How ayush-nagar55/veridoc-document-intelligence works
VeriDoc turns uploaded files into validated, reviewable, searchable knowledge, then answers questions with evidence and confidence. The main path runs from the browser through the document API and background pipeline, from format-aware reading through extraction, independent validation, confidence-based review routing, and indexing. Separate user workflows cover human review, grounded multi-document queries, comparison, exports, and analytics. The README documents additional capabilities whose route-to-service wiring is not shown in the sampled source; those are represented without speculative call edges.
Main entry points: pipeline.py Processing pipeline · router.py Route registry · database.py Application records · documents.py Document routes
Components 28
Web application
- Web pages
frontend/src/pages - API client
frontend/src/api/client.ts
API workflows
- API application
backend/app/main.py - Route registry
backend/app/api/router.py - Document routes
backend/app/api/routes/documents.py - Review routes
backend/app/api/routes/review.py - Query routes
backend/app/api/routes/query.py - Comparison routes
backend/app/api/routes/compare.py - Analytics routes
backend/app/api/routes/analytics.py
Document processing
- Processing pipeline
backend/app/services/pipeline.py - Format processors
backend/app/services/processors/factory.py - OCR
backend/app/services/ocr_service.py - Field extraction
backend/app/services/extraction_service.py - LLM provider
backend/app/services/llm/factory.py - Independent validation
backend/app/services/validation_service.py - Review threshold
backend/app/services/confidence_service.py - Quality and summary
backend/app/services/summary_service.py - PII detection
backend/app/services/pii_service.py - Duplicate detection
backend/app/services/duplicate_service.py - Document comparison
backend/app/services/comparison_service.py
Verified knowledge
- Grounded RAG
backend/app/services/rag_service.py - Embedding service
backend/app/services/embedding_service.py - Vector index
backend/app/services/vector_store.py - File storage
backend/app/services/storage.py - Application records
backend/app/database.py - Audit trail
backend/app/services/audit_service.py
Other components
- User
- Configured model
Connections 33
- User → Web pages: uses
- Web pages → API client: requests
- API client → Route registry: HTTP calls
- API application → Route registry: mounts routes
- Route registry → Document routes: dispatches
- Route registry → Review routes: dispatches
- Route registry → Query routes: dispatches
- Route registry → Comparison routes: dispatches
- Route registry → Analytics routes: dispatches
- Document routes → Processing pipeline: starts processing
- Document routes → File storage: stores files
- Processing pipeline → Format processors: reads document
- Format processors → OCR: OCR fallback
- Processing pipeline → Field extraction: extracts fields
- Field extraction → LLM provider: requests extraction
- LLM provider → Configured model: calls provider
- Processing pipeline → Independent validation: validates fields
- Processing pipeline → Review threshold: routes uncertain fields
- Processing pipeline → Quality and summary: computes quality
- Processing pipeline → PII detection: scans text
- Processing pipeline → Duplicate detection: checks duplicates
- Processing pipeline → Grounded RAG: indexes document
- Grounded RAG → Embedding service: embeds text
- Grounded RAG → Vector index: updates and searches
- Query routes → Grounded RAG: asks question
- Query routes → Application records: logs query
- Review routes → Application records: records decisions
- Review routes → Audit trail: records review
- Processing pipeline → Audit trail: records processing
- Comparison routes → Document comparison: compares documents
- Document routes → Application records: reads and writes records
- Processing pipeline → Application records: persists results
- Analytics routes → Application records: reads metrics