GoModel: A Lightweight Open-Source AI Gateway Written in Go

GoModel is an open-source AI gateway written in Go that sits between your application and model providers like OpenAI, Anthropic, Gemini, and others. It provides a unified OpenAI-compatible API interface while handling provider-specific differences internally.

Key Features and Differences

The project was built to solve several practical problems: tracking AI usage and cost per client or team, switching models without changing application code, debugging request flows more easily, and reducing AI spending with exact and semantic caching.

Key differentiators from alternatives:

• ~17MB Docker image (LiteLLM's image is ~746MB on amd64, making GoModel 44x lighter)
• Request workflow is visible and easy to inspect
• Configuration is environment-variable-first by default

Quick Start

Basic deployment with Docker:

docker run --rm -p 8080:8080 \
  -e OPENAI_API_KEY="your-openai-key" \
  enterpilot/gomodel

For production, avoid passing secrets via command line and use:

docker run --env-file .env enterpilot/gomodel

Make your first API call:

curl http://localhost:8080/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{ "model": "gpt-5-chat-latest", "messages": [{"role": "user", "content": "Hello!"}] }'

Supported Providers

GoModel supports multiple LLM providers with automatic detection based on supplied credentials:

• OpenAI (OPENAI_API_KEY)
• Anthropic (ANTHROPIC_API_KEY)
• Google Gemini (GEMINI_API_KEY)
• Groq (GROQ_API_KEY)
• OpenRouter (OPENROUTER_API_KEY)
• Z.ai (ZAI_API_KEY)
• xAI/Grok (XAI_API_KEY)
• Azure OpenAI (AZURE_API_KEY + AZURE_BASE_URL)
• Oracle (ORACLE_API_KEY + ORACLE_BASE_URL)
• Ollama (OLLAMA_BASE_URL)

The gateway supports chat completions, embeddings, file processing, batch operations, and passthrough capabilities across most providers. For Oracle, you may need to set ORACLE_MODELS=openai.gpt-oss-120b,xai.grok-3 when the upstream /models endpoint is unavailable.

Alternative Setup Methods

You can also run from source (Go 1.26.2+ required) or use Docker Compose for infrastructure components including Redis, PostgreSQL, MongoDB, and Adminer.

This type of gateway is particularly useful for teams managing multiple AI models across different providers, needing cost tracking, or wanting to maintain flexibility to switch providers without code changes. The lightweight Docker image makes it suitable for resource-constrained environments.