RAG Insight

A Streamlit RAG Chatbot for querying PDF documents using a structure-aware chunking approach.

Features

PDF Indexer

The PDF Indexer uses the LLMSherpa API internally for parsing the PDF document. The main sections thus obtained are split recursively using a chunk size of 2048 characters, into subsections that fit within this limit. The hierarchical structure of the document is maintained by returning entire sections rather than arbitrary slices of text. The resulting text chunks are used for building a LlamaIndex query engine, on top of an in-memory VectorStoreIndex.

Streamlit Chatbot

The Streamlit Chatbot allows users to:

1. Input an OpenAI API key
2. Upload a PDF document
3. Ask questions about the document

Installation

Set up LLMSherpa API

1. Install the nlm-ingestor server:
   Follow the instructions at https://github.com/nlmatics/nlm-ingestor

2. The local llmsherpa_url will be: http://localhost:5001/api/parseDocument?renderFormat=all

Deploy the Chatbot

1. Clone the repository

git clone https://github.com/IoanaDragan/rag-insight

2. Install dependencies

pip install -r requirements.txt

3. (Optional) Set up environment variables: Create a .env file and add your OpenAI API key:
   OPENAI_API_KEY=your_api_key_here

4. Launch the Streamlit server

cd rag-insight
streamlit run rag-chatbot.py

5. Access the app at http://localhost:8501 in your browser

Customization

The index_pdf method of the PDFIndexer accepts the following parameters:

Parameter	Description	Default
chunk_size	Size of document chunks	2048
first_n_chunks	Number of chunks to index (for testing)	None (all)
add_summary	Add chunk summaries as metadata	False
retrieve_top_k	Number of similar documents to retrieve per query	2
similarity_threshold	Minimum similarity score for retrieved documents	0.8

License

Distributed under the MIT License. See LICENSE for more information.