How rohiit7762-cloud/hcl_hackathon works
The product is an enterprise knowledge assistant: employees submit questions through Streamlit, the backend resolves trusted identity, decomposes and authorizes tasks, retrieves documents and/or queries structured data, checks evidence, drafts cited answers, and validates them before release. A separate ingestion path accepts uploads and indexes document content or structured data. The README documents the main workflow and security guarantees; several implementation details below are represented only at subsystem level because their source was not sampled.
Main entry points: query_graph.py Query workflow · ingestion.py Ingestion service · dependencies.py Service wiring · task_graph.py Task execution
Components 29
User interface
- Streamlit UI
frontend/streamlit_app.py - Upload page
frontend/pages/1_Upload_documents.py - HTTP client
frontend/api_client.py
API and orchestration
- FastAPI application
backend/app/main.py - Query endpoint
backend/app/api/routes/query.py - Ingest endpoint
backend/app/api/routes/ingest.py - Service wiring
backend/app/api/dependencies.py - Query service
backend/app/services/orchestration.py - Query workflow
backend/app/graph/query_graph.py - Planner and policy check
backend/app/agents/planner.py - Task execution
backend/app/graph/task_graph.py - Evidence assessment
backend/app/agents/evidence.py - Cited answer drafting
backend/app/agents/answer.py - Shared contracts
backend/app/schemas - Runtime configuration
backend/app/core/config.py - Privacy-safe logging
backend/app/core/logging.py
Knowledge ingestion and retrieval
- Document retrieval
backend/app/agents/retrieval.py - Structured-data queries
backend/app/agents/sql.py - Ingestion service
backend/app/services/ingestion.py - Document parsers
backend/app/ingestion/parsers - Semantic chunking
backend/app/ingestion/semantic_chunker.py - Image OCR
backend/app/ingestion/ocr.py - Structured loader
backend/app/ingestion/structured_loader.py
Security and data
- Final validation
backend/app/agents/compliance.py - Data-layer factory
backend/app/services/data_layer.py - Persistence adapters
backend/app/repositories - Access policy
backend/app/security/policy.py - Identity resolution
backend/app/security/identity.py
Other components
- Employee
Connections 28
- Employee → Streamlit UI: asks questions
README.md - Streamlit UI → HTTP client: uses
frontend/streamlit_app.py - Upload page → HTTP client: uploads through
frontend/pages/1_Upload_documents.py - HTTP client → FastAPI application: HTTP requests
frontend/api_client.py - FastAPI application → Query endpoint: mounts routes
backend/app/main.py - FastAPI application → Ingest endpoint: mounts routes
backend/app/main.py - Query endpoint → Service wiring: resolves services
backend/app/api/routes/query.py - Ingest endpoint → Service wiring: resolves services
backend/app/api/routes/ingest.py - Service wiring → Query workflow: builds workflow
backend/app/api/dependencies.py - Service wiring → Data-layer factory: loads factory
backend/app/api/dependencies.py - Query endpoint → Query service: submits query
- Query service → Query workflow: runs workflow
backend/app/services/orchestration.py - Query workflow → Planner and policy check: plans tasks
README.md - Query workflow → Task execution: dispatches tasks
backend/app/graph/query_graph.py - Task execution → Document retrieval: searches documents
README.md - Task execution → Structured-data queries: queries structured data
README.md - Task execution → Evidence assessment: checks evidence
backend/app/graph/task_graph.py - Query workflow → Cited answer drafting: drafts answer
backend/app/graph/query_graph.py - Query workflow → Final validation: validates response
backend/app/graph/query_graph.py - Planner and policy check → Access policy: checks access
- Document retrieval → Access policy: filters evidence
backend/app/agents/retrieval.py - Ingest endpoint → Ingestion service: accepts ingestion
- Ingestion service → Document parsers: parses documents
backend/app/services/ingestion.py - Ingestion service → Semantic chunking: chunks content
backend/app/services/ingestion.py - Ingestion service → Image OCR: extracts image text
backend/app/services/ingestion.py - Ingestion service → Structured loader: loads structured data
backend/app/services/ingestion.py - Data-layer factory → Persistence adapters: provides data access
- Service wiring → Identity resolution: resolves identity
backend/app/api/dependencies.py