24.4K stars! Manage the world's mainstream AI models with one click. This open source API gateway is amazing!

One API It is the most powerful AI model API management platform. Developers only need to deploy one service to uniformly access 18+ mainstream AI models such as OpenAI, Azure, Claude, Gemini, etc. It solves the three major pain points in enterprise-level applications:

1. Key management chaos : centrally manage all your API keys
2. Difficulty in cost control : Intelligent allocation of request channels
3. Complex system connection : unified and standardized API interface

# Typical deployment commands (Docker version)
docker run --name one-api -d --restart always \
-p 3000:3000 -e TZ=Asia/Shanghai \
-v /home/ubuntu/data/one-api:/data \
justsong/one-api

Core Features Highlights

 Unified access to all models

Support includes:

• International giants : OpenAI GPT-4o, Google Gemini Pro
• Domestic large models : Wenxin Yiyan, iFlytek Spark
• Open source upstarts : ChatGLM, DeepSeek
• Industry-specific : ByteDance, Tencent Hunyuan

 Intelligent routing engine

// Configuration example for automatically selecting the optimal channel
{
  "strategy" :  "weighted-random" ,  // weighted random
  "retry" :  3 ,                     // Number of failed retries
  "timeout" :  30 // Timeout (seconds)                  
}

Accurate cost control

• Dynamically calculate token consumption (supports 2x completion rate of GPT-4)
• Multi-dimensional statistical reports (user/channel/model dimensions)
• Real-time usage warning (email/webhook notification)

Enterprise-grade security

# Security Configuration Example
MEMORY_CACHE_ENABLED=true # Enable cache acceleration
GLOBAL_API_RATE_LIMIT=300 # Single IP request limit per minute
RELAY_PROXY=http://proxy:8080 # Enterprise-level proxy support

? Scalable architecture

## Technical architecture analysis
| Components | Technology Selection | Advantages |
|-----------------|---------------------|----------------------------|
| Core framework | Golang | High performance concurrent processing |
| Front-end interface | React + Ant Design | Enterprise-level management experience |
| Deployment plan | Docker single file | 5 minutes quick deployment |
| Database support | SQLite/MySQL | Lightweight and high availability, free choice |
| Cache mechanism | Memory cache + batch update | Throughput increased by 300% |
| Monitoring system | Prometheus + built-in indicators | Real-time channel health detection |

## Six application scenarios
1. **SaaS product development**: Rapid integration of multi-model capabilities
2. **Enterprise internal system**: unified management of AI resources
3. Educational and scientific research institutions: flexible allocation of computing resources
4. **Developer Studio**: Reduce API call costs
5. AI Application Market: Building a Model Distribution Platform
6. **Cross-border business system**: intelligent switching of regional nodes


![](https://files.mdnice.com/user/74286/e7e59d34-2c2d-4a50-9ec6-4f73906aeb88.png)


## Comparison of similar projects
| Project Name | Core Advantages | Limitations | Applicable Scenarios |
|---------------|-------------------------|-------------------------|----------------------|
| **One API** | Supports the most models/easiest to deploy | Technical interface | Enterprise-level API gateway |
| LobeChat | Excellent interactive experience | Only supports basic models | Individual developers |
| FastGPT | Powerful knowledge base | Depends on specific cloud services | Intelligent customer service system |
| API Gateway | General API Management | No AI-specific optimization | Traditional microservice architecture |

## Actual combat case demonstration
**Scenario**: E-commerce customer service system needs to use GPT-4 and Wenxinyiyan at the same time

```Python
# Unified call example
import requests

def ask_ai(prompt):
    url =  "https://api.yourcompany.com/v1/chat/completions"
    headers = { "Authorization" :  "Bearer sk-xxxx" }
    data = {
        "model" :  "gpt-4" ,   # Automatically route to available channels
        "messages" : [{ "role" :  "user" ,  "content" : prompt}]
    }
    response = requests.post(url, json=data, headers=headers)
    return  response.json()

Project Deployment Guide

1. Basic deployment (suitable for individual developers)

# Quick Start with SQLite
./one-api --port 3000 -- log -dir ./logs

1. Enterprise-level deployment (high availability solution)

# docker-compose-pro.yml
version: '3'
services:
one-api:
    image: justsong/one-api
    environment:
      - DATABASE_URL=mysql://user:pass@db:3306/oneapi
      - MEMORY_CACHE_ENABLED=true
    ports:
      - "3000:3000"
    depends_on:
      -db​

db:
    image: mysql:8.0
    environment:
      - MYSQL_ROOT_PASSWORD=securepassword

Ecosystem expansion recommendation

• FastGPT : A knowledge base system based on a large model
• ChatGPT-Next-Web : Cross-platform client
  VChart : Intelligent visualization solution
• CherryStudio : Enterprise-level AI development platform

Interface Effects