AdaptiveProxy

Intelligent multi-provider LLM routing library for Go applications

A high-performance Go library that routes requests across multiple LLM providers (OpenAI, Anthropic, Gemini, DeepSeek, Groq) with intelligent fallback, caching, and cost optimization.

🚀 Quick Start

Installation

go get github.com/Egham-7/adaptive-proxy

Basic Usage

package main

import (
    "log"
    "os"

    "github.com/Egham-7/adaptive-proxy/pkg/config"
)

func main() {
    // Configure proxy with OpenAI and Anthropic fallback
    builder := config.New().
        Port("8080").
        AddOpenAICompatibleProvider("openai",
            config.NewProviderBuilder(os.Getenv("OPENAI_API_KEY")).Build(),
        ).
        AddAnthropicCompatibleProvider("anthropic",
            config.NewProviderBuilder(os.Getenv("ANTHROPIC_API_KEY")).Build(),
        )

    // Start server
    srv := config.NewProxyWithBuilder(builder)
    if err := srv.Run(); err != nil {
        log.Fatal(err)
    }
}

Your proxy is now running on http://localhost:8080 with automatic fallback!

⭐ Key Features

• Zero Vendor Lock-in - Switch providers without code changes
• Multi-Provider Support - OpenAI, Anthropic, Groq, DeepSeek, Gemini
• Production-Ready - Redis caching, rate limiting, graceful shutdown
• Type-Safe - Fluent builder API with full Go type safety

📚 Documentation

Full documentation available in docs/

• Quick Start - Get running in 5 minutes
• Installation - Installation and setup
• Basic Usage - Core concepts and patterns
• Providers - Configure OpenAI, Anthropic, Gemini, and custom providers
• Caching - Redis-backed prompt caching and semantic search
• Routing - Intelligent model selection and cost optimization
• Fallback - Multi-provider fallback strategies
• Middleware - Rate limiting, timeouts, and custom middleware

💡 API Usage

Once your proxy is running, make requests using the OpenAI-compatible API:

Chat Completions

curl http://localhost:8080/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-4",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

Intelligent Routing with Cost Optimization

{
  "model": "",
  "messages": [{ "role": "user", "content": "Complex analysis task" }],
  "model_router": {
    "cost_bias": 0.3,
    "models": [
      { "provider": "openai" },
      { "provider": "anthropic", "model_name": "claude-3-5-sonnet-20241022" }
    ]
  },
  "cache": {
    "enabled": true,
    "semantic_threshold": 0.85
  },
  "fallback": {
    "mode": "sequential"
  }
}

Other Endpoints

• GET /v1/models - List available models
• GET /health - Health check
• POST /v1/messages - Anthropic-compatible messages endpoint

🛠️ Development

Running from Source

git clone https://github.com/Egham-7/adaptive-proxy.git
cd adaptive-proxy
cp .env.example .env.local  # Add your API keys
go run cmd/api/main.go

Code Quality (Required Before Commit)

gofmt -w . && golangci-lint run && go mod tidy && go vet ./... && go test ./...

Install Development Tools

go install github.com/golangci/golangci-lint/cmd/golangci-lint@latest

🏗️ Architecture

See CLAUDE.md for detailed architecture documentation and AGENTS.md for AI agent development guidelines.

🤝 Contributing

Contributions welcome! Please ensure all pre-commit checks pass:

gofmt -w . && golangci-lint run && go mod tidy && go vet ./... && go test ./...

📝 License

See LICENSE for details.

🆘 Support

• Documentation: docs/
• Issues: GitHub Issues
• Discussions: GitHub Discussions