AI Gateway (Go)

Archived. This service moved into zetaoss/bob as /aigate.

A simple Go-based AI proxy gateway. It centralizes AI model calls and routes requests to multiple providers such as Gemini and Ollama under a single /v1 API.

Supported Endpoints

• GET /healthz
• GET /v1/models (enabled models)
• GET /v1/models?all (available models from providers, with enabled flag)
• GET /v1/models/{provider} (enabled models for one provider)
• GET /v1/models/{provider}?all (available models for one provider, with enabled flag)
• POST /v1/chat/completions

Configuration Example

server:
  port: 8080
  logLevel: info
  validateModelsOnStartup: true

fallback:
  rounds: 2
  perAttemptTimeout: 30s

models:
# format: {provider}/{model}
- gemini/gemini-2.5-flash
- ollama/gemma4:31b

providers:
  gemini:
    apiKey: YOUR_GEMINI_API_KEY
  ollama:
    apiKey: YOUR_OLLAMA_API_KEY

Provider endpoints are fixed in code for gemini and ollama.

server.port defaults to 8080 when omitted. server.logLevel supports debug, info, error (default: info). Use debug to print upstream provider request/response logs. The gateway loads each provider's available model list once during startup and keeps it in memory for /v1/models?all. server.validateModelsOnStartup verifies each configured model against that startup model list (default: true). models uses {provider}/{model} and splits on the first /, so model ids containing / are supported. model omission is always allowed and routes by configured model order. When model is specified, fallback is always allowed within the resolved model chain. fallback.rounds controls how many full passes of the resolved model chain are attempted (default: 2). fallback.perAttemptTimeout is per-attempt HTTP timeout.

Usage Examples

List available models

curl http://localhost:8080/v1/models?all

Create a chat completion

curl -X POST http://localhost:8080/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{"model":"gemini/gemini-2.5-flash","messages":[{"role":"user","content":"Hello"}]}'