v6.9.18-0

Changelog

• 29dba03 Comment out system message check in Qwen executor
• 2df3544 Fix executor compat helpers
• 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
• c744179 Merge PR router-for-me#479
• b468ca7 Merge branch 'dev' of github.com:router-for-me/CLIProxyAPI into dev
• 3e78a8d Merge branch 'main' into dev
• 5ae3858 Merge branch 'main' into plus
• 088c1d0 Merge branch 'main' into plus
• 6a43d72 Merge branch 'main' into plus
• 7223fee Merge branch 'pr-488'
• a337ecf Merge branch 'router-for-me:main' into feat(vertex)/add-prefix-field
• cd64829 Merge branch 'router-for-me:main' into main
• e7a66ae Merge branch 'router-for-me:main' into main
• 140faef Merge branch 'router-for-me:main' into main
• e9dc576 Merge branch 'router-for-me:main' into main
• 75da02a Merge branch 'router-for-me:main' into main
• 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
• 1587ff5 Merge pull request router-for-me#2389 from router-for-me/claude
• c422d16 Merge pull request router-for-me#2398 from 7RPH/fix/responses-sse-framing
• 6570692 Merge pull request router-for-me#2400 from router-for-me/revert-2374-codex-cache-clean
• 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
• 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
• f2c0f3d Merge pull request router-for-me#476 from hungthai1401/fix/ghc-gpt54mini
• 9720b03 Merge pull request router-for-me#477 from ben-vargas/plus-main
• 516d22c Merge pull request router-for-me#484 from Ve-ria/main
• 98509f6 Merge pull request router-for-me#485 from kunish/fix/copilot-premium-request-inflation
• 3882494 Merge pull request router-for-me#486 from router-for-me/plus
• 4607356 Merge pull request router-for-me#491 from Ve-ria/main
• 9a9ed99 Merge pull request router-for-me#494 from router-for-me/plus
• 6bb9bf3 Merge pull request router-for-me#495 from Ve-ria/main
• f43d25d Merge pull request router-for-me#496 from kunish/fix/copilot-premium-request-inflation
• a279192 Merge pull request router-for-me#498 from router-for-me/plus
• 03a1bac Merge upstream v6.9.9 (PR router-for-me#483)
• f63cf6f Normalize Claude temperature for thinking
• 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
• 13aa5b3 Revert "fix(codex): restore prompt cache continuity for Codex requests"
• 73cda6e Update CodeBuddy DeepSeek model description
• 341b4be Update internal/runtime/executor/codebuddy_executor.go
• e08f68e chore(auth): drop reconcile test file from pr
• f5aa68e chore: add workflow to prevent AGENTS.md modifications in pull requests
• 70efd4e chore: add workflow to retarget main PRs to dev automatically
• f3ab8f4 chore: update antigravity UA version to 1.21.9
• 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
• ada8e29 feat(api): enhance proxy resolution for API key-based auth
• 09e4800 feat(auth): add support for managing custom headers in auth files
• 6d8de0a feat(auth): implement weighted provider rotation for improved scheduling fairness
• 9a5f142 feat(codebuddy): add glm-5v-turbo model and update context lengths
• 4ba1053 feat(docs): add Poixe AI sponsorship details to README files
• 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
• 3774b56 feat(misc): add background updater for Antigravity version caching
• a824e7c feat(models): add GPT-5.3, GPT-5.4, and GPT-5.4-mini with enhanced "thinking" levels
• c1a8adf feat(registry): add GitHub Copilot gemini-3.1-pro-preview model
• bd09c0b feat(registry): add gpt-5.4-mini model to GitHub Copilot registry
• adb580b feat(security): add configuration to toggle Gemini CLI endpoint access
• 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
• 163d683 feat: support socks5h scheme for proxy settings
• 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
• a0bf33e fix(antigravity): preserve fallback and honor config gate
• 25feceb fix(antigravity): reorder model parts to prevent tool_use↔tool_result pairing breakage
• 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
• 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
• 145e0e0 fix(claude): add default max_tokens for models
• 07d6689 fix(claude): add interleaved-thinking beta header, AMP gzip error decoding, normalizeClaudeBudget max_tokens
• 5f58248 fix(claude): clamp max_tokens to model limit in normalizeClaudeBudget
• f033d3a fix(claude): enhance ensureModelMaxTokens to use registered max_completion_tokens and fallback to default
• d11936f fix(codex): add default instructions for /responses/compact
• 105a215 fix(codex): centralize session management with global store and add tests for executor session lifecycle
• 330e12d fix(codex): conditionally set Session_id header for Mac OS user agents and clean up redundant logic
• 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
• b849bf7 fix(copilot): address code review — SSE reasoning, multi-choice, agent detection
• 578c312 fix(copilot): lower static Claude context limits and expose them to Claude Code
• 59af2c5 fix(copilot): reduce premium request inflation and enable thinking
• 87bf0b7 fix(copilot): use dynamic API limits to prevent prompt token overflow
• 91e7591 fix(executor): add transient 429 resource exhausted handling with retry logic
• c8b7e2b fix(executor): ensure empty stream completions use output_item.done as fallback
• 754b126 fix(executor): remove commented-out code in QwenExecutor
• bea13f9 fix(executor): support non-stream requests for CodeBuddy
• d54f816 fix(executor): update Qwen user agent and enhance header configuration
• e34b2b4 fix(gemini): clean tool schemas and eager_input_streaming
• 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
• 08e078f fix(openai): route copilot Gemini preview models to chat endpoint
• 1b44364 fix(qwen): update CLI simulation to v0.13.2
• 06405f2 fix(security): enforce stricter localhost validation for GeminiCLIAPIHandler
• d390b95 fix(tests): update test cases
• fcc59d6 fix(translator): add unit tests to validate output_item.done fallback logic for Gemini and Claude
• 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
• 66eb122 fix: clear stale thinking signature when no block is open
• 76b53d6 fix: finalize pending thinking block before next summary part
• 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
• d1fd2c4 fix: repair websocket custom tool calls
• 8f0e66b fix: repair websocket custom tool calls
• b6c6379 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
• 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
• f611dd6 refactor(auth): dedupe route-aware model support checks
• 4f8acec refactor(logging): centralize websocket handshake recording
• 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
• 0805989 更新CodeBuddy CN的模型列表