Ingestion lifecycle
Current pipeline
No active ingestion job.
System control
Overview
Not refreshed
Idle
Automation
Daily refresh
Source coverage
PDF crawler
Extraction quality
Dataset completeness
Trust funnel
Review coverage
Exception drivers
Top quality gaps
Managed operation
Refresh committee data
Scrape new committee records, resolve and download missing PDFs, parse documents, and rebuild the database as one persisted background job.
Historical quality repair
Re-adjudicate historical records
Re-runs the current extraction, identity, proposal-type, outcome, and provenance rules across all stored records. It does not crawl or download PDFs.
Parser recovery
Rebuild from local PDFs
Use when a value visible in a downloaded PDF is missing from the stored record. Re-extracts and reparses existing local PDFs, then refreshes the database without crawling CDSCO.
Persistent job
Pipeline progress
No ingestion job is currently running.
Crawler diagnostics
Source inventory
Evidence-first review
Decision queue
Select a proposal
Open a source-backed exception to inspect the evidence, then confirm the outcome in one step or correct it in the full form.
Proposal
Review proposal
Controlled action
Train forecasting model
Training uses only source-verified, automatically validated parser rows. Human-finalized exceptions can publish to intelligence without silently changing the model corpus.
Local intelligence
LLM service
Model registry
Recent training runs
Class coverage