dzakwanf / brag

bRAG [WIP]

bRAG is an example repo how you can do Basic Retrieval Augmented Generation the alternative way. No langchain. No ORMs. No alembic/migration generation tools. Instead we go with PDF. What in the world? pugsql, dbmate & fastapi. That's it.

To better understand this project, you should read the following article: Build your own LLM RAG API without langchain using OpenAI, qdrant and PDF stack

More coming.

Or to be more precise, we will use:

1. plain python openai client to interact with our LLM
2. qdrant as our vector database of choice and their client.
3. For the embeddings we will use openai Qdrant's fastembed lib.
4. the API will be written using fastapi
5. no ORM for us, instead we will roll with pugsql
6. migrations will be done using plain sql and dbMate.
7. instead of poetry, we will go for pyenv + pip-tools
8. on the db front - postgres
9. the app will be containerised - docker
10. simple ingestion with faststream and nats

Does that sound weird to you? Because it is! Even more fun.

Basics

One click deployment

docker-compose up

local development for the api + db & vdb in docker

docker compose up -d database qdrant nats api worker
make run-dev

that's it. Then head over to http://localhost:8000/docs to see the swagger docs.

To run the worker locally with faststream: o run the worker processing the reviews:

faststream run --reload app.process.subscriber:app

lastly to ingest:

python -m app.process.ingest

To get run-dev working proceed through the following steps:

To start, let's make sure you have pyenv installed. What is pyenv? You can read a bit about it on my blog. Pyenv, poetry and other rascals - modern Python dependency and version management.

Long story short it's like a virtualenv but for python versions that keeps everything isolated, without polluting your system interpreter or interpreters of other projects, which would be bad. You don't need to know much more than that.

If you are on mac you can just

brew install pyenv

or if you do not like homebrew/are linux based:

curl https://pyenv.run | bash

Remember to set up your shell for pyenv... In short you have to add some stuff to either your .bashrc, .zshrc or .profile. Why? So the proper 'commands' are available in your terminal and so that stuff works. How?

# For bash:
echo 'export PYENV_ROOT="$HOME/.pyenv"' >> ~/.bashrc
echo 'command -v pyenv >/dev/null || export PATH="$PYENV_ROOT/bin:$PATH"' >> ~/.bashrc
echo 'eval "$(pyenv init -)"' >> ~/.bashrc

# For Zsh:
echo 'export PYENV_ROOT="$HOME/.pyenv"' >> ~/.zshrc
echo 'command -v pyenv >/dev/null || export PATH="$PYENV_ROOT/bin:$PATH"' >> ~/.zshrc
echo 'eval "$(pyenv init -)"' >> ~/.zshrc

Done? Now you just need to get pyenv to get python 3.11 downloaded & installed locally (no worries it won't change your system interpreter):

pyenv install 3.11

When you are done installing pyenv, we can get going.

We will name our project bRAG - the name coming from basic Retrieval Augmented Generation API. First of all, let's create a directory of the project:

mkdir bRAG
cd bRAG

Now we can set up a python version we want to use in this particular directory. How? With pyenv and pyenv-virtualenv, which is nowadays installed by default with the basic pyenv installer. If you need to, check the article I referred before to understand what's happening.

pyenv virtualenv 3.11 bRAG-3-11  # this creates a 'virtualenv' based on python 3.11 named bRAG-3-11
pyenv local bRAG-3-11

after that just install stuff from requirements.txt.

Project structure

The file structure is:

 |-- app/  # Main codebase directory
 |---- chat/
 |------ __init__.py
 |------ api.py
 |------ constants.py
 |------ message.py
 |------ models.py
 |------ streams.py
 
 |---- core/
 |------ __init__.py
 |------ api.py
 |------ logs.py
 |------ middlewares.py
 |------ models.py
 
 |---- __init__.py
 |---- db.py
 |---- main.py
 
 
 |-- db/  # Database/migration/pugsql related code
 |---- migrations/
 |---- queries/
 |---- schema.sql
 
 |-- settings/  # Settings files directory
 |---- base.py
 |---- gunicorn.conf.py
 |-- tests/  # Main tests directory
 |-- requirements/
 
 |-- .dockerignore
 |-- .env.example
 |-- .gitignore
 |-- pre-commit-config.yaml  # for linting in during development
 |-- docker-compose.yml
 |-- Dockerfile
 |-- Makefile  # useful shortcuts
 |-- README.md