Skip to content

proxy.md

Latest commit

 

History

History
99 lines (80 loc) · 3.53 KB

proxy.md

File metadata and controls

99 lines (80 loc) · 3.53 KB
Raw

API Proxy Server Documentation

Overview

This is a Cloudflare Worker script functioning as a proxy server for various AI models and APIs. It acts as a gateway between clients and different AI providers, offering features like caching, rate limiting, and security.

Configuration

Constants

  • API_HOST: Main API host
  • POLLINATIONS_HOST: Pollinations.ai service host
  • GITHUB_HOST: GitHub Pages host for static contents
  • MODEL_ID: Default model ID for Cloudflare AI
  • MAX_CONTENT_LENGTH: Max request size (50KB)
  • CUSTOM_MAX_LENGTH: Extended max size for certain endpoints (100KB)
  • CACHE_CONTROL: Caching policy (4 hours)
  • CACHE_FOREVER: Permanent caching (1 year)
  • CORS_HEADERS: Cross-Origin Resource Sharing headers configuration

Main Functions

1. API Endpoint Handling

The server handles various routes:

  • /api/[provider]/ - API calls for specific providers
  • /ai/[query] - Simplified AI query interface
  • /media/, /files/, /search/ - Media and file handling
  • /dist/ - Static content
  • /backend-api/ - Backend API routes

2. Supported AI Providers

const models = {
  "nvidia", "openrouter", "deepinfra", "groq", "ollama",
  "azure", "grok", "x.ai", "gemini", "typegpt", "pollinations",
  "api.airforce", "gpt4free.pro", "nectar", "audio", "perplexity",
  "huggingface", "puter"
}

3. Security Features

  • Rate Limiting: Limits requests (1 per 10 seconds for API paths)
  • Content-Length Validation: Checks request size limits
  • CORS Handling: Manages cross-origin requests
  • API Key Rotation: Randomly selects from multiple API keys
  • Content-Type Validation: Protects against unwanted content types

Key Functions

handleRequest(request, env, ctx)

Main function that processes all incoming requests:

  1. OPTIONS requests: Handles CORS preflight
  2. Model list endpoints: Returns available models
  3. Caching: Caches GET/HEAD responses
  4. Rate limiting: Protects against abuse
  5. Content validation: Enforces request size limit
  6. Routing: Forwards requests to appropriate providers

forwardApi(request, newUrl, liteRequest, ctx, cache_control)

Forwards API requests with caching and CORS support.

forwardWorker(env, request, liteRequest, pathname, ctx)

Special handling for Cloudflare AI worker endpoints.

shield(url, options)

Performs safety checks:

  • Content-Type validation
  • Error handling standardization
  • Removing sensitive headers
  • Adding CORS headers

retrieveCache(request, liteRequest, pathname, ctx, host, cache_control)

Caching mechanism for repeated requests.

Caching Strategy

  • Static contents: Forever cache (1 year)
  • API responses: Time-limited cache (4 hours)
  • Media/Files: Cache-based distribution

Error Handling

  • Structured JSON error responses
  • Rate limit errors with Retry-After header
  • Content-Length errors
  • Provider-specific error handling

Special Features

  • Auto-Model Selection: Random provider selection
  • JSON response formatting: Automatic JSON formatting
  • Audio support: Text-to-Speech integration
  • Image Generation: Pollinations.ai integration
  • Documentation Generation: Dynamic docs creation

Deployment

This code is optimized for Cloudflare Workers with:

  • Environment variables for API keys
  • Cloudflare KV for rate limiting
  • Cloudflare AI for local models
  • Cloudflare cache for performance

This proxy server provides a robust, scalable API gateway that enhances security, performance, and flexibility for integrating various AI services.