KnowledgeGPT provides accurate, cited answers from uploaded documents, targeting users who need to quickly extract information and verify sources. It leverages a Langchain-based architecture to process documents and interact with LLMs, offering a user-friendly Streamlit interface for document interaction.

How It Works

The system processes user-uploaded documents, likely chunking them and creating embeddings for efficient retrieval. When a question is posed, it performs a similarity search against the document embeddings to find relevant text snippets. These snippets are then fed to a Large Language Model (LLM) along with the question, prompting the LLM to generate an answer and cite the source text. This approach ensures answers are grounded in the provided documents and their origins are transparent.

Quick Start & Requirements

• Install with poetry install and run with streamlit run main.py.
• Requires Python ≥3.10.
• An OpenAI API key is necessary for optimal performance.
• Docker installation is an alternative for running the application.
• Official Demo: http://localhost:8501 (local)

Highlighted Details

• Accurate answers with instant citations from uploaded documents.
• Built using Streamlit for the UI and Langchain for LLM tooling.
• Supports customization of upload file size via .streamlit/config.toml.

Maintenance & Community

• Active development with a public roadmap.
• Open to contributions.
• Social links include Twitter: https://twitter.com/mm_sasmitha

Licensing & Compatibility

• MIT License.
• Permissive license suitable for commercial use and integration into closed-source projects.

Limitations & Caveats

The current implementation primarily supports document formats that Langchain can process out-of-the-box; support for formats like webpages and PPTX is planned. OCR for scanned documents is also a future enhancement.