GenRAG is a terminal tool designed to set up a Retrieval-Augmented Generation Pipeline locally from scratch, without utilizing any high-level frameworks like LangChain or vector databases. It includes features such as recursive text splitting, chunking, and building embeddings. The embeddings are stored in a CSV file,without using any Vector Databases and searching is based on cosine similarity.
- Recursive text splitting
- Text chunking
- Building embeddings
- Storing embeddings in a CSV file
- Searching based on cosine similarity
- Inference from local LLM/API
-
Clone the repository:
git clone [repo name] cd GenRAG -
Create and activate a virtual environment:
python3 -m venv env source env/bin/activate # On Windows use `env\Scripts\activate`
-
Install the required dependencies:
pip install -r requirements.txt
-
Add a PDF file to the
datafolder and change the path in the script as necessary. -
Create embeddings by running:
python3 create_embeddings.py
You should see a CSV file generated in the
datafolder. -
Run the main script:
python3 main.py
Enter your query when prompted.
You can use both a local LLM or an LLM from an API like Gemini for generating responses.
-
Local LLM: If you have the capability to run a local LLM, you can use it for generating responses. Cause mine is too slow :(
-
LLM from API: If your system is not powerful enough for local inference, you can use an API like Gemini. To do this, create a
.envfile and pass the Gemini API key.
- Create a
.envfile in the root directory of the project. - Add your Gemini API key to the
.envfile:GEMINI_API_KEY=your_api_key_here
- The system will use the API key for generating responses through the Gemini API.
- Patrick Lewis ., "Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks" arXiv:2005.11401
- Vaswani et al., "Attention is All You Need" arXiv:1706.03762
- Reimers et al., "Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks" arXiv:1908.10084
This project is licensed under the MIT License - see the LICENSE file for details.