Product
One artifact. Every downstream use.
Corpus2Frame produces a Frame: Field Catalog + Values Tables + Evidence Map, versioned as a data product.
The Frame artifact
Field Catalog
Schema + taxonomy + types + normalization rules. Auto-discovered, human-refinable.
Values Table(s)
Wide or long format. Record/subject/time keyed. Status + confidence per cell.
Evidence Map
Cell-level provenance. Every value traces to exact source spans.
How it plugs in
Warehouse / Lakehouse
Export structured tables directly to Snowflake, BigQuery, Databricks, or S3.
Feature Store
Push typed features to Feast, Tecton, or custom ML pipelines.
Agent Tool Interface
Read/write against fields with policy gates. No free-text edits.
APIs + Batch Modes
REST API for real-time access. Batch processing for corpus-scale refresh.
Governance & control
Pin & Approve
Promote fields into production schema
Confidence Thresholds
Set min confidence per field
Human Gates
Require confirmation for sensitive updates
Audit Logs
Full lineage and change history
See a Frame in action
Explore a live demo with pre-built corpora or upload your own documents.