v6.9.41-patch

Changelog

• faad8e3 Add CPA Usage Keeper to README ecosystem list
• 736018a Add GPT-5.5 Codex model support
• 5fb7e5b Add PostForceRefreshTokens
• 29dba03 Comment out system message check in Qwen executor
• af210b7 Fix EOF retry logic for network errors (#15)
• 65e9e89 Fix missing response.completed.usage for late-usage OpenAI-compatible streams
• c10f8ae Fixed: router-for-me#2420
• 108895f Harden Responses SSE framing against partial chunk boundaries
• da3a498 Keep Claude Code compatibility work low-risk and reviewable
• 9b5ce8c Keep Claude builtin helpers aligned with the shared helper layout
• ab9ebea Merge PR router-for-me#2474
• b468ca7 Merge branch 'dev' of github.com:router-for-me/CLIProxyAPI into dev
• 3e78a8d Merge branch 'main' into dev
• a337ecf Merge branch 'router-for-me:main' into feat(vertex)/add-prefix-field
• d09dd4d Merge commit '15c2f274ea690c9a7c9db22f9f454af869db5375' into dev
• 0e4f189 Merge pull request router-for-me#1302 from dinhkarate/feat(vertex)/add-prefix-field
• c1818f1 Merge pull request router-for-me#1940 from Blue-B/fix/claude-interleaved-thinking-amp-gzip-budget
• ea43361 Merge pull request router-for-me#2121 from destinoantagonista-wq/main
• 000fcb1 Merge pull request router-for-me#2298 from snoyiatk/feat/add-gitstore-branch
• b851208 Merge pull request router-for-me#2332 from RaviTharuma/fix/claude-thinking-signature
• c422d16 Merge pull request router-for-me#2398 from 7RPH/fix/responses-sse-framing
• d267522 Merge pull request router-for-me#2403 from CharTyr/clean-pr
• 486cd4c Merge pull request router-for-me#2409 from sususu98/fix/tool-use-pairing-break
• 39dc455 Merge pull request router-for-me#2412 from sususu98/feat/signature-cache-toggle
• 7c3c2e9 Merge pull request router-for-me#2417 from CharTyr/fix/amp-streaming-thinking-regression
• faae9c2 Merge pull request router-for-me#2422 from MonsterQiu/fix/codex-compact-instructions
• b09b03e Merge pull request router-for-me#2424 from possible055/fix/websocket-transcript-replacement
• 1623194 Merge pull request router-for-me#2426 from xixiwenxuanhe/feature/antigravity-credits
• bc3a6e4 Merge pull request router-for-me#2434 from MonsterQiu/fix/codex-responses-null-instructions
• e783d0a Merge pull request router-for-me#2441 from MonsterQiu/issue-2421-alias-before-suspension
• b05f575 Merge pull request router-for-me#2444 from 0oAstro/fix/codex-nonstream-finish-reason-tool-calls
• 1c7003f Merge pull request router-for-me#2452 from Lucaszmv/fix-qwen-cli-v0.13.2
• 7170eee Merge pull request router-for-me#2454 from buddingnewinsights/add-proxypal-to-readme
• 66cd50f Merge pull request router-for-me#2468 from router-for-me/ip
• 474fb04 Merge pull request router-for-me#2476 from router-for-me/cherry-pick/pr-2438-to-dev
• e3eb048 Merge pull request router-for-me#2489 from Soein/upstream-pr
• a59e924 Merge pull request router-for-me#2490 from router-for-me/logs
• c3f8dc3 Merge pull request router-for-me#2491 from mpfo0106/feature/claude-code-safe-alignment-sentinels
• fcba912 Merge pull request router-for-me#2492 from davidwushi1145/main
• 42c062b Merge pull request router-for-me#2509 from adamhelfgott/fix-claude-thinking-temperature
• f389667 Merge pull request router-for-me#2513 from lonr-6/codex/fix-ws-custom-tool-repair-v2
• ae37ccf Merge pull request router-for-me#2520 from Arronlong/main
• 9eabdd0 Merge pull request router-for-me#2522 from aikins01/fix/strip-tool-use-signature
• 8430b28 Merge pull request router-for-me#2526 from rensumo/main
• 9f5bdfa Merge pull request router-for-me#2531 from jamestut/openai-vertex-token-usage-fix
• 2ee938a Merge pull request router-for-me#2535 from rensumo/main
• d1f6224 Merge pull request router-for-me#2569 from LucasInsight/fix/record-zero-usage
• 6a27bce Merge pull request router-for-me#2576 from zilianpn/fix/disable-cooling-auth-errors
• cad45ff Merge pull request router-for-me#2578 from LemonZuo/feat_socks5h
• 12b9671 Merge pull request router-for-me#2592 from router-for-me/tests
• b2c0cdf Merge pull request router-for-me#2621 from wykk-12138/fix/oauth-extra-usage-detection
• 2a97037 Merge pull request router-for-me#2670 from sususu98/feat/antigravity-prefer-prod-url
• 9ded75d Merge pull request router-for-me#2702 from AllenReder/docs/add-quota-inspector
• 26fd3ef Merge pull request router-for-me#2715 from sususu98/pr/antigravity-32mb-bypass-signatures
• 1dc3018 Merge pull request router-for-me#2716 from sususu98/pr/antigravity-bypass-log-noise
• 61e0447 Merge pull request router-for-me#2723 from sususu98/fix/drop-redacted-thinking-blocks
• c4459c4 Merge pull request router-for-me#2724 from sususu98/fix/skip-schema-cleanup-empty-tools
• 3dea1da Merge pull request router-for-me#2782 from sususu98/fix/strip-invalid-signature-thinking-blocks
• f56cf42 Merge pull request router-for-me#2800 from sususu98/fix/antigravity-max-output-tokens-cap
• d1508ca Merge pull request router-for-me#2816 from sususu98/feat/session-affinity
• ba454db Merge pull request router-for-me#2817 from sususu98/fix/antigravity-strip-billing-header
• 8ced7a5 Merge pull request router-for-me#2834 from muzhi1991/fix/openai-compat-host-header
• c6baa64 Merge pull request router-for-me#2892 from router-for-me/fix-provider
• e05abec Merge pull request router-for-me#2898 from octo-patch/fix/issue-2866-remove-obsolete-1m-beta-header
• 3444820 Merge pull request router-for-me#2939 from stringer07/fix/codex-stream-output-backfill
• 8eb56e5 Merge pull request router-for-me#2962 from MoYeRanqianzhi/feat/codex-image-generation-tool-injection
• 12195a2 Merge pull request router-for-me#2971 from sususu98/feat/antigravity-credits-fallback
• a325533 Merge pull request router-for-me#2972 from XYenon/feat/amp-thread-id
• 1576d14 Merge pull request router-for-me#2989 from ben-vargas/gpt-5-5-support
• 36cc762 Merge pull request router-for-me#3007 from sususu98/fix-antigravity-credits-stream-fallback
• c5bea6f Merge pull request router-for-me#3020 from Matthias319/fix/codex-error-classification
• be0fe6f Merge pull request router-for-me#3022 from Willxup/main
• 736ae61 Merge pull request router-for-me#3051 from philipbankier/fix/gpt55-free-tier-test
• 34027da Merge pull request router-for-me#3088 from sususu98/codex-claude-reasoning-signature
• b34f3be Merge pull request router-for-me#3114 from Xbang0222/fix/antigravity-credits-ua
• c4965be Merge pull request router-for-me#3117 from sususu98/fix/restore-antigravity-ua-handling
• 96056d0 Merge remote-tracking branch 'upstream/main' into fix/oauth-extra-usage-detection
• f63cf6f Normalize Claude temperature for thinking
• c523101 Preserve Codex reasoning signatures for Claude
• abc293c Prevent malformed Responses SSE frames from breaking stream clients
• e166e56 Reconcile registry model states on auth changes
• 34339f6 Refactor websocket logging and error handling
• 095e143 Restore Codex prompt cache key synthesis
• e08f68e chore(auth): drop reconcile test file from pr
• 7b89583 chore(models): remove GPT-5.5 model entry from registry JSON
• a4c1e32 chore(models): remove outdated GPT-5 and related model entries from registry JSON
• 7afef97 chore: add fork upstream sync workflow and DockerHub settings
• f5aa68e chore: add workflow to prevent AGENTS.md modifications in pull requests
• 70efd4e chore: add workflow to retarget main PRs to dev automatically
• 8fac296 chore: remove Qwen support from SDK and internal components
• f5dc648 chore: remove iFlow-related modules and dependencies
• f3ab8f4 chore: update antigravity UA version to 1.21.9
• 38f0ae5 docs(antigravity): document signature validation spec alignment
• 6f58518 docs(readme): remove redundant GITSTORE_GIT_BRANCH description in README files
• 04ba8c8 feat(amp): sanitize signatures and handle stream suppression for Amp compatibility
• 88dd9c7 feat(antigravity): add AI credits quota fallback
• 14d46a0 feat(antigravity): conductor-level credits fallback for Claude models
• cf24958 feat(antigravity): configurable signature cache with bypass-mode validation
• d801393 feat(antigravity): prefer prod URL as first priority
• 1716a84 feat(api): add support for HEAD requests to /healthz endpoint
• 0a7c6b0 feat(api): enhance model assignment logic in image handlers
• ada8e29 feat(api): enhance proxy resolution for API key-based auth
• 28d7827 feat(api): implement protocol multiplexer and Redis queue for usage integration
• 894baad feat(api): integrate auth index into key retrieval endpoints for Gemini, Claude, Codex, OpenAI, and Vertex
• 5bfaf80 feat(auth): add configurable worker pool size for auto-refresh loop
• d949921 feat(auth): add proxy URL override support to auth constructors and executors
• e6866ff feat(auth): add refresh backoff for ineffective token updates
• 09e4800 feat(auth): add support for managing custom headers in auth files
• a7e92e2 feat(auth): disallow free-tier Codex auth during selection process
• a583463 feat(auth): implement auto-refresh loop for managing auth token schedule
• 31934ae feat(codex): enable image generation for all Codex upstream requests
• 01e16a8 feat(codex): handle thinking-signature conversion for reasoning content
• f1ba615 feat(codex): pass base model to enable conditional image_generation tool injection
• 3857305 feat(config): add support for disabling OpenAI compatibility providers
• 4ba1053 feat(docs): add Poixe AI sponsorship details to README files
• 8f4a4ea feat(docs): add VisionCoder sponsorship details and optimize external links
• c7b28ba feat(executor): add support for Codex image generation tool usage tracking
• 22a1a24 feat(executor): add tests for preserving key order in cache control functions
• c2d4137 feat(executor): enhance Qwen system message handling with strict injection and merging rules
• d2419ed feat(executor): ensure default system message in QwenExecutor payload
• 058793c feat(gitstore): honor configured branch and follow live remote default
• 25137b1 feat(logging): add AI API path support for image routes
• 3774b56 feat(misc): add background updater for Antigravity version caching
• 5dcca69 feat(models): add Claude Opus 4.7 model entry to registry JSON
• ea670ef feat(models): add Codex Auto Review model entry to registry JSON
• a824e7c feat(models): add GPT-5.3, GPT-5.4, and GPT-5.4-mini with enhanced "thinking" levels
• 7d5f6d9 feat(models): add GPT-5.5 model entry to registry JSON
• 4fc2c61 feat(models): add Kimi K2.6 model entry to registry JSON
• e935196 feat(models): add hardcoded GPT-Image-2 model support in Codex
• adb580b feat(security): add configuration to toggle Gemini CLI endpoint access
• 2c626ef feat(security): implement IP ban for repeated management key and Redis AUTH failures
• 7c24d54 feat(session-affinity): add session-sticky routing for multi-account load balancing
• 86c856f feat(translator): add partial and full image generation support in Codex-GPT and Codex-Gemini flows
• 8435c3d feat(tui): show time in usage details
• fdeef48 feat(vertex): Add Prefix field to VertexCredentialStorage for per-file model namespacing
• 14cb2b9 feat(vertex): add --vertex-import-prefix flag for model namespacing
• 7ee37ee feat: add /healthz endpoint and test coverage for health check
• 3724933 feat: add opt-in experimental Claude cch signing
• 8d5e470 feat: dynamically fetch antigravity UA version from releases API
• 4d6457e feat: support extracting X-Amp-Thread-Id header as session id for session affinity
• 163d683 feat: support socks5h scheme for proxy settings
• 0e12351 fix antigravity client agent headers
• 5f5d593 fix antigravity credits stream fallback
• e78d45a fix antigravity user agent handling
• b15453c fix(amp): address PR review - stream thinking suppression, SSE detection, test init
• 279cbbb fix(amp): don't suppress thinking blocks in streaming mode
• b0653ce fix(amp): strip signature from tool_use blocks before forwarding to Claude
• 727221d fix(antigravity): allow 32MB bypass signatures
• 8fecd62 fix(antigravity): cap maxOutputTokens using registry max_completion_tokens
• 41ae2c8 fix(antigravity): discard thinking blocks with non-Claude-format signatures
• 65158cc fix(antigravity): drop redacted thinking blocks with empty text
• a0bf33e fix(antigravity): preserve fallback and honor config gate
• 8ed290c fix(antigravity): reduce bypass mode log noise
• 30e94b6 fix(antigravity): refine 429 handling and credits fallback
• 4de5c29 fix(antigravity): remove credits fallback from CountTokens, fix gofmt
• 25feceb fix(antigravity): reorder model parts to prevent tool_use↔tool_result pairing breakage
• e75daa2 fix(antigravity): respect pinned auth in credits fallback, release deferred body on success
• f5ed5c7 fix(antigravity): skip full schema cleanup for empty tool requests
• d4a6a5a fix(antigravity): strip billing header from system instruction before upstream call
• 278a898 fix(antigravity): strip thinking blocks with empty signatures instead of rejecting
• 10b55b5 fix(antigravity): use E-prefixed fake signature in strict bypass test
• a992dee fix(antigravity): use real antigravity UA when polling credits balance
• ad8e396 fix(auth): add retry logic for 429 status with Retry-After and improve testing
• f130846 fix(auth): break credits cold-start deadlock by keeping unknown-hint auths as fallback candidates
• 941334d fix(auth): handle OAuth model alias in retry logic and refine Qwen quota handling
• 0ea7680 fix(auth): honor disable-cooling and enrich no-auth errors
• 730809d fix(auth): preserve and restore ready view cursors during index rebuilds
• 07b7c1a fix(auth): resolve oauth aliases before suspension checks
• 17363ed fix(auth): skip downtime for request-scoped 404 errors in model state management
• f09ed25 fix(auth): tighten registry model reconciliation
• 5e81b65 fix(auth, executor): normalize Qwen base URL, adjust RefreshLead duration, and add tests
• d0e4b1d fix(ci): keep fork releases on patch tags
• 07d6689 fix(claude): add interleaved-thinking beta header, AMP gzip error decoding, normalizeClaudeBudget max_tokens
• 0f45d89 fix(claude): address PR review feedback for OAuth cloaking
• 5f58248 fix(claude): clamp max_tokens to model limit in normalizeClaudeBudget
• f780c28 fix(claude): map question/skill to TitleCase instead of removing them
• ac36119 fix(claude): preserve OAuth tool renames when filtering tools
• 7e8e222 fix(claude): reduce forwarded OAuth prompt to minimal tool reminder
• e8d1b79 fix(claude): remap OAuth tool names to Claude Code style to avoid third-party fingerprint detection
• f0c20e8 fix(claude): remove invalid cache_control scope from static system block
• 7cdf8e9 fix(claude): sanitize forwarded third-party prompts for OAuth cloaking
• d11936f fix(codex): add default instructions for /responses/compact
• bb8408c fix(codex): backfill streaming response output
• 105a215 fix(codex): centralize session management with global store and add tests for executor session lifecycle
• 4056c25 fix(codex): classify known upstream failures
• 330e12d fix(codex): conditionally set Session_id header for Mac OS user agents and clean up redundant logic
• d85e13b fix(codex): include content field in reasoning item initialization
• 39b9a38 fix(codex): normalize null instructions across responses paths
• d3b94c9 fix(codex): normalize null instructions for compact requests
• bd855ab fix(codex): normalize null instructions for responses requests
• 1734aa1 fix(codex): prioritize websocket-enabled credentials across priority tiers in scheduler logic
• ec77f4a fix(codex): set finish_reason to tool_calls in non-streaming response when tool calls are present
• 51fd58d fix(codex): use normalizeCodexInstructions to set default instructions
• 1267fdd fix(docker-build): improve argument handling and error messaging for usage option
• 91e7591 fix(executor): add transient 429 resource exhausted handling with retry logic
• f4eb161 fix(executor): drop obsolete context-1m-2025-08-07 beta header (fixes router-for-me#2866)
• c8b7e2b fix(executor): ensure empty stream completions use output_item.done as fallback
• 69b950d fix(executor): fix OAuth extra usage detection by Anthropic API
• 0ab1f54 fix(executor): handle 429 Retry-After header and default retry logic for quota exhaustion
• 5ab9afa fix(executor): handle OAuth tool name remapping with rename detection and add tests
• 65ce863 fix(executor): implement immediate retry with token refresh on 429 for Qwen and add associated tests
• 613fe67 fix(executor): inject full Claude Code system prompt blocks with proper cache scopes
• 754b126 fix(executor): remove commented-out code in QwenExecutor
• d54f816 fix(executor): update Qwen user agent and enhance header configuration
• e34b2b4 fix(gemini): clean tool schemas and eager_input_streaming
• 1dba2d0 fix(handlers): add base URL validation and improve API key deletion tests
• 7b03f04 fix(handlers): include execution session metadata and skip idempotency key when absent
• fd71960 fix(handlers): remove handling of unsupported n parameter in OpenAI image handlers
• a188159 fix(handlers): remove references to unsupported n parameter in OpenAI image handlers
• f32c8c9 fix(handlers): update listener to bind on all interfaces instead of localhost
• c26936e fix(management): stabilize auth-index mapping
• acf98ed fix(openai): add session reference counter and cache lifecycle management for websocket tools
• d1c07a0 fix(openai): add websocket tool call repair with caching and tests to improve transcript consistency
• a3e21df fix(openai): avoid developer transcript resets
• c1d7599 fix(openai): handle transcript replacement after websocket compaction
• caa529c fix(openai): improve client IP retrieval in websocket handler
• 1b44364 fix(qwen): update CLI simulation to v0.13.2
• 06405f2 fix(security): enforce stricter localhost validation for GeminiCLIAPIHandler
• 32ef158 fix(test): remove free tier from GPT-5.5 inclusion test
• a64141a fix(tests): remove obsolete config_auth_index_test file
• da43f63 fix(tests): update Gemini family test case numbers for consistency
• d9a3b3e fix(tests): update model lookup references and enhance Claude executor tests
• d390b95 fix(tests): update test cases
• fcc59d6 fix(translator): add unit tests to validate output_item.done fallback logic for Gemini and Claude
• 04a336f fix(usage_helpers): skip zero-token usage in additional model records
• eba561b fix(util): also keep Host in header map for synthetic requests
• 8f9e662 fix(util): forward custom Host header to upstream
• 36efcc6 fix(vertex): include prefix in auth filename and validate at import
• 249f969 fix: Claude API 请求使用 utls Chrome TLS 指纹
• c03883c fix: address responses SSE review feedback
• 9e0ab4d fix: build cache_control JSON manually to avoid sjson map marshaling
• 8783caf fix: buildTextBlock cache_control sjson path issue
• 66eb122 fix: clear stale thinking signature when no block is open
• 76b53d6 fix: finalize pending thinking block before next summary part
• 8e49c79 fix: forward HTTP headers to executor Options so session affinity can read X-Amp-Thread-Id
• 73b22ec fix: omit empty signature field from thinking blocks
• a34dfed fix: preserve Claude thinking signatures in Codex translator
• 134a9ea fix: preserve SSE event boundaries for Responses streams
• 15c2f27 fix: preserve cloak config defaults when mode omitted
• 8b9dbe1 fix: record zero usage
• e2e3c7d fix: remove invalid org scope and match Claude Code block layout
• d1fd2c4 fix: repair websocket custom tool calls
• b6c6379 fix: repair websocket custom tool calls
• 8f0e66b fix: repair websocket custom tool calls
• 5fc2bd3 fix: retain codex thinking signature until item done
• c31ae2f fix: retain previously captured thinking signature on new summary part
• 25d1c18 fix: scope experimental cch signing to billing header
• f73d55d fix: simplify responses SSE suffix handling
• 0fcc02f fix: tighten responses SSE review follow-up
• f6f4640 fix: use sjson to build system blocks, avoid raw newlines in JSON
• bb44671 fix: 修复反代检测对抗的 3 个问题
• 4045378 fix: 增强 Claude 反代检测对抗能力
• 7ad1900 perf(antigravity): async credits hint refresh for warm tokens
• b6781d6 perf(codex): avoid repeated output patch writes
• f611dd6 refactor(auth): dedupe route-aware model support checks
• 6c0a1ef refactor(auth): simplify auth directory scanning and improve JSON processing logic
• 828df80 refactor(executor): remove immediate retry with token refresh on 429 for Qwen and update tests accordingly
• 4f8acec refactor(logging): centralize websocket handshake recording
• 920b6ef refactor(logging): strip unrelated deferred body changes, keep credits-only logging
• 51a4379 refactor(openai): remove websocket body log truncation limit
• d2c7e4e refactor(runtime): move executor utilities to helps package and update references
• ca11b23 refactor(runtime, openai): simplify header management and remove redundant websocket logging logic
• f5e9f01 test(amp): update tests to expect thinking blocks to pass through during streaming
• ff7dbb5 test(amp): update tests to expect thinking blocks to pass through during streaming
• 9fb6a49 test(api): add validation for unsupported models in OpenAI image handlers