RAG5 - Ollama, Docker, FastAPI, & Langchain

This is a project for me to learn serveral thing at once

• Python
• Docker
• FastAPI
• Langchain
• SQLite
• Vector databses
• Project archetecture

Feel free to jump in with suggestions and/or observations. I'm learning as I go.

Front End

• using react
• webpack.confif.js - the devServer has a proxy to the backend where context has the variouls routes that match the routes in config/routes.py. e.g. context: ['/api', '/auth', '/users', '/admin', '/home', '/convo'], // Proxy all backend routes

FastAPI-Users

• https://fastapi-users.github.io/fastapi-users/latest/installation/
• https://fastapi-users.github.io/fastapi-users/latest/configuration/full-example/#__tabbed_1_2
• https://github.com/fastapi-users/fastapi-users/tree/master/examples/sqlalchemy

Langchain

• Conceptual Guide: https://python.langchain.com/v0.2/docs/concepts/#langgraph
• https://python.langchain.com/v0.1/docs/use_cases/question_answering/chat_history/#setup
• https://python.langchain.com/v0.1/docs/expression_language/how_to/message_history/
• https://python.langchain.com/v0.2/docs/concepts/#prompt-templates
• https://api.python.langchain.com/en/latest/prompts/langchain_core.prompts.chat.ChatPromptTemplate.html

SQLAlchemy

• DeclaritiveBase - https://docs.sqlalchemy.org/en/20/orm/mapping_api.html#sqlalchemy.orm.DeclarativeBase

• registry - https://docs.sqlalchemy.org/en/20/orm/mapping_api.html#sqlalchemy.orm.registry

Project Description

Create an API that can deployed on Docker that offers private use of Ollama models. The Ollama models will be stored on a volume and the API will be run on a separate container with the following features:

• User accounts - simple email and password
• Conversation Tracker
• Employ Langchain to use the LLMs
• Employ some combination of databases to store documents and the vector files

Refactoring Tasks

• No need for user id in messages at all.