ramzeng / ai-endpoint

An AI endpoint that provides a unified interface for AI models.

AI Endpoint

Introduction

English | 简体中文

This project is an AI endpoint that provides a unified interface for AI models. It is designed to be multi-tenant, and supports load balancing, rate limiting, and logging capabilities. It is also containerized for easy deployment.

Capabilities

• Azure API Proxy
  • Compatible with OpenAI API format
  • API Keys load balancing
  • Weighted round-robin
  • Adaptive weight
• Multi-tenant
  • Request authentication
  • Configuration isolation
• Rate limiting capabilities
  • Tenant-level rate limiting
  • Model-level rate limiting
• Logging capabilities
  • Multi-channel writing
  • File splitting
• Containerized deployment

Supported routes

• POST /v1/chat/completions

Invocation method

curl --location 'localhost:8080/v1/chat/completions' \
--header 'Content-Type: application/json' \
--header 'Authorization: Bearer xxxx-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx' \
--data '{
    "stream": true,
    "model": "gpt-3.5-turbo",
    "messages": [
        {
            "role": "system",
            "content": "You are a helpful assistant."
        },
        {
            "role": "user",
            "content": "Does Azure OpenAI support customer managed keys?"
        },
        {
            "role": "assistant",
            "content": "Yes, customer managed keys are supported by Azure OpenAI."
        },
        {
            "role": "user",
            "content": "Do other Azure Cognitive Services support this too?"
        }
    ]
}'

Dependencies

• gin
• google/uuid
• redis/go-redis
• spf13/cast
• spf13/viper
• go.uber.org/zap
• gorm.io/mysql
• gorm.io/gorm
• tidwall/gjson

Configuration file

logger:
  channels:
    - name: app
      filename: /var/log/app.log
      maxSize: 1
      maxAge: 30
      maxBackups: 10
      compress: false
      level: info
    - name: request
      filename: /var/log/request.log
      maxSize: 1
      maxAge: 30
      maxBackups: 10
      compress: false
      level: info
database:
  addr: root:password@tcp(127.0.0.1:3306)/endpoints?charset=utf8mb4&parseTime=True&loc=Local
redis:
  addr: 127.0.0.1:6379
  password: ""
  db: 0
azure:
  openai:
    models:
      - gpt-3.5-turbo
      - gpt-4
      - gpt-4-32k
    peers:
      - key: xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
        endpoint: https://api.openai.com
        weight: 20
        deployments:
          - model: gpt-3.5-turbo
            # if you want to use the original openai model, set isOpenAI to true
            isOpenAI: true
          - model: gpt-4
            isOpenAI: true
          - model: gpt-4-32k
            isOpenAI: true
      - key: xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
        endpoint: https://xxxxx.openai.azure.com
        weight: 20
        deployments:
          - name: gpt-35-turbo
            model: gpt-3.5-turbo
            version: 2023-03-15-preview
          - name: gpt-4
            model: gpt-4
            version: 2023-03-15-preview
          - name: gpt-4-32k
            model: gpt-4-32k
            version: 2023-03-15-preview