QA with Documents (Streamlit + LlamaIndex)

Retrieval-augmented QA app that lets you upload a document (PDF/TXT), indexes it with LlamaIndex + OpenAI embeddings, and answers questions via an OpenAI LLM. Frontend is Streamlit.

Quickstart

# 1) Create and activate venv (example)
python3 -m venv venv
source venv/bin/activate

# 2) Install deps
pip install -r requirements.txt

# 3) Set your OpenAI key
export OPENAI_API_KEY="sk-..."
# or create a .env file with OPENAI_API_KEY=...

# 4) Run the app
streamlit run StreamlitApp.py

How it works

• Upload a document in the UI; it is saved into Data/.
• load_data reads the file(s) and produces documents.
• create_or_load_index builds (or reloads) a vector index using OpenAI embeddings and persists it to storage/.
• Queries are sent to the index; responses come from the OpenAI LLM.

Project layout

• StreamlitApp.py – Streamlit UI and request flow
• QAWithPDF/data_ingestion.py – loads uploaded files into LlamaIndex documents
• QAWithPDF/embedding.py – builds/loads the vector index and sets embedding/LLM settings
• QAWithPDF/model_api.py – loads the OpenAI LLM
• QAWithPDF/logger.py – basic logging configuration
• requirements.txt – full dependency list
• storage/ – persisted index artifacts (created at runtime)
• Data/ – uploaded documents are written here

Configuration

• OPENAI_API_KEY must be set in your environment or .env.
• Adjust embedding/LLM models or chunking in QAWithPDF/embedding.py if needed.

Troubleshooting

• If imports fail, ensure you run from the project root so QAWithPDF is on PYTHONPATH.
• Delete storage/ to rebuild the index from fresh documents if you change embeddings or data.