museai

本页保存的是公开项目资料快照，阅读过程不需要连接 GitHub。

Muse AI

OpenAI-compatible API gateway for muse.ai personal workspaces, featuring native 9Router integration.

---

Architecture

OpenAI SDK / 9Router / Web Client
              │
              ▼ HTTP (Bearer Token)
       ┌──────────────┐
       │   Muse AI    │ FastAPI Proxy (Port 18610)
       └──────┬───────┘
              │ Chrome DevTools Protocol (CDP over WebSocket)
              ▼
       ┌──────────────┐
       │   Chromium   │ Headless Browser (Isolated Context per Account)
       └──────┬───────┘
              │ HTTPS / WSS (hatch_sess + hatch_gw + datr)
              ▼
          muse.ai

---

Technical Highlights

• OpenAI Wire Compatibility: Direct drop-in for OpenAI SDKs, LangChain, LobeChat, NextChat, Cherry Studio, and autonomous coding agents.
• Native 9Router Provider: Ships with scripts/connect_9router.py to auto-register model routes directly into 9Router’s SQLite database (~/.9router/db/data.sqlite).
• Resilient Cookie Injection: Automatically URL-decodes percent-encoded cookie tokens (%3A -> :) and registers sessions across dual-domain scopes (.muse.ai and muse.ai) via CDP.
• Meta Edge Proxy Compliance: Supports datr cookie passing to prevent device-integrity redirects on Meta infrastructure.
• Account Pooling & Failover: Multi-account scheduling (affinity, lru, round_robin), concurrency limits, automatic error cooldown, and transparent retries.
• Deterministic Offline Testing: 100% offline test suite powered by MockDriver (23 passed in < 0.5s).

---

Quickstart

1. Installation

Requires Python 3.10+ and a local Chromium or Google Chrome binary.

git clone https://github.com/d4ncboz/museai.git
cd museai

python3 -m venv .venv
source .venv/bin/activate
pip install -e '.[dev]'

2. Configuration

Copy the example environment configuration:

cp .env.example .env

Default settings in .env:

MUSEAI_DRIVER=browser
MUSEAI_HOST=127.0.0.1
MUSEAI_PORT=18610
MUSEAI_API_KEY=sk-museai-local-key
MUSEAI_ADMIN_KEY=sk-museai-admin-key
MUSEAI_POOL_STRATEGY=affinity
MUSEAI_KEEPALIVE_ENABLED=true

(Note: Chrome executable is auto-detected on macOS /Applications/Google Chrome.app and Linux /usr/bin/chromium. Set MUSEAI_CHROMIUM_PATH if using a custom path).

3. Run the Service

python -m museai

The server binds to http://127.0.0.1:18610.

---

Authentication & Account Setup

1. Account Registration

An active muse.ai account is required. If you do not have an account yet, follow the step-by-step registration guide here:

• Registration Guide: Telegram @bozdrop/8447

2. Export Session Cookies

Export your session cookies from an active muse.ai browser session (DevTools F12 ──> Application ──> Cookies ──> https://muse.ai):

• hatch_sess: Session authentication token
• hatch_gw: Gateway routing cookie
• hatch_native_auth_device: Registered device UUID
• hatch_vml: Workspace lease token (optional/dynamic)
• datr: Meta device verification cookie (recommended)

Import via Admin API

curl -X POST http://127.0.0.1:18610/admin/accounts \
  -H "Authorization: Bearer sk-museai-admin-key" \
  -H "Content-Type: application/json" \
  -d '{
    "label": "primary-account",
    "cookies": {
      "hatch_sess": "...",
      "hatch_gw": "...",
      "hatch_native_auth_device": "...",
      "hatch_vml": "...",
      "datr": "..."
    }
  }'

Alternatively, use the helper script to convert raw Netscape / DevTools JSON exports:

python scripts/extract_cookies.py exported_cookies.txt --label primary-account --out account.json
curl -X POST http://127.0.0.1:18610/admin/accounts \
  -H "Authorization: Bearer sk-museai-admin-key" \
  -H "Content-Type: application/json" \
  -d @account.json

---

9Router Multi-Model Gateway Hook

To register museai into a local 9Router instance:

python scripts/connect_9router.py --port 18610 --api-key sk-museai-local-key --prefix muse

Call the model through 9Router immediately:

curl -s -X POST http://127.0.0.1:20128/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "muse/muse-chat",
    "messages": [{"role": "user", "content": "ping"}]
  }'

Available model IDs routed by 9Router:

• muse/muse-chat: Primary personal agent conversational model
• muse/gpt-4o: OpenAI tooling alias
• muse/gpt-5: High-reasoning alias
• muse/claude-sonnet-4: Sonnet alias
• muse/muse-video: Text / first-frame image-to-video

---

API Usage

Streaming Chat Completion (curl)

curl -N -X POST http://127.0.0.1:18610/v1/chat/completions \
  -H "Authorization: Bearer sk-museai-local-key" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "muse-chat",
    "messages": [
      {"role": "system", "content": "You are a concise engineering assistant."},
      {"role": "user", "content": "Explain raft consensus in two sentences."}
    ],
    "stream": true
  }'

Python SDK (openai)

from openai import OpenAI

client = OpenAI(
    base_url="http://127.0.0.1:18610/v1",
    api_key="sk-museai-local-key"
)

response = client.chat.completions.create(
    model="muse-chat",
    messages=[{"role": "user", "content": "Write a thread-safe singleton in Python"}],
    stream=False
)

print(response.choices[0].message.content)

---

Endpoints

Method	Route	Description
GET	/healthz · /readyz	Service liveness and driver readiness checks
GET	/v1/models	OpenAI-compliant model catalog and alias mapping
POST	/v1/chat/completions	Multi-turn chat (streaming SSE & buffered JSON)
POST	/v1/images/generations	Text-to-image synthesis
POST	/v1/videos	Asynchronous video generation task dispatch
GET	/v1/videos/{id}	Task status polling
GET/POST	/admin/accounts	Account pool CRUD and session renewal
GET	/admin/status	Real-time driver stats, tabs, and pool health

---

Development & Testing

# Run unit tests (MockDriver, zero external network calls)
pytest

# Code style & linting
ruff check .
ruff format .

See AGENTS.md for machine-readable architecture contracts, protocol framing, and contribution guidelines.

---

License

MIT © 2026 D4NNBOZ