README
¶
Nox API
Unified AI gateway. One endpoint, every model.
Route requests to 40+ AI providers through a single OpenAI-compatible API. Built-in billing, rate limiting, auth, and a modern admin dashboard.
简体中文 · 繁體中文 · English · Français · 日本語
Why Nox API
- One API, every provider — OpenAI, Claude, Gemini, AWS Bedrock, Azure, DeepSeek, Mistral, Cohere, Ollama, and 30+ more behind a single
/v1/chat/completionsendpoint. - Drop-in compatible — Anything that speaks OpenAI can talk to Nox API. No SDK changes needed.
- Smart routing — Weighted load balancing, automatic failover, per-user model rate limiting.
- Pay-as-you-go billing — Token-level metering with cache-aware billing for all major providers.
- Multi-format relay — Native support for OpenAI Chat/Responses, Claude Messages, Gemini, and automatic format conversion between them.
- Enterprise-ready auth — JWT, OAuth (GitHub, Discord, OIDC), Telegram, WebAuthn/Passkeys, 2FA.
- Modern dashboard — React + Radix UI + Tailwind. Data visualization, channel management, user admin.
Quick Start
docker run -d --name nox-api --restart always \
-p 3000:3000 -v ./data:/data \
drleomeng/nox-api:latest
Open http://localhost:3000 — default admin credentials: root / 123456
Docker Compose
git clone https://github.com/LeoMengTCM/nox-api.git
cd nox-api
# edit docker-compose.yml to your liking
docker compose up -d
Supported Providers
| Provider | Type | Provider | Type |
|---|---|---|---|
| OpenAI | Chat, Responses, Realtime, Images, Audio, Embeddings | Anthropic Claude | Chat (Messages), native format |
| Google Gemini | Chat, native format | AWS Bedrock | Claude, etc. |
| Azure OpenAI | Full OpenAI compat | Google Vertex AI | Gemini, Claude |
| DeepSeek | Chat | Mistral | Chat |
| Cohere | Chat, Rerank | xAI (Grok) | Chat |
| Ollama | Local models | Cloudflare Workers AI | Chat |
| Baidu (Wenxin) | Chat | Alibaba (Qwen) | Chat, Images |
| Zhipu (GLM) | Chat | Tencent Hunyuan | Chat |
| iFlytek (Spark) | Chat | Moonshot (Kimi) | Chat |
| MiniMax | Chat, TTS | Volcengine (Doubao) | Chat, TTS |
| SiliconFlow | Chat | Perplexity | Chat |
| Jina | Rerank, Embeddings | Dify | ChatFlow |
| OpenRouter | Multi-model | Replicate | Inference |
| Coze | Bot API | Midjourney Proxy | Image generation |
| Suno API | Music | Jimeng | Image generation |
Key Features
Routing & Reliability
- Channel weighted random & priority routing
- Automatic retry on failure with configurable retry count
- Channel affinity — sticky routing per user/token
- Per-user and per-model rate limiting
Billing & Quotas
- Token-level pay-per-use billing
- Cache-aware billing (prompt cache, context caching)
- Online top-up (Stripe, EPay)
- Subscription plans with quota management
- Redemption codes
Format Conversion
- OpenAI Chat ↔ Claude Messages — bidirectional
- OpenAI Chat → Gemini — automatic
- Gemini → OpenAI Chat — text support
- OpenAI Chat ↔ Responses — in progress
- Reasoning effort suffixes (
-high,-medium,-low) for o3, GPT-5, Gemini, Claude thinking models
Channel Management
- Per-channel system prompt injection (prepend or override mode)
- Model selector with provider-based categorization (OpenAI, Anthropic, Google, etc.)
- Upstream model fetching with search, filter, and individual selection
- 57 channel types supported
Auth & Security
- JWT tokens with scoped permissions
- OAuth: GitHub, Discord, LinuxDO, Telegram, OIDC
- WebAuthn / Passkeys
- Two-factor authentication (TOTP)
- User avatar system with upload and display
- IP-based and model-based rate limiting
Admin Dashboard
- Real-time usage statistics and charts
- Channel health monitoring
- User and token management
- Model pricing configuration
- Multi-language UI (zh, en, fr, ja, vi, ru)
Deployment
Requirements
| Component | Requirement |
|---|---|
| Database | SQLite (default), MySQL >= 5.7.8, or PostgreSQL >= 9.6 |
| Runtime | Docker recommended, or build from source (Go 1.22+) |
| Cache | Redis (recommended for production), or in-memory |
Environment Variables
| Variable | Description | Default |
|---|---|---|
SQL_DSN |
Database connection string (empty = SQLite) | — |
REDIS_CONN_STRING |
Redis connection string | — |
SESSION_SECRET |
Session secret (required for multi-node) | — |
CRYPTO_SECRET |
Encryption key (required with Redis) | — |
STREAMING_TIMEOUT |
Stream timeout in seconds | 300 |
Full list: Environment Variables
Documentation
| Installation | docs.noxapi.dev/en/docs/installation |
| API Reference | docs.noxapi.dev/en/docs/api |
| Environment Variables | docs.noxapi.dev/en/docs/installation/config-maintenance/environment-variables |
| FAQ | docs.noxapi.dev/en/docs/support/faq |
Acknowledgements
Nox API is built upon the foundation of New API by Calcium-Ion. We are grateful for their pioneering work on the AI API gateway ecosystem.
Contributing
Bug reports, feature requests, and pull requests are welcome at GitHub Issues.
License
GNU Affero General Public License v3.0 (AGPLv3)
For commercial licensing inquiries: leomengtcm@gmail.com
If Nox API is useful to you, consider giving it a star.
Documentation
¶
There is no documentation for this package.
Source Files
Directories
Click to show internal directories.
Click to hide internal directories.