How vijay-rodge/polarnexus works
This is an NCPOR polar-science knowledge, analysis, and outreach portal. Users enter through the web portal or REST API; assistant queries are normalized and classified, routed to retrieval, scientific analysis, or catalog searches, then synthesized and citation-validated. Separate product branches support DSpace harvesting and manuscript extraction, scientific data analysis, catalog exploration, media, and educational outreach. The README documents several capabilities whose page or engine internals are not sampled; those relationships are shown only where documented, with uncertain wiring omitted or dashed.
Main entry points: graph.py LangGraph router · controlled_tools.py Controlled data tools · app.py Streamlit portal · dspace_search_harvester.py DSpace harvesting
Components 23
User interfaces
- Streamlit portal
streamlit_app/app.py - AI assistant
streamlit_app/pages/01_assistant.py - Catalog explorers
streamlit_app/pages/02_stations.py - Outreach guides
streamlit_app/pages/08_outreach.py - REST API
api/main.py
Query assistant
- LangGraph router
agents/graph.py - Query normalization
agents/nodes/normalize.py - Intent routing
agents/nodes/classify.py - Document retrieval
agents/nodes/retrieve_rag.py - Data analysis route
agents/nodes/analyze_data.py - Catalog search routes
agents/nodes/search_nodes.py - Answer synthesis
agents/nodes/generate_answer.py - Citation validation
agents/nodes/validate_citations.py
Science and catalogs
- Controlled data tools
agents/tools/controlled_tools.py - Scientific analysis
scientific_engine/analyzer.py - Manuscript analysis
scientific_engine/document_analyzer.py - Domain classification
ml/classification/domain_classifier.py
Repository ingestion
- DSpace harvesting
crawler/dspace_search_harvester.py - Repository crawler
crawler/engine.py
Data and retrieval
- Polar metadata database
database/models.py - Chroma knowledge store
rag/vectorstore/chroma_store.py
Other components
- Portal user
- NCPOR DSpace
Connections 27
- Portal user → Streamlit portal: uses
- Portal user → REST API: queries
- Streamlit portal → AI assistant: offers
- AI assistant → LangGraph router: submits query
- REST API → LangGraph router: executes query
api/main.py - LangGraph router → Query normalization: normalizes
agents/graph.py - LangGraph router → Intent routing: classifies
agents/graph.py - LangGraph router → Document retrieval: routes to
agents/graph.py - LangGraph router → Data analysis route: routes to
agents/graph.py - LangGraph router → Catalog search routes: routes to
agents/graph.py - Document retrieval → Controlled data tools: searches documents
agents/nodes/retrieve_rag.py - Data analysis route → Controlled data tools: requests analysis
agents/nodes/analyze_data.py - Catalog search routes → Controlled data tools: searches catalogs
agents/nodes/search_nodes.py - LangGraph router → Answer synthesis: dispatches synthesis
agents/graph.py - LangGraph router → Citation validation: dispatches validation
agents/graph.py - Controlled data tools → Scientific analysis: computes statistics
agents/tools/controlled_tools.py - Controlled data tools → Polar metadata database: queries metadata
agents/tools/controlled_tools.py - Controlled data tools → Chroma knowledge store: searches chunks
agents/tools/controlled_tools.py - Answer synthesis → Manuscript analysis: uses coordinates
agents/nodes/generate_answer.py - DSpace harvesting → NCPOR DSpace: harvests records
- Repository crawler → DSpace harvesting: complements harvest
- Repository crawler → Manuscript analysis: extracts manuscript data
- DSpace harvesting → Manuscript analysis: extracts coordinates and figures
crawler/dspace_search_harvester.py - DSpace harvesting → Chroma knowledge store: stores text chunks
crawler/dspace_search_harvester.py - Streamlit portal → Catalog explorers: provides explorers
- Streamlit portal → Outreach guides: provides outreach
- Domain classification → Polar metadata database: tracks model metadata