Practical overview
When it fits
Use Kg Rag when you want to explore a locally configured, graph-enriched language-model workflow, particularly for disease-related biomedical questions.
Who it is for
- Biomedical and bioinformatics researchers exploring graph-grounded language-model answers.
- Developers who want to study or adapt a notebook-based knowledge-graph RAG workflow.
- Technical users comfortable configuring a local Python project and its model settings.
Documented uses
- Add biomedical knowledge-graph context to prompts about diseases.
- Compare language-model prompting with and without retrieved graph context.
- Run a documented GPT or Llama generation workflow from a local terminal.
- Explore adapting graph-backed retrieval to knowledge-intensive questions in another domain.
Requirements & limits
The code repository is public and identifies the Apache-2.0 license.
- The README describes the framework broadly as task-agnostic, but its run instructions say the implementation is currently designed for disease-related prompts.
- Setup requires a local Python environment, dependencies, and configuration in config.yaml; setup creates a disease vector database.
- The project documents GPT workflows through OpenAI or Azure and a separate Llama workflow.
- Replacing or extending the biomedical graph is a possible adaptation, but the repository materials do not document a tested graph migration.
- This is a research codebase with notebook and terminal workflows, rather than a documented hosted service or API.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Set up the disease knowledge-graph data
Prepare the local repository by creating its disease vector database, with an optional Llama model download.
Before you start
- A local Python environment; the README specifies Python 3.10.9 for its scripts.
- The KG_RAG repository and its requirements installed.
- A populated config.yaml file.
Steps
Clone the repository and enter its directory.
Create and activate a Python 3.10.9 environment.
Install dependencies with pip install -r requirements.txt.
Update config.yaml with the required local configuration.
From the KG_RAG directory, run python -m kg_rag.run_setup and follow the interactive setup; choose whether to download Llama.
Expected outcome
The setup script creates the disease vector database; downloading Llama is optional.
Guide limitations
- The README says the implementation is currently designed for disease-related prompts.
- This setup recipe does not run a generation prompt.
Run a GPT-based KG-RAG prompt from the terminal
Use the repository's documented GPT text-generation entry point after completing setup and configuration.
Before you start
- A configured KG_RAG checkout with dependencies installed and setup completed.
- GPT_API_TYPE configured for OpenAI or Azure, as indicated in the README.
- A GPT model selection supported by the documented command examples.
Steps
From the KG_RAG directory, select the OpenAI or Azure GPT_API_TYPE configuration.
Run the documented kg_rag.rag_based_generation.GPT.text_generation module with the -g argument and a supported model name, such as gpt-4 or the README's provider-specific GPT-3.5 Turbo name.
For the documented interactive mode, add -i True to the command.
Expected outcome
The repository starts its GPT-based generation workflow; interactive mode lets the user go through each process step interactively.
Guide limitations
- The README's run instructions describe a disease-focused implementation.
- The documented generation example is a local terminal workflow, not a hosted chat service.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 15 October 2026
- Original website title
- BaranziniLab/KG_RAG · Official homepage title · retained in its source language