LLM - Large Language Model

This processes queries from the user by interfacing with the frontend and sending the inputs to a LLM, which can either be stored locally or hosted online. Data is fetched from DMAS and reformatted into a context string which is also passed into the LLM to make the responses more relevant.

Setup

Use Poetry to install all relevant packages. Run poetry install

Download the Gemma model and put the path to it in .env

Running without docker

poetry install
uvicorn llm.app.main:app --reload

Running with docker

docker build -t llm_image .
docker run -d --name llm -p 8081:8081 llm_image

Endpoints:

/query_llm When a post request is made to this endpoint, the LLM will process the given query from the user and return a suggestion based on the question and the plant growth data from the database.

Returns: {"llm_response": llm_response}