Milestones
List view
Establish the foundational data layer ingest all industrial document types and build the entity-linked knowledge graph.Issues to create: Implement PDF ingestion pipeline Build pdf_processor.py to parse engineering specs, OEM manuals, MSDS sheets using LlamaIndex. Support chunking, metadata extraction, and storage to MinIO + Qdrant. Build P&ID parser with YOLOv8 Fine-tune YOLOv8 to detect P&ID symbols, extract equipment tags, and pipe structured output to the knowledge graph via pid_processor.py. OCR pipeline for scanned forms Integrate Azure Document Intelligence in spreadsheet_processor.py and email_processor.py to handle inspection logs, permit-to-work forms, and email archives. Neo4j entity extractor & graph builder Implement entity_extractor.py and graph_builder.py to auto-create Equipment → Procedure → Risk relationships from ingested documents. Batch ingestion API endpoint Wire up POST /ingest/batch with async job queuing (Redis) and GET /ingest/status/{job_id} for progress tracking.
Overdue by 2 month(s)•Due by June 24, 2026