Play 38
Document Understanding v2
Advanced document processing — multi-page PDF, table relationships, handwriting, entity linking.
Advanced document processing pipeline for multi-page PDF understanding, table extraction with relationship preservation, handwriting recognition, cross-document entity linking, and structured output generation. Azure AI Document Intelligence handles OCR and layout analysis, Azure OpenAI provides semantic understanding, Blob Storage manages document ingestion, and Cosmos DB stores structured extraction results. Successor to Play 06 with multi-document correlation and entity graph linking.
Architecture Pattern
Advanced DocProc: multi-page, table relationships, entity linking, structured output
Azure Services
DevKit (.github Agentic OS)
- agent.md — root orchestrator with builder→reviewer→tuner handoffs
- 3 agents — Document V2 Builder (gpt-4o), Reviewer (gpt-4o-mini), Tuner (gpt-4o-mini)
- 3 skills — deploy (106 lines), evaluate (107 lines), tune (103 lines)
- 4 prompts — /deploy, /test, /review, /evaluate with agent routing
- .vscode/mcp.json — FrootAI MCP with Doc Intelligence + OpenAI key inputs + envFile
TuneKit (AI Config)
- config/openai.json — extraction and analysis prompts
- config/extraction.json — field schemas, table relationship rules
- config/guardrails.json — PII masking, confidence thresholds
- evaluation/eval.py — Extraction accuracy >90%, Entity linking >85%
Tuning Parameters
Machine evidence
FrootAI evidence lifecycle
This is an internal evidence maturity label, not third-party certification, accreditation, legal compliance, or a production guarantee. Missing or expired evidence demotes automatically; catalog claims cannot promote a play.
This play currently has design evidence only. A runnable scenario, endpoint evaluation, and build receipts are the next contiguous gates.
Repo Intelligence
v1A no-clone, revision-pinned map for agents and humans. Observed evidence is separated from inferred flow so the output stays useful without pretending to be a full call graph.