epistasislab / alzkb-site Goto Github PK

Move the neo4j_v4 branch to this repo and a tag.

Model will be deployed locally on a server.

Performance Requirement:
The model needs to respond within a reasonable amount of time. Multiple methods should be tested and benchmarked:

• Local Model (benchmark different models)
• Cloud (hugginface inference api)

We need to reference the GitHub repository and add the current release version on the alzkb-site about page.

We should moderate the responses from the model to filter out any inappropriate language. We can check if prompt-engineering is enough for this, otherwise we could still see if it makes sense to implement an OpenAI endpoint for moderation

This is similar to the functionality from ExecGPT.
Need to decide on the tech used:

• Server (nodejs or fastapi)
• Database (SQL vs NoSQL)

alzkb.ai is currently deployed from the old repository. We need to deploy from this new repo, for consistency.

Add a GUI for the chat functionality.

Deploy on MongoDB Docker Image.
Build scripts (possibly a combo of python, yaml, sql and shell scripts) to build and deploy a basic Mongo Database for this project.
Some tables to include:

• chats
• messages
• models

We need to add a workflow to automated the unit tests.