
data.set(EventMetadataKey.GEMINI_CLI_START_SESSION_MODEL, event.model);
|
||||
data.set(
|
||||
EventMetadataKey.GEMINI_CLI_START_SESSION_EMBEDDING_MODEL,
|
||||
event.embedding_model,
|
||||
);
|
||||
data.set(
|
||||
EventMetadataKey.GEMINI_CLI_START_SESSION_SANDBOX,
|
||||
event.sandbox_enabled.toString(),
|
||||
);
|
||||
data.set(
|
||||
EventMetadataKey.GEMINI_CLI_START_SESSION_CORE_TOOLS,
|
||||
event.core_tools_enabled,
|
||||
);
|
||||
data.set(
|
||||
EventMetadataKey.GEMINI_CLI_START_SESSION_APPROVAL_MODE,
|
||||
event.approval_mode,
|
||||
);
|
||||
data.set(
|
||||
EventMetadataKey.GEMINI_CLI_START_SESSION_API_KEY_ENABLED,
|
||||
event.api_key_enabled.toString(),
|
||||
);
|
||||
data.set(
|
||||
EventMetadataKey.GEMINI_CLI_START_SESSION_VERTEX_API_ENABLED,
|
||||
event.vertex_ai_enabled.toString(),
|
||||
);
|
||||
data.set(
|
||||
EventMetadataKey.GEMINI_CLI_START_SESSION_DEBUG_MODE_ENABLED,
|
||||
event.debug_enabled.toString(),
|
||||
);
|
||||
data.set(
|
||||
EventMetadataKey.GEMINI_CLI_START_SESSION_MCP_SERVERS,
|
||||
event.mcp_servers,
|
||||
);
|
||||
data.set(
|
||||
EventMetadataKey.GEMINI_CLI_START_SESSION_TELEMETRY_ENABLED,
|
||||
event.telemetry_enabled.toString(),
|
||||
);
|
||||
data.set(
|
||||
EventMetadataKey.GEMINI_CLI_START_SESSION_TELEMETRY_LOG_USER_PROMPTS_ENABLED,
|
||||
event.telemetry_log_user_prompts_enabled.toString(),
|
||||
);
|
||||
|
||||
this.enqueueLogEvent(this.createLogEvent(start_session_event_name, data));
|
||||
this.flushIfNeeded();
|
||||
}
|
||||
|
||||
logNewPromptEvent(event: UserPromptEvent): void {
|
||||
const data: Map<EventMetadataKey, string> = new Map();
|
||||
|
||||
data.set(
|
||||
EventMetadataKey.GEMINI_CLI_USER_PROMPT_LENGTH,
|
||||
JSON.stringify(event.prompt_length),
|
||||
);
|
||||
|
||||
this.enqueueLogEvent(this.createLogEvent(new_prompt_event_name, data));
|
||||
this.flushIfNeeded();
|
||||
}
|
||||
|
||||
logToolCallEvent(event: ToolCallEvent): void {
|
||||
const data: Map<EventMetadataKey, string> = new Map();
|
||||
|
||||
data.set(EventMetadataKey.GEMINI_CLI_TOOL_CALL_NAME, event.function_name);
|
||||
data.set(
|
||||
EventMetadataKey.GEMINI_CLI_TOOL_CALL_DECISION,
|
||||
JSON.stringify(event.decision),
|
||||
);
|
||||
data.set(
|
||||
EventMetadataKey.GEMINI_CLI_TOOL_CALL_SUCCESS,
|
||||
JSON.stringify(event.success),
|
||||
);
|
||||
data.set(
|
||||
EventMetadataKey.GEMINI_CLI_TOOL_CALL_DURATION_MS,
|
||||
JSON.stringify(event.duration_ms),
|
||||
);
|
||||
data.set(
|
||||
EventMetadataKey.GEMINI_CLI_TOOL_ERROR_MESSAGE,
|
||||
JSON.stringify(event.error),
|
||||
);
|
||||
data.set(
|
||||
EventMetadataKey.GEMINI_CLI_TOOL_CALL_ERROR_TYPE,
|
||||
JSON.stringify(event.error_type),
|
||||
);
|
||||
|
||||
this.enqueueLogEvent(this.createLogEvent(tool_call_event_name, data));
|
||||
this.flushIfNeeded();
|
||||
}
|
||||
|
||||
logApiRequestEvent(event: ApiRequestEvent): void {
|
||||
const data: Map<EventMetadataKey, string> = new Map();
|
||||
|
||||
data.set(EventMetadataKey.GEMINI_CLI_API_REQUEST_MODEL, event.model);
|
||||
|
||||
this.enqueueLogEvent(this.createLogEvent(api_request_event_name, data));
|
||||
this.flushIfNeeded();
|
||||
}
|
||||
|
||||
logApiResponseEvent(event: ApiResponseEvent): void {
|
||||
const data: Map<EventMetadataKey, string> = new Map();
|
||||
|
||||
data.set(EventMetadataKey.GEMINI_CLI_API_RESPONSE_MODEL, event.model);
|
||||
data.set(
|
||||
EventMetadataKey.GEMINI_CLI_API_RESPONSE_STATUS_CODE,
|
||||
JSON.stringify(event.status_code),
|
||||
);
|
||||
data.set(
|
||||
EventMetadataKey.GEMINI_CLI_API_RESPONSE_DURATION_MS,
|
||||
JSON.stringify(event.duration_ms),
|
||||
);
|
||||
data.set(
|
||||
EventMetadataKey.GEMINI_CLI_API_ERROR_MESSAGE,
|
||||
JSON.stringify(event.error),
|
||||
);
|
||||
data.set(
|
||||
EventMetadataKey.GEMINI_CLI_API_RESPONSE_INPUT_TOKEN_COUNT,
|
||||
JSON.stringify(event.input_token_count),
|
||||
);
|
||||
data.set(
|
||||
EventMetadataKey.GEMINI_CLI_API_RESPONSE_OUTPUT_TOKEN_COUNT,
|
||||
JSON.stringify(event.output_token_count),
|
||||
);
|
||||
data.set(
|
||||
EventMetadataKey.GEMINI_CLI_API_RESPONSE_CACHED_TOKEN_COUNT,
|
||||
JSON.stringify(event.cached_content_token_count),
|
||||
);
|
||||
data.set(
|
||||
EventMetadataKey.GEMINI_CLI_API_RESPONSE_THINKING_TOKEN_COUNT,
|
||||
JSON.stringify(event.thoughts_token_count),
|
||||
);
|
||||
data.set(
|
||||
EventMetadataKey.GEMINI_CLI_API_RESPONSE_TOOL_TOKEN_COUNT,
|
||||
JSON.stringify(event.tool_token_count),
|
||||
);
|
||||
|
||||
this.enqueueLogEvent(this.createLogEvent(api_response_event_name, data));
|
||||
this.flushIfNeeded();
|
||||
}
|
||||
|
||||
logApiErrorEvent(event: ApiErrorEvent): void {
|
||||
const data: Map<EventMetadataKey, string> = new Map();
|
||||
|
||||
data.set(EventMetadataKey.GEMINI_CLI_API_ERROR_MODEL, event.model);
|
||||
data.set(
|
||||
EventMetadataKey.GEMINI_CLI_API_ERROR_TYPE,
|
||||
JSON.stringify(event.error_type),
|
||||
);
|
||||
data.set(
|
||||
EventMetadataKey.GEMINI_CLI_API_ERROR_STATUS_CODE,
|
||||
JSON.stringify(event.status_code),
|
||||
);
|
||||
data.set(
|
||||
EventMetadataKey.GEMINI_CLI_API_ERROR_DURATION_MS,
|
||||
JSON.stringify(event.duration_ms),
|
||||
);
|
||||
|
||||
this.enqueueLogEvent(this.createLogEvent(api_error_event_name, data));
|
||||
this.flushIfNeeded();
|
||||
}
|
||||
|
||||
logEndSessionEvent(event: EndSessionEvent): void {
|
||||
const data: Map<EventMetadataKey, string> = new Map();
|
||||
|
||||
data.set(
|
||||
EventMetadataKey.GEMINI_CLI_END_SESSION_ID,
|
||||
event?.session_id?.toString() ?? '',
|
||||
);
|
||||
|
||||
this.enqueueLogEvent(this.createLogEvent(end_session_event_name, data));
|
||||
// Flush immediately on session end.
|
||||
this.flushToClearcut();
|
||||
}
|
||||
|
||||
shutdown() {
|
||||
const event = new EndSessionEvent(this.config);
|
||||
this.logEndSessionEvent(event);
|
||||
}
|
||||
}
|
||||
Reference in New Issue
Block a user