atkouti / sematic

Context-enhanced question answering, local setup: GPT4ALL, HuggingFace Embeddings model, FAISS, LangChain

Sematic

This Jupyter notebook represents my attempt to implement a context-enhanced question answering setup using open source tools that can be executed locally:

• The HuggingFace model all-mpnet-base-v2 is utilized for generating vector representations of text
• The resulting embedding vectors are stored, and a similarity search is performed using FAISS
• Text generation is accomplished through the utilization of GPT4ALL.

The objective of this personal project is to address the issue of data that cannot be shared with OpenAI or similar APIs. Additionally, it serves as my initial encounter with LangChain, a framework designed for developing applications powered by language models.

Running the notebook

To run the notebook, you may try accessing it through Google Colab or import the .ipynb file from this repository into a new Google Colab environment. Subsequently, please refer to the instructions provided within the notebook itself and/or the accompanying Youtube video for guidance.