First make sure that docker desktop is running.
cd ./langchaindocker (Wherever you happen to put the repo)
then run:
docker compose build
docker compose up --detach
(I believe docker compose up will build it if it isn't already, so docker compose build might not be necessary)
The llm-server container should automatically serve ollama
Then you need to find the NAME of the app container. It should be langchaindocker-app-1. Run the following command and the name should be under NAMES
docker container ls
This will show all running containers
You can then open a separate terminal and run
docker attach langchaindocker-app-1
ollama pull nomic-embed-text
ollama pull mistral
(if the NAME is different, use that NAME instead. You can also use another model besides mistral)
Now to run a test. try node test-chromadb.js. If this doesn't work, type npm install then attempt to run the test script again.
When you're done, ctrl-C the server (if it is open) and /bye -> exit the app terminal to leave. The run:
docker compose down
- It has to install mistral every time. It shouldn't have to do this.
- Doesn't recognize GPU yet, I'll need to consult here: https://hub.docker.com/r/ollama/ollama to figure out how to do that with multiple GPU types (automatically recognize)
- Currently it runs with chromadb. Queries are successful, but now I'll need a better interface for uploading documents. It also needs to be integrated with langchain.