tenantai

README

tenantai

A single-binary, production-ready AI API proxy with per-tenant isolation. Not a framework — a deployable backend. Connect to 300+ models via OpenRouter, get conversation logging for free, extend only what you need.

Features

• Multi-tenant Isolation — Each service gets its own DuckDB database file
• OpenRouter Integration — Unified interface to 300+ AI models
• Streaming Support — Server-Sent Events (SSE) for real-time responses
• Async Conversation Logging — Non-blocking logging to preserve API latency
• Service Management API — Create, list, and delete isolated services at runtime
• Conversation History — Query history with pagination and session filtering
• Configuration — YAML-based config with environment variable expansion

Quick Start

git clone https://github.com/hra42/tenantai.git
cd tenantai
export OPENROUTER_API_KEY="your-api-key-here"
go run main.go

Verify with:

curl http://localhost:8080/health

Using the Quickstart Script

The quickstart script creates a config file, sets up a default service, and starts the backend:

export OPENROUTER_API_KEY="your-api-key-here"
./scripts/quickstart.sh

Docker Compose

Run the full stack (backend + frontend) with Docker Compose:

cd examples
cp .env.example .env
# Edit .env and set your OPENROUTER_API_KEY
docker compose up

The frontend will be available at http://localhost:3000 and the backend at http://localhost:8080.

Configuration

Config loads from config.yaml with ${ENV_VAR} expansion (12-factor style).

server:
  port: 8080
  env: development

openrouter:
  api_key: "${OPENROUTER_API_KEY}"
  base_url: "https://openrouter.ai/api/v1"

database:
  services_dir: "./data/services"
  max_connections: 10

services:
  - id: default
    name: "Default Service"

Required: OPENROUTER_API_KEY environment variable.

API Overview

See docs/API.md for full endpoint documentation with curl examples.

Endpoint	Description
GET /v1/models	List available models from OpenRouter
POST /v1/chat/completions	Chat completion (requires X-Service-ID header)
POST /services	Create a new service
GET /services	List all services
GET /services/:id	Get service details
DELETE /services/:id	Delete a service
GET /services/:id/conversations	Query conversation history

Quick Example

# Create a service
curl -X POST http://localhost:8080/services \
  -H "Content-Type: application/json" \
  -d '{"id": "my-app", "name": "My App"}'

# Send a chat completion
curl -X POST http://localhost:8080/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "X-Service-ID: my-app" \
  -d '{
    "model": "openai/gpt-4-turbo",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

Architecture

See docs/ARCHITECTURE.md for detailed design documentation.

Request flow: HTTP → Fiber middleware (extract X-Service-ID, load DB) → Handler → OpenRouter SDK → Response + async conversation logging to DuckDB.

Key packages: config/, service/, database/, handler/, middleware/, openrouter/, models/

Why not X?

Feature	tenantai	LangChain	LiteLLM	OpenRouter Direct
Multi-tenant isolation	Per-service DuckDB	Manual	Manual	None
Single binary	Yes	No (Python)	No (Python)	N/A
Conversation logging	Built-in, async	Manual setup	Manual setup	None
Model access	300+ via OpenRouter	Multiple providers	Multiple providers	300+ models
Streaming	SSE built-in	Framework-dependent	Yes	Yes
Complexity	Minimal, focused	High (large framework)	Moderate	API only
Extensibility	Middleware + interfaces	Plugins/chains	Callbacks	N/A

LangChain is a comprehensive AI framework with chains, agents, and memory abstractions. If you need orchestration primitives, use LangChain. If you need a deployable backend with tenant isolation, use tenantai.

LiteLLM is a Python proxy that normalizes LLM APIs. It focuses on API compatibility across providers. tenantai focuses on multi-tenant isolation and conversation logging as first-class features, deployed as a single Go binary.

OpenRouter Direct gives you model access but no tenant isolation, conversation logging, or service management. tenantai wraps OpenRouter and adds the infrastructure layer.

Build & Test

go build -o tenantai .      # Build
go test ./...                  # Test
go test -race ./...            # Race detection
go test -cover ./...           # Coverage
golangci-lint run              # Lint

Extending

See docs/EXTENDING.md for guides on adding JWT auth, RAG, cost tracking, model routing, prompt versioning, fine-tuning management, and webhook integration.

License

This project is licensed under the Unlicense.