How shv2312/university-chatbot-and-assistant works
The product is a Streamlit university-policy assistant. Its main workflow accepts uploaded or bundled documents, extracts page-aware text, chunks it with citation metadata, persists embeddings, retrieves relevant passages for a question, and returns a grounded answer with source references. The graph also shows document administration, local ChromaDB persistence, and optional Gemini services. The source excerpts are partial, so some dashboard-management and app-to-answer wiring is documented rather than directly visible in the sampled calls.
Main entry points: app.py Streamlit dashboard · vector_store.py Index and search · rag_pipeline.py Retrieval and guardrails · document_processor.py Page extraction
Components 12
Dashboard
- Streamlit dashboard
app.py - Document management
app.py
Document ingestion
- Page extraction
src/document_processor.py - Metadata chunking
src/document_processor.py
Knowledge base
- Index and search
src/vector_store.py - Persistent vector collection
Grounded answers
- Retrieval and guardrails
src/rag_pipeline.py - Answer and citations
src/rag_pipeline.py
External services
- Google embeddings
- Google Gemini
Other components
- Student or faculty
- Policy documents
Connections 15
- Student or faculty → Streamlit dashboard: asks or uploads
README.md - Policy documents → Streamlit dashboard: uploads or selects
README.md - Streamlit dashboard → Page extraction: extracts text
app.py - Page extraction → Metadata chunking: produces pages
src/document_processor.py - Streamlit dashboard → Metadata chunking: chunks documents
app.py - Streamlit dashboard → Index and search: indexes chunks
app.py - Index and search → Persistent vector collection: persists and searches
README.md - Streamlit dashboard → Document management: manages documents
README.md - Document management → Index and search: lists, deletes, resets
README.md - Streamlit dashboard → Retrieval and guardrails: submits question
app.py - Retrieval and guardrails → Index and search: searches chunks
src/rag_pipeline.py - Index and search → Google embeddings: embeds text
README.md - Retrieval and guardrails → Google Gemini: requests grounded answer
src/rag_pipeline.py - Retrieval and guardrails → Answer and citations: returns answer and sources
src/rag_pipeline.py - Streamlit dashboard → Answer and citations: displays citations
app.py