chore: sync 140 upstream commits while preserving Pinable runtime contracts

.github/workflows/pinable-runtime.yml

@@ -6,14 +6,19 @@ on:
   pull_request:
     branches: [main]
     paths:
+      - 'cmd/server/**'
+      - 'internal/**'
+      - 'sdk/**'
+      - 'go.mod'
+      - 'go.sum'
       - '.github/workflows/pinable-runtime.yml'
       - 'scripts/pinable/runtime.py'
       - 'scripts/pinable/test_runtime.py'
       - 'scripts/pinable/test_publish_runtime.py'
+      - 'scripts/pinable/test_smoke_runtime.py'
       - 'scripts/pinable/smoke_runtime.py'
       - 'scripts/pinable/publish_runtime.py'
       - 'docs/pinable/desktop-runtime.md'
-      - 'internal/watcher/**'
   workflow_dispatch:
     inputs:
       version:

.github/workflows/pinable-upstream-check.yml

@@ -26,8 +26,8 @@ jobs:
           persist-credentials: false
       - name: Verify synchronization baseline ancestry
         run: |
-          git merge-base --is-ancestor 2eb057869ca11301edf1768e6ff68d8f95521e10 HEAD
-          git merge-base --is-ancestor a270e7b9e57aaecd8f82555f44c2108518ad2330 HEAD
+          git merge-base --is-ancestor fa10525c30391ebade420d23302425eecc9678c1 HEAD
+          git merge-base --is-ancestor 67465884ca179a8f9098328d03a361b50003fdd7 HEAD
       - name: Test safe synchronization script
         run: |
           bash -n scripts/pinable/sync-upstream.sh
@@ -42,6 +42,10 @@ jobs:
         run: go test ./...
       - name: Repeat timeout ordering regressions
         run: go test ./internal/runtime/executor -run 'TestCodex(Websockets)?Executor_BootstrapBuffering_(TimeBudgetReleasesStream|OverloadDirectlyAfterTimeoutDeliveredInStream|StatusBearingErrorAfterTimeoutDeliveredInStream)$' -count=20
+      - name: Repeat shutdown acknowledgement regressions with race detection
+        run: go test -race ./internal/api/handlers/management -run '^TestPostRuntimeShutdown(WaitsForResponseFlush|AcknowledgesBeforeImmediateConnectionClose)$' -count=50
+      - name: Repeat HTTP interrupt cancellation regression
+        run: go test ./sdk/api/handlers/openai -run '^TestResponsesInterruptStopsHTTPUpstream$' -count=100
       - name: Probe version and discovery output
         shell: bash
         run: |

AGENTS.md

@@ -50,6 +50,7 @@ go build -o test-output ./cmd/server && rm test-output # Verify compile (REQUIRE
 - As a rule, do not make standalone changes to `internal/translator/`. You may modify it only as part of broader changes elsewhere.
 - If a task requires changing only `internal/translator/`, run `gh repo view --json viewerPermission -q .viewerPermission` to confirm you have `WRITE`, `MAINTAIN`, or `ADMIN`. If you do, you may proceed; otherwise, file a GitHub issue including the goal, rationale, and the intended implementation code, then stop further work.
 - `internal/runtime/executor/` should contain executors and their unit tests only. Place any helper/supporting files under `internal/runtime/executor/helps/`.
+- Payload configuration MUST be the final semantic barrier before sending requests in every executor, including streaming, WebSocket, continuation, retry/fallback, image, and token-count paths where applicable. Complete all built-in payload translation, normalization, injection, and cleanup first, then evaluate and apply user payload rules exactly once to the final business payload. No subsequent logic may overwrite configured values or restore filtered fields. Only necessary transport framing, serialization, signature calculation, and read-only validation may follow without changing business payload semantics. Add regression tests when introducing or changing request-building paths to enforce this ordering.
 - Follow `gofmt`; keep imports goimports-style; wrap errors with context where helpful
 - Do not use `log.Fatal`/`log.Fatalf` (terminates the process); prefer returning errors and logging via logrus
 - Shadowed variables: use method suffix (`errStart := server.Start()`)

README.md

@@ -16,23 +16,31 @@ You can access the following providers locally and with multiple CLI accounts th
     </tr>
     <tr>
         <td align="center"><a href="https://www.kimi.com/code/?aff=cliproxyapi"><img src="./assets/logo/kimi.svg" alt="Kimi" width="28" height="28" /></a></td>
-        <td>Kimi series models (Kimi K3, Kimi K2.7 Code, etc.). <a href="https://platform.kimi.ai/docs/guide/kimi-k3-quickstart">Kimi K3</a> is Moonshot AI’s most capable model and the world’s first open 3T-class model. With 2.8 trillion parameters, native vision, and a 1-million-token context window, K3 is built for long-horizon coding, knowledge work, and reasoning. CLIProxyAPI supports Kimi through OAuth or compatible API interfaces. Try a <strong>Kimi Code plan</strong> (<a href="https://www.kimi.com/code?aff=cliproxyapi">中文站</a> | <a href="https://www.kimi.ai/code?aff=cliproxyapi">Global</a>), or get an <strong>API key</strong> from the Kimi Open Platform (<a href="https://platform.kimi.com?track_id=track-f15622e7182046baa22ca35e006e13a7&aff=cliproxyapi">中文站</a> | <a href="https://platform.kimi.ai?track_id=track-8a28e4b291d84f62af2fccc3e7a21cb3&aff=cliproxyapi">Global</a>). Thanks to Kimi for supporting CLIProxyAPI and the open-source community!</td>
+        <td>Kimi series models (Kimi K3, K2.8 Preview, etc.). <a href="https://platform.kimi.ai/docs/guide/kimi-k3-quickstart">Kimi K3</a> is Moonshot AI’s most capable model and the world’s first open 3T-class model. With 2.8 trillion parameters, native vision, and a 1-million-token context window, K3 is built for long-horizon coding, knowledge work, and reasoning. CLIProxyAPI supports Kimi through OAuth or compatible API interfaces. Try a <strong>Kimi Code plan</strong> (<a href="https://www.kimi.com/code?aff=cliproxyapi">中文站</a> | <a href="https://www.kimi.ai/code?aff=cliproxyapi">Global</a>), or get an <strong>API key</strong> from the Kimi Open Platform (<a href="https://platform.kimi.com?track_id=track-f15622e7182046baa22ca35e006e13a7&aff=cliproxyapi">中文站</a> | <a href="https://platform.kimi.ai?track_id=track-8a28e4b291d84f62af2fccc3e7a21cb3&aff=cliproxyapi">Global</a>). Thanks to Kimi for supporting CLIProxyAPI and the open-source community!</td>
     </tr>
     <tr>
-        <td align="center"><a href="https://platform.openai.com/docs/guide/gpt-5.6"><img src="./assets/logo/openai.svg" alt="OpenAI" width="28" height="28" /></a></td>
-        <td>OpenAI GPT series models (GPT 5.6, GPT 5.5, etc.). GPT-5.6 sets a new quality and efficiency baseline for complex production workflows. GPT-5.6 is especially token-efficient and improves frontend aesthetics, including layout, visual hierarchy, and design judgment.</td>
+        <td align="center"><a href="https://developers.openai.com/api/docs/models"><img src="./assets/logo/openai.svg" alt="OpenAI" width="28" height="28" /></a></td>
+        <td>OpenAI GPT-6 models (GPT-6 Astra, GPT-6.1 Sol, GPT-6 Luna), including access through Codex OAuth. Astra is the flagship for complex reasoning and coding, Sol balances capability and cost, and Luna handles focused, high-volume work.</td>
     </tr>
     <tr>
         <td align="center"><a href="https://www.anthropic.com/claude/fable"><img src="./assets/logo/claude.svg" alt="Anthropic" width="28" height="28" /></a></td>
-        <td>Anthropic Claude series models (Claude Fable, Claude Opus, Claude Sonnet, etc.). Claude Fable 5 is Anthropic's most capable widely released model, built for the most demanding reasoning and long-horizon agentic work.</td>
+        <td>Anthropic Claude models (Claude Fable 5.1, Claude Opus 5.5, Claude Sonnet 5.5). Fable 5.1 is built for ambitious, long-running coding and knowledge work; Opus 5.5 brings strong agentic coding at a lower cost.</td>
     </tr>
     <tr>
         <td align="center"><a href="https://antigravity.google/"><img src="./assets/logo/antigravity.svg" alt="Antigravity" width="28" height="28" /></a></td>
-        <td>Google Gemini series models (Gemini 3.5 Flash, Gemini 3.1 Pro, etc.). Gemini 3.5 Flash provides sustained frontier-level intelligence optimized for real-world tasks at a higher speed and lower cost. Designed for the agentic era, it excels at sub-agent deployment, multi-step workflows, and long-horizon tasks at scale. This model is particularly effective for rapid agentic loops involving complex coding cycles and iterations.</td>
+        <td>Google Gemini models include Gemini 3.8 Flash and Gemini 3.1 Pro Preview. CLIProxyAPI supports Gemini API, AI Studio, Vertex AI, Gemini CLI, and Antigravity accounts; model availability varies by channel. Gemini 3.8 Flash is Google's latest Flash model for long-horizon software engineering and agentic workflows.</td>
     </tr>
     <tr>
-        <td align="center"><a href="https://x.ai/grok"><img src="./assets/logo/xai.svg" alt="xAI" width="28" height="28" /></a></td>
-        <td>xAI Grok series models (Grok 4.5, Grok Composer 2.5 Fast, etc.). Grok 4.5 is SpaceXAI's frontier model built for coding, agentic tasks, and knowledge work. It was trained in SpaceXAI's data centers in Memphis with new datasets spanning science, engineering, and math.</td>
+        <td align="center"><a href="https://docs.x.ai/developers/grok-4-7"><img src="./assets/logo/xai.svg" alt="xAI" width="28" height="28" /></a></td>
+        <td>xAI Grok models (Grok 4.7, Grok 4.7 Build Fast, etc.). Grok 4.7 is SpaceXAI's latest model for coding, agentic tasks, and knowledge work.</td>
+    </tr>
+    <tr>
+        <td align="center"><a href="https://dev.meta.ai/docs/overview"><img src="./assets/logo/meta.svg" alt="Meta" width="28" height="28" /></a></td>
+        <td>Meta Muse models (Muse Spark 1.3, Muse Spark 1.2, etc.). CLIProxyAPI supports Muse Code accounts through Meta login and Meta Model API keys, with Muse Spark 1.3 for coding and agentic workflows.</td>
+    </tr>
+    <tr>
+        <td align="center"><a href="https://devin.ai/cli">Devin</a></td>
+        <td>Devin models (SWE-2, GPT-6 Astra, Claude Fable 5.1, etc.). Connect a Devin account with <code>--devin-login</code> to route the models available to that account.</td>
     </tr>
 </tbody>
 </table>
@@ -81,10 +89,6 @@ PackyCode provides special discounts for our software users: register using <a h
 <td>Thanks to APIMart for sponsoring this project! APIMart is a low-cost API platform for AI image &amp; video generation — GPT-Image-2 from &#36;0.006/image, 160+ images per dollar. One async API covers both image and video: submit a task, get an ID, fetch results via polling or callback. Batch tens of thousands of images without timeouts, switch models without changing code. Pay-as-you-go with no monthly fee — <a href="https://go.apimart.ai/gh-cliproxyapi">sign up here</a> to get started.</td>
 </tr>
 <tr>
-<td width="180"><a href="https://www.axisnow.io/zh"><img src="./assets/axisnow.png" alt="AxisNow" width="150"></a></td>
-<td>Protect and accelerate websites and APIs while optimizing access from both mainland China and the rest of the world. Extend acceleration and security to native and mobile apps through client SDKs — <b>self-hosted private CDN | subscription-based DDoS-protected CDN | independently controlled, flexibly composable CDN networks.</b></td>
-</tr>
-<tr>
 <td width="180"><a href="https://www.swiftproxy.net/?code=PR67S9A95"><img src="./assets/swiftproxy.png" alt="Swiftproxy" width="150"></a></td>
 <td>Swiftproxy provides 90M+ clean residential IPs across 220+ locations, supporting HTTP(S)/SOCKS5, IP rotation, Sticky Sessions, and precise location targeting. It helps AI API tools and automation workflows access online services reliably from different locations, making it ideal for API requests, web access, data collection, and location-based testing. Residential proxies from &#36;0.7/GB. Free testing is available, with 10% off using code PROXY90. <a href="https://www.swiftproxy.net/?code=PR67S9A95">Try Swiftproxy Now</a></td>
 </tr>
@@ -150,6 +154,10 @@ Standalone persistence and visualization service for CLIProxyAPI, with periodic
 
 Full CLIProxyAPI management center with request-level monitoring and cost estimates. CPA-Manager tracks collected requests by account, model, channel, latency, status, and token usage; estimates cost with editable model prices and one-click LiteLLM price sync; persists events in SQLite; and provides Codex account-pool operations with batch inspection, quota detection, unhealthy account discovery, cleanup suggestions, and one-click execution for day-to-day multi-account maintenance.
 
+### [Oh-My-CPA](https://github.com/WizisCool/oh-my-cpa)
+
+Modern Ant Design-based management console for CLIProxyAPI v8+, combining CPAMC-style administration with SQLite-backed request records and usage analytics. Covers OAuth accounts, API providers, client keys, quotas, pricing and CPA operations in one interface. Tracks request-level latency, TTFT, tokens and cost, with multi-dimensional filtering, live dashboards and token heatmaps. OpenRouter price sync, custom rates and per-request price snapshots keep historical costs stable; a built-in Agent and MCP tools support assisted investigation and management.
+
 ## SDK Docs
 
 - Usage: [docs/sdk-usage.md](docs/sdk-usage.md)
@@ -291,6 +299,10 @@ Claude Code status line for CLIProxyAPI: per-account Codex / Grok / Antigravity
 
 Cross-platform Electron tray dashboard for CLIProxyAPI. Shows the real OAuth quota windows per account across ChatGPT/Codex, Claude, Gemini/Antigravity, Grok, Kimi and Cursor, estimates usage-queue cost, and tracks OpenAI/Claude service status. Dark terminal-style UI; runs on Windows and Linux.
 
+### [panel4cliproxyapi](https://github.com/yaanlaan/panel4cliproxyapi)
+
+A modern, responsive web dashboard and management console for CLIProxyAPI built with React and Tailwind CSS. Features multi-provider OAuth login with LAN redirect assistance, visual zero-YAML configuration, real-time traffic analytics charts, client API key management, core version checking & hot-reloading, and a live SSE streaming playground.
+
 > [!NOTE]  
 > If you developed a project based on CLIProxyAPI, please open a PR to add it to this list.
 
@@ -312,6 +324,10 @@ OmniRoute is an AI gateway for multi-provider LLMs: an OpenAI-compatible endpoin
 
 This is a tool built with Tauri 2 + Vue 3 for managing multiple OpenAI Codex desktop accounts. Switch between saved ChatGPT/Codex certification profiles, check 5-hour and weekly quota usage in real time, verify token health, view active account details, and import or save auth.json files without manual copying.
 
+### [cliproxy-rs](https://github.com/vayungodara/cliproxy-rs)
+
+A Rust port of CLIProxyAPI that reads the same `config.yaml` and auth files and serves the same routes and v8 Management API, so you can switch between the two in either direction. Ships as a single binary with the management dashboard built in (per-account 5-hour and weekly quota view, account sign-in, client setup guides), and adds an opt-in `soonest-reset` routing strategy that spends the account whose weekly window resets first.
+
 > [!NOTE]  
 > If you have developed a port of CLIProxyAPI or a project inspired by it, please open a PR to add it to this list.