BioMANIA-Server

This repository contains the FastAPI app to provide a RESTful API for BioMANIA-backend. It is used in the BioMANIA-Next web application.

Installation

Local Build

You can also run the following code to build and start the server locally:

docker build -t biomania-server .
docker run -p 5001:5001 -d biomania-server

API docs

API docs are work in progress. Use swagger to explore the API by opening the server's /docs URI. For example, for localhost it will be:

http://0.0.0.0:5001/docs

Developer

We build the Docker image using the following command to account for ARM and AMD64 architectures:

docker buildx build --tag biocypher/biochatter-server:latest --tag biocypher/biochatter-server:<version> --platform linux/amd64,linux/arm64 --push .

The version on Dockerhub should match the version in the pyproject.toml file.

Work with API Agent

To introduce Scanpy API Agent, we need to create biochatter.RagAgent and inject into conversation:

from biochatter.rag_agent import RagAgent, RagAgentModeEnum

def _conversation_factory():
    conversation = GptConversation(
        model_name="gpt-4o",
        correct=False,
        prompts={},
    )
    conversation.set_api_key(os.getenv("OPENAI_API_KEY"), user="test")
    return conversation

api_agent = RagAgent(
    mode=RagAgentModeEnum.API_SCANPY,
    conversation_factory=_create_conversation,
    use_prompt=True,
)
chatter.set_rag_agent(api_agent)

response = chatter.query(text)