Product Overview¶
Last verified: 2026-07-26 against dev at f0388abb.
Smart Document Processing turns document sets into structured, reviewable outputs. Users upload documents into an assessment, the system classifies and processes them through configured extractors/schemas, and analysts review, correct, validate, and export extracted data.
Product Promise¶
Help teams process messy document packages faster while keeping the extracted results reviewable, auditable, and scoped to the right organization.
Primary Jobs¶
- Create an assessment for a reporting/review workflow.
- Upload and validate documents.
- Process documents with registered extractors and schemas.
- Track batch/run/document processing status.
- Review extracted items and fields.
- Correct or validate extracted results.
- Export structured outputs.
- Evaluate extraction/checklist quality with repeatable experiments.
Main Product Surfaces¶
- React/Vite frontend for assessments, document workflows, validation, and extraction/redaction mode UI.
- FastAPI backend for assessments, documents, batches, processors, extracted items, corrections, exports, tenancy, and auth.
- Azure Function extractor apps triggered by Service Bus queues.
- Schema and extractor registries for processor generation.
- Phoenix/Arize evaluation tooling for extraction and checklist accuracy.
Boundaries¶
SDP is a shared document-processing engine. It is not the Research Assistant chat/RAG product and is not the AltaForge control plane. Consumer applications can call SDP, but browser clients must not carry long-lived LLM keys; use a server-side BFF for BYOK flows.
Where to Go Next¶
- For local setup, use the first local run tutorial.
- For API facts, use API surface.
- For architecture context, use Architecture.