refactor: 移除 attribution header 中的 per-message fingerprint 以提升缓存稳定性

getAttributionHeader() 不再接收 fingerprint 参数，cc_version 仅用
MACRO.VERSION。原来 cc_version=2.1.888.abc 中 abc 是 per-message 的
fingerprint，每次请求都变，即使 header 标记为 cacheScope:null 也不利于
未来的缓存优化。

- system.ts: 移除 fingerprint 参数，cc_version 仅用 MACRO.VERSION
- claude.ts: 移除 computeFingerprintFromMessages import 及调用
- sideQuery.ts: 移除 fingerprint 计算、extractFirstUserMessageText 及
  computeFingerprint import

Co-Authored-By: deepseek-v4-pro <deepseek-ai@claude-code-best.win>

Author: claude-code-best

src/constants/system.ts

@@ -58,7 +58,7 @@ function isAttributionHeaderEnabled(): boolean {
 
 /**
  * Get attribution header for API requests.
- * Returns a header string with cc_version (including fingerprint) and cc_entrypoint.
+ * Returns a header string with cc_version and cc_entrypoint.
  * Enabled by default, can be disabled via env var or GrowthBook killswitch.
  *
  * When NATIVE_CLIENT_ATTESTATION is enabled, includes a `cch=00000` placeholder.
@@ -70,22 +70,18 @@ function isAttributionHeaderEnabled(): boolean {
  * We use a placeholder (instead of injecting from Zig) because same-length
  * replacement avoids Content-Length changes and buffer reallocation.
  */
-export function getAttributionHeader(fingerprint: string): string {
+export function getAttributionHeader(): string {
   if (!isAttributionHeaderEnabled()) {
     return ''
   }
 
-  const version = `${MACRO.VERSION}.${fingerprint}`
+  const version = MACRO.VERSION
   const entrypoint = process.env.CLAUDE_CODE_ENTRYPOINT ?? 'unknown'
 
   // cch=00000 placeholder is overwritten by Bun's HTTP stack with attestation token
   const cch = feature('NATIVE_CLIENT_ATTESTATION') ? ' cch=00000;' : ''
   // cc_workload: turn-scoped hint so the API can route e.g. cron-initiated
-  // requests to a lower QoS pool. Absent = interactive default. Safe re:
-  // fingerprint (computed from msg chars + version only, line 78 above) and
-  // cch attestation (placeholder overwritten in serialized body bytes after
-  // this string is built). Server _parse_cc_header tolerates unknown extra
-  // fields so old API deploys silently ignore this.
+  // requests to a lower QoS pool. Absent = interactive default.
   const workload = getWorkload()
   const workloadPair = workload ? ` cc_workload=${workload};` : ''
   const header = `x-anthropic-billing-header: cc_version=${version}; cc_entrypoint=${entrypoint};${cch}${workloadPair}`

src/services/api/claude.ts

@@ -72,7 +72,6 @@ import {
 import { resolveAppliedEffort } from '../../utils/effort.js'
 import { isEnvTruthy } from '../../utils/envUtils.js'
 import { errorMessage } from '../../utils/errors.js'
-import { computeFingerprintFromMessages } from '../../utils/fingerprint.js'
 import { captureAPIRequest, logError } from '../../utils/log.js'
 import {
   createAssistantAPIErrorMessage,
@@ -1386,11 +1385,6 @@ async function* queryModel(
     postNormalizedMessageCount: messagesForAPI.length,
   })
 
-  // Compute fingerprint from first user message for attribution.
-  // Must run BEFORE injecting synthetic messages (e.g. deferred tool names)
-  // so the fingerprint reflects the actual user input.
-  const fingerprint = computeFingerprintFromMessages(messagesForAPI)
-
   // When the delta attachment is enabled, deferred tools are announced
   // via persisted deferred_tools_delta attachments instead of this
   // ephemeral prepend (which busts cache whenever the pool changes).
@@ -1438,7 +1432,7 @@ async function* queryModel(
   // filter(Boolean) works by converting each element to a boolean - empty strings become false and are filtered out.
   systemPrompt = asSystemPrompt(
     [
-      getAttributionHeader(fingerprint),
+      getAttributionHeader(),
       getCLISyspromptPrefix({
         isNonInteractive: options.isNonInteractiveSession,
         hasAppendSystemPrompt: options.hasAppendSystemPrompt,

src/utils/sideQuery.ts

@@ -30,7 +30,6 @@ import {
 import { getModelBetas, modelSupportsStructuredOutputs } from './betas.js'
 import { logForDebugging } from './debug.js'
 import { errorMessage } from './errors.js'
-import { computeFingerprint } from './fingerprint.js'
 import { getAPIProvider } from './model/providers.js'
 import { normalizeModelStringForAPI } from './model/model.js'
 import { getOpenAIClient } from '../services/api/openai/client.js'
@@ -97,21 +96,6 @@ export type SideQueryOptions = {
   optional?: boolean
 }
 
-/**
- * Extract text from first user message for fingerprint computation.
- */
-function extractFirstUserMessageText(messages: MessageParam[]): string {
-  const firstUserMessage = messages.find(m => m.role === 'user')
-  if (!firstUserMessage) return ''
-
-  const content = firstUserMessage.content
-  if (typeof content === 'string') return content
-
-  // Array of content blocks - find first text block
-  const textBlock = content.find(block => block.type === 'text')
-  return textBlock?.type === 'text' ? textBlock.text : ''
-}
-
 /**
  * Extract system prompt text from the `system` option.
  */
@@ -219,12 +203,7 @@ export async function sideQuery(opts: SideQueryOptions): Promise<BetaMessage> {
     betas.push(STRUCTURED_OUTPUTS_BETA_HEADER)
   }
 
-  // Extract first user message text for fingerprint
-  const messageText = extractFirstUserMessageText(messages)
-
-  // Compute fingerprint for OAuth attribution
-  const fingerprint = computeFingerprint(messageText, MACRO.VERSION)
-  const attributionHeader = getAttributionHeader(fingerprint)
+  const attributionHeader = getAttributionHeader()
 
   // Build system as array to keep attribution header in its own block
   // (prevents server-side parsing from including system content in cc_entrypoint)