Fullstack AI Chatbot

A demo of a fullstack AI chatbot using FastAPI, Redis and Hugging Face Inference API.

• FastAPI is used as the webserver talking to Client with REST and Websocket endpoints. The server uses two concurrent coroutines to talk to the clients and Redis streams.
• Client is a simple js/jquery based UI.
• Redis enterprise cloud DB (free) instance is used as cache to save chat data. Redis is also used for message queues (message channel and response channel) between Server and Worker.
• Worker connects to Redis streams/channels to communicate with server and also connects to Hugging Face On Demand Inference API to query a hosted model.
• The backend is based on an in depth tutorial by Stephen Sanwo written in collaboration with Redis. The frontend UI is based on a simple UI project.

Fig: Architecture showing components and interfaces

Environment variables:

Set app environment to development (for dev). export APP_ENV=development

Add a .env file under /server and one under /worker.
Add Redis config to both files:
export REDIS_URL=
export REDIS_USER=
export REDIS_PASSWORD=
export REDIS_HOST=
export REDIS_PORT=

Add Hugging Face config to .env under worker:
export HUGGINGFACE_INFERENCE_TOKEN=
export MODEL_URL=

UI Sample: