Cubby Chat is a micro-services container based demo app. The use case is an AI Chat interface, but it also works in a non-AI mode for when you need it to load quickly and/or in a small resource footprint. AI is provided by an ollama container which loads tinyllama as model by default.
The app has 4 micro-services:
- React frontend
- Go backend
- Postgres database to store chat history
- Ollama for LLM chat (optional)
Everything should build and run with:
docker compose up -d
To run with AI, use this:
OLLAMA_ENABLED=true docker compose --profile ai up
This app was created to help demonstrate ConfigHub. Follow this example to see how you can manage a global deployment footprint of Cubby Chat with ConfigHub.
- Check this for enabling GPU: https://www.substratus.ai/blog/kind-with-gpus
- ...but also this: https://chariotsolutions.com/blog/post/apple-silicon-gpus-docker-and-ollama-pick-two/