@pi-stef/qwen-proxy

Guest-mode proxy for chat.qwen.ai with OpenAI + Anthropic compatibility. No Qwen account required — the proxy handles Baxia anti-bot tokens via headless Chromium (Chrome CDP).

Quick start

Docker (recommended)

cd packages/qwen-proxy/docker
SF_QWEN_API_KEY=your-secret-key docker compose up -d

Pulls ghcr.io/sfiorini/pi-stef/qwen-proxy:latest and starts the proxy on port 7790. See the Docker guide for details, volumes, image tags, and reverse-proxy setup.

Native

pnpm --filter @pi-stef/qwen-proxy dev

Verify

curl http://127.0.0.1:7790/v1/health
# {"status":"ok"}

Features

• Guest mode — no Qwen account or login required; the proxy talks directly to chat.qwen.ai
• Baxia anti-bot tokens — headless Chromium (Chrome CDP) fetches __baxia__ tokens; 25-minute cache with background refresh and pre-warm at startup
• OpenAI-compatible API — /v1/chat/completions, /v1/models
• Anthropic-compatible API — /v1/messages with claude-* model fallback to qwen3-max
• Tool calling — OpenAI-style tools/tool_choice (upstream prompt-engineering translation)
• Model aliases — map friendly names to upstream Qwen models (e.g. gpt-4o → qwen3-max)
• Global throttle — SF_QWEN_MIN_REQUEST_GAP_MS paces all requests (±50% jitter) to look human
• Inline retry-on-empty — retries empty completions up to SF_QWEN_EMPTY_RETRY_MAX (default 3, SF_QWEN_EMPTY_RETRY_GAP_MS gap) before a flat 10s pool cooldown
• Admin dashboard — /admin with Baxia cache-status panel (optional, gated by SF_QWEN_ADMIN_KEY)
• Docker — multi-arch image (linux/amd64, linux/arm64) on GHCR, non-root uid 1000, bundled Chromium + fonts

Configuration

All configuration is via environment variables (prefix SF_QWEN_).

| Variable | Default | Description | |----------|---------|-------------| | SF_QWEN_HOST | 127.0.0.1 (0.0.0.0 in Docker) | Server bind host | | SF_QWEN_PORT | 7790 | Server port | | SF_QWEN_DB | ./data/qwen-proxy.db (/data/qwen-proxy.db in Docker) | SQLite database path | | SF_QWEN_API_KEY | (required) | Client API keys, comma-separated | | SF_QWEN_ADMIN_KEY | (unset) | Admin dashboard key; unset → /admin returns 404 (D15) | | SF_QWEN_EMPTY_COOLDOWN_MS | 10000 (10s) | Flat pool cooldown applied only AFTER inline empty-retries are exhausted | | SF_QWEN_EMPTY_RETRY_MAX | 3 | Inline retries on an empty completion (Baxia CAPTCHA flag) before giving up. 0 disables (immediate cooldown) | | SF_QWEN_EMPTY_RETRY_GAP_MS | 1000 (1s) | Sleep between inline empty-retry attempts | | SF_QWEN_MIN_REQUEST_GAP_MS | 4000 (4s) | Global look-human throttle (±50% jitter); 0 disables | | SF_QWEN_MAX_CONCURRENCY | 1 | Max in-flight chat.qwen.ai calls (1 = serialize, like the web chat). Baxia flags the IP on concurrent upstream connections; raise only if you accept that risk | | SF_QWEN_PROXY_COUNT | 0 | Proxy pool size for NordVPN SOCKS5 rotation. 0 = legacy (single-IP, no rotation); >1 = enable rotation across N proxies | | SF_QWEN_PROXY_URLS | (unset) | Comma-separated SOCKS5 proxy URLs (overrides auto-discovery; e.g. socks5://user:pass@host:1080,...) | | SF_QWEN_PROXY_USER | (unset) | NordVPN service username (used for auto-discovery; ignored if SF_QWEN_PROXY_URLS is set) | | SF_QWEN_PROXY_PASS | (unset) | NordVPN service password | | SF_QWEN_PROXY_COUNTRIES | (unset) | Comma-separated country codes for auto-discovery (e.g. us,de,gb) | | SF_QWEN_TIMEOUT_MS | 60000 | TTFB timeout in ms — aborts if no response headers arrive within this window (cleared on headers, never aborts mid-stream) | | SF_QWEN_FIRST_PAYLOAD_TIMEOUT_MS | 30000 | Abort an upstream completion that produces no payload chunk within this many ms after headers (0 disables). The abort throws EmptyCompletionError → token eviction + proxy rotation, and always releases the concurrency slot | | SF_QWEN_STREAM_IDLE_TIMEOUT_MS | 30000 | Abort an upstream stream that goes silent for this many ms after content started flowing (0 disables). Ends the response gracefully with the partial content — no token eviction | | SF_QWEN_MODEL_ALIASES | (unset) | JSON object mapping alias → upstream model | | SF_QWEN_LOG_LEVEL | info | Log level | | SF_QWEN_CHROME_PATH | (unset) | Path to Chrome/Chromium; unset → autodetect (/usr/bin/chromium in Docker) | | SF_QWEN_BAXIA_CACHE_TTL_MS | 1500000 (25min) | Baxia token cache TTL | | SF_QWEN_BAXIA_READINESS_TIMEOUT_MS | 30000 (30s) | Max wait for Baxia SDK readiness per token mint (chrome-error:// pages abort instantly); clamped ≥5000 | | SF_QWEN_BAXIA_VERSION | 2.5.37 | Baxia bx-v version | | SF_QWEN_BAXIA_PRE_WARM | true | Eagerly fetch the first token at startup (exit 1 on failure) | | SF_QWEN_BAXIA_FALLBACK | false | Return last-known token on fetch failure |

See the full documentation for architecture, API surface, and known limitations.

Proxy rotation (NordVPN SOCKS5 pool)

By default the proxy uses a single IP for all upstream requests. When Qwen's Baxia anti-bot flags that IP, every subsequent request hits the same ceiling — retries rotate the same token but the IP stays the same.

Rotation mode (SF_QWEN_PROXY_COUNT > 1 or SF_QWEN_PROXY_URLS set) distributes completion requests across N SOCKS5 proxies. Each proxy is tried once per request (budget = N); if all N are exhausted (empty / network / 5xx), the proxy returns 429 with a cooldown.

How it works

| Mode | Trigger | Behavior | |------|---------|----------| | Legacy (N ≤ 1) | — | Single IP; SF_QWEN_EMPTY_RETRY_MAX inline retries + SF_QWEN_EMPTY_COOLDOWN_MS cooldown | | Rotation (N > 1) | emptyRetryMax ignored | Rotate on empty/network/5xx; budget = N attempts; all-burned → 429 + cooldown |

• Rotate triggers: EmptyCompletionError, NetworkError (TTFB timeout, connection reset), ServerError (5xx), TypeError (fetch internals), raw Error (e.g. SOCKS connect failure).
• Terminal (no rotate): ClientError (4xx, incl. data_inspection_failed), RateLimitError (429), UnknownError.
• Token generation is proxy-affine — in rotation mode, a local loopback SOCKS5 bridge injects NordVPN credentials for Chromium; Baxia tokens are cached per-proxy so the token's issuing IP matches the completion's egress IP. Legacy (N≤1) still generates tokens directly.
• Stream rotation is pre-first-content only — once the first content token has been yielded to the client, no rotation occurs (would duplicate already-sent chunks). A post-content error surfaces directly.
• No refreshBaxiaToken on rotation exhaustion — the token is fine; the IP ceiling is the bottleneck.

Setup

1. NordVPN service credentials — obtain from my.nordaccount.com → Service credentials (not your NordVPN account password).
2. Auto-discovery (recommended): set SF_QWEN_PROXY_USER + SF_QWEN_PROXY_PASS; the proxy queries api.nordvpn.com for SOCKS5 servers, sorted by load, filtered by SF_QWEN_PROXY_COUNTRIES if set. Takes the N lowest-load servers.
3. Explicit URLs: set SF_QWEN_PROXY_URLS (comma-separated, e.g. socks5://user:pass@host1:1080,socks5://user:pass@host2:1080). Overrides auto-discovery.
4. Docker: the image now depends on socks-proxy-agent; ensure outbound port 1080 is open.

Graceful degradation

• If discovery returns fewer servers than SF_QWEN_PROXY_COUNT, the proxy uses what it got (with a warning).
• If discovery returns 0 usable servers (or credentials are missing), the proxy falls back to legacy mode silently.
• If SF_QWEN_PROXY_COUNT ≤ 1 and no explicit URLs, legacy mode applies (byte-for-byte backward-compatible).

Per-proxy token affinity (resolved)

Token generation now egresses through the same SOCKS5 proxy as completion requests. A pure-Node loopback SOCKS5 bridge (net.Server + socks) binds 127.0.0.1, injects NordVPN credentials, and routes Chromium through it. Tokens are cached per-proxy key. Legacy mode (no proxy pool) remains byte-for-byte unchanged.

Documentation

• Service docs — configuration, API surface, architecture, known limitations
• Docker guide — deployment, volumes, reverse-proxy setup

License

MIT