Overview1 · Sourcing & Ingestion2 · Parsing & JSON3 · HITL Curation4 · Hybrid Retrieval5 · MCP & Agents6 · Delivery
Enterprise module · flagship
Document Refinery — end-to-end lifecycle
Unstructured financial content becomes governed, machine-ready intelligence: sourced with lineage, refined with layout awareness, curated by stewards, and delivered to search, agents and workflows.
Documents refined (24h)
184,302
+6.2%
Median ingest → ready
3m 41s
−22s
Parsing accuracy
98.4%
+0.3pt
Search nDCG@10
0.812
+0.024
HITL straight-through
96.1%
+1.1pt
Agent tool calls (24h)
1.27M
+18%
Data lifecycle — click any stage to walk through
01
Source & Ingest
Filings, transcripts, statements, research, news via API, SFTP, RSS, web.
02
Parse & Enrich
Layout-aware parsing, OCR, chunking, entity resolution, JSON contracts.
03
Curate (HITL)
Exception routing for low-confidence extractions with full audit trail.
04
Retrieve
Hybrid BM25 + dense vectors, re-ranking, cited bounding boxes.
05
Serve Agents
MCP tools for LLMs and autonomous agents to consume the refinery.
06
Deliver
Spreadsheet & data-toolkit exports, event webhooks to workflows.
Ingestion throughput (k docs / hour)
-14hnow
Live pipeline state
- Refined4
- Parsing1
- In Review3
- Queued0
- Open exceptions5