Repolode
Submit
SaiAkhil066

SaiAkhil066/CORTEX-AI-SUPER-RAG

A local retrieval-augmented generation app that answers questions over your PDFs, Word and text files with citations to file and page, running on Ollama.

View on GitHub
Stars
1,981
Forks
290
Open issues
14
Last commit
last month

What is it?

Cortex RAG lets you upload PDFs, Word documents or text files and ask questions in plain language. Each answer lists the file and page it came from, along with the passage used. Everything runs on your own machine through Ollama, so no API key or cloud upload is needed.

Behind each answer is a retrieval pipeline that combines BM25 keyword search with FAISS vector search, then adds an entity co-occurrence graph and a cross-encoder reranker. Source routing limits search to a single file when a question names it, and follow-up questions are rewritten into standalone queries. Techniques such as RAG-Fusion, HyDE, Corrective RAG and Contextual Retrieval can be switched on or off from the sidebar.

The project also includes an evaluation harness that runs a question set against different pipeline settings and reports hit rate, MRR, correctness and faithfulness. Knowledge bases are saved to disk and reopen automatically. The interface is built with Streamlit, and the project currently has no REST API or multi-user login.

Key features

  • 01Answers cite the source file and page number for each passage used.
  • 02Hybrid retrieval combines BM25 keyword search with FAISS vector search.
  • 03Source routing restricts search to a file when the question names it.
  • 04Cross-encoder reranking orders candidate passages by relevance.
  • 05Built-in evaluation harness compares retrieval settings on your own question set.
  • 06Knowledge bases are saved to disk and can be added to later.

Use cases

  • Querying a folder of internal handbooks, contracts or reports without sending them to a cloud service.
  • Experimenting with RAG techniques and measuring which ones help on your own documents.
  • Prototyping a document Q&A tool on a laptop with Ollama models.

Quick start

  1. 1

    Clone the repository

    git clone https://github.com/SaiAkhil066/CORTEX-AI-SUPER-RAG.git
  2. 2

    Enter the project folder and install Python dependencies

    cd CORTEX-AI-SUPER-RAG pip install -r requirements.txt
  3. 3

    Pull the chat model and the embedding model in Ollama

    ollama pull llama3.1:8b ollama pull nomic-embed-text
  4. 4

    Start the app and open http://localhost:8501

    python -m streamlit run app.py

Things to know

  • Requires Ollama installed and running locally, plus a machine able to run the models.
  • On a laptop, answers with an 8B model can take roughly 30 to 90 seconds.
  • Calculations such as ratios remain weak for the default 8B model, per the README benchmark.

In collections

Private AI knowledge bases and RAGRun LLMs locally

Topics

#rag#ollama#local-llm#document-qa#vector-search#bm25#faiss#streamlit

FAQ

What is CORTEX-AI-SUPER-RAG?
A local retrieval-augmented generation app that answers questions over your PDFs, Word and text files with citations to file and page, running on Ollama. Cortex RAG lets you upload PDFs, Word documents or text files and ask questions in plain language. Each answer lists the file and page it came from, along with the passage used. Everything runs on your own machine through Ollama, so no API key or cloud upload is needed.
Is CORTEX-AI-SUPER-RAG free to use? What license does it have?
CORTEX-AI-SUPER-RAG is open source under the MIT license. Check the license terms before using it in commercial products.
Is CORTEX-AI-SUPER-RAG still maintained?
Updated within 6 months. The last commit to CORTEX-AI-SUPER-RAG was on Sep 28, 2026.
What can I use CORTEX-AI-SUPER-RAG for?
Querying a folder of internal handbooks, contracts or reports without sending them to a cloud service. Experimenting with RAG techniques and measuring which ones help on your own documents. Prototyping a document Q&A tool on a laptop with Ollama models.
How do I install CORTEX-AI-SUPER-RAG?
Clone the repository: git clone https://github.com/SaiAkhil066/CORTEX-AI-SUPER-RAG.git Enter the project folder and install Python dependencies: cd CORTEX-AI-SUPER-RAG pip install -r requirements.txt Pull the chat model and the embedding model in Ollama: ollama pull llama3.1:8b ollama pull nomic-embed-text Start the app and open http://localhost:8501: python -m streamlit run app.py
Where can I get CORTEX-AI-SUPER-RAG?
The source code and installation instructions are on GitHub: https://github.com/SaiAkhil066/CORTEX-AI-SUPER-RAG

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects