GFM-RAG Development

This page is for contributors maintaining the repository and documentation site.

Requirements

Name	Installation	Purpose
Python 3.12	https://www.python.org/downloads/	Runtime and package development
Poetry	https://python-poetry.org/docs/#installation	Dependency management and packaging
CUDA toolkit	NVIDIA or conda packages	Builds the rspmm extension and supports GPU workflows

Local Setup

Bash

poetry install
pre-commit install

If CUDA is installed manually, make sure CUDA_HOME is set:

Bash

export CUDA_HOME=/usr/local/cuda-12.6

Repository Structure

Text Only

gfm-rag/
├── docs/
│   ├── experiment/          # Script-first paper reproduction guides
│   ├── getting_started/     # Minimal user entrypoints
│   ├── workflow/            # General user workflow guides
│   ├── config/              # Configuration reference
│   ├── api/                 # API reference pages
│   ├── DEVELOPING.md
│   └── CHANGELOG.md
├── gfmrag/
│   ├── gfmrag_retriever.py
│   ├── graph_index_construction/
│   ├── graph_index_datasets/
│   ├── models/
│   ├── trainers/
│   ├── workflow/
│   │   ├── config/
│   │   │   ├── gfm_rag/
│   │   │   └── gfm_reasoner/
│   │   ├── index_dataset.py
│   │   ├── sft_training.py
│   │   ├── qa.py
│   │   └── qa_ircot_inference.py
│   ├── evaluation/
│   ├── llms/
│   └── utils/
├── scripts/
│   ├── gfm-rag/
│   └── g-reasoner/
├── tests/
├── mkdocs.yml
├── pyproject.toml
└── poetry.lock

Common Commands

Serve the docs locally:

Bash

mkdocs serve

Run repository checks:

Bash

pre-commit run --all-files

Build the package:

Bash

poetry build