Functionary Docker Image

Ready-to-deploy Docker image including the Functionary LLM served via an OpenAI-Compatible API.

[!WARNING] This LLM model supports Function Calling. However It needs some hacking on the client-side because Functionary vLLm script doesn't respect Open-AI API Function Calling format. I will provide an example project on how to achieve that in the coming days.

[!IMPORTANT] There is no possibility to set an API Key at the moment. But I will work on adding that option as an environment variable. In the meantime, don't share your endpoint with anybody.

Specifications

API

• Default Port: 8000
• Path: /v1

Example config for your OpenAI-compatible client (here using a RunPod endpoint):

{
  "model": "musabgultekin/functionary-7b-v1",
  "api_base": "https://[YOUR_CONTAINER_ID]-8000.proxy.runpod.net/v1",
  "api_key": "functionary", // Dummy API Key since it can't be `null`
  "api_type": "open_ai"
}

Docker Image

Base

• nvidia/pytorch

Content

• Python 3.9
• Functionary LLM
• vLLM Server

Deployment

RunPod