Skip to content

Latest commit

 

History

History
58 lines (36 loc) · 1.09 KB

README.md

File metadata and controls

58 lines (36 loc) · 1.09 KB

Semantic Search and RAG on a FOSS stack

Slides

➡️ rti.github.io/barebone-rag/slides.html

Quickstart

Required software:

  • Docker Engine
  • Docker Compose 2

Start the stack

docker compose up --build --wait

Import the goodwiki dataset

Download a goodwiki dump from huggingface https://huggingface.co/datasets/euirim/goodwiki (direct link)

Rename the file to goodwiki.parquet and run

docker compose run app python import_dump.py

This will take some time. But you can start querying already while it's running.

Start the REPL to query the dataset

docker compose run app python repl.py

Development

Python dependencies

pip install -r requirements.txt

Generate the slides

Slides are generated from markdown using https://marp.app/

marp-cli --watch slides.md

Or, with nix-wrap:

wrap -n nix run nixpkgs#marp-cli -- --watch slides.md