OpenChef

OpenChef is a Go-based AI agent orchestrator that embeds NATS for messaging and exposes an OpenAI-compatible HTTP API. It unifies your custom AI logic (across any language or framework) into a single, easy-to-run system—whether you need a simple “direct model” or a sophisticated multi-agent swarm.

---

Why OpenChef?

1. Single Binary, Instant Setup

• No complicated cloud or multi-service overhead—just download/run one binary.
• Embedded NATS means you get robust, high-speed messaging out-of-the-box.

2. Future-Proof Your AI

• Start small with a direct agent (like a single custom LLM) and scale up to orchestrated or swarm-based multi-agent workflows.
• Plug in advanced AI logic or external libraries in Python, Node, or Go—OpenChef unifies them via NATS.

3. OpenAI-Like API

• Your existing tools (LangChain, openai-python) can talk to OpenChef by changing api_base.
• Let your custom “model” or agent appear as if it's a standard OpenAI endpoint, speeding up integration.

4. Collaborative Agent Ecosystem

• Easily register new agents, from a Summarizer to a Database Query bot, and let them cooperate or self-organize.
• Use “Orchestrator” mode for step-by-step control, or “Swarm” mode for fully decentralized, claim-based problem-solving.

5. Simple Local UI (optional)

• A minimal dashboard to see “Registered Agents,” recent requests, or partial logs—helping you debug and track usage.

---

Key Features

1. Embedded NATS – High-performance messaging, zero extra infra needed.
2. OpenAI-Endpoints – POST /v1/chat/completions, returning JSON or partial SSE chunks.
3. Modes:
   • Direct: route to a single agent subject (myAgent)
   • Orchestrator: built-in “supervisor agent” that calls multiple sub-agents in sequence/parallel
   • Swarm: broadcast tasks, advanced “claim” logic for truly distributed solutions
4. Agent Registration – Agents can connect from Python/Node/Go, sign up with a name and subject, handle requests.
5. Optional File Endpoint – If you want a local /v1/files for storing metadata or small assets.

---

Quick Example

# 1) Run OpenChef
./openchef run --port=8080

# 2) A Python agent using NATS
pip install nats-py
python agent.py  # Subscribes to "model.direct.myAgent.request"

# 3) Call OpenChef's endpoint
curl -X POST http://localhost:8080/v1/chat/completions \
     -H "Content-Type: application/json" \
     -d '{
       "model": "myAgent",
       "messages": [{"role": "user", "content": "Hello?"}]
     }'
# => Passes request to your agent, returns AI response in OpenAI JSON


## Quickstart

1. **Download / Build**

```bash
# Example if you have source
git clone https://github.com/yourorg/openchef.git
cd openchef
go build -o openchef cmd/openchef/main.go


## Development

1. Clone the repository
```bash
git clone https://github.com/epuerta9/openchef.git
cd openchef

2. Create your environment file

cp .env.template .env

3. Install dependencies and tools

make setup

4. Start the development server

make dev

Documentation

See the docs directory for detailed documentation.