Document Workflow Delivery Kit

Turn incoming documents into validated business data. Production-ready FastAPI + Next.js source code for document intake, classification, structured extraction, confidence review, approvals, exports, webhooks, audit logs, and commercial client deployments.

This is a complete document workflow — not merely OCR or an extraction API. Each tier ships a distinct, purpose-built architecture rather than a strict feature ladder, so you buy the one that matches how your client or product actually needs to work with documents.

Three architectures, three approaches

• Developer — Validated Extraction Engine: A resilient, single-pipeline extraction system built around a 14-rule business validation engine and a Human-in-the-Loop (HITL) review queue. This is the tier closest to a classic document-operations pipeline: extract, validate, route low-confidence fields to a human, approve, export, audit.
• Studio — Conversational Document Intelligence: Swaps the fixed validation engine for a LangGraph stateful multi-agent RAG workflow over your ingested documents — multi-turn conversational search with citations, Qdrant hybrid vector retrieval, SSE token streaming — plus Stripe subscription and credit billing for teams building a metered document-AI product.
• Agency — Dynamic Schema Platform: Built for agencies onboarding many clients with different document shapes. Define extraction schemas at runtime per client or collection (no migrations, no redeploys), with a 3-tier schema precedence engine, non-destructive schema versioning, and SchemaGuard protections, plus the same conversational RAG chat capability as Studio.

Core capabilities across the product

• Multi-engine VLM & OCR pipeline: Google Gemini, OpenAI, DeepSeek, Anthropic Claude, Groq, and local air-gapped engines (Docling, PaddleOCR, Ollama Moondream) — swap providers without rewriting extraction logic.
• Confidence-based routing: low-confidence fields route to a review queue instead of silently shipping bad data downstream.
• Owned workflow logic: validation rules, schemas, and review flows live in your codebase, not a vendor's black box — customize freely for invoices, purchase orders, forms, or logistics documents.
• Provider portability: no lock-in to a single OCR/VLM vendor; the extraction layer is pluggable by design.
• Enterprise delivery: dual storage (local disk & Amazon S3), HMAC SHA-256 signed webhooks, streaming CSV/JSON exports, and immutable audit logging ship in every tier.

An honest note on the tiers

Developer, Studio, and Agency are not additive, and a higher price does not mean a more complete codebase. Developer is the only tier with the 14-rule validation engine and human review queue. Studio adds conversational RAG chat and Stripe billing but drops validation entirely. Agency adds dynamic per-client schemas but has neither validation nor billing, and ships with a local-dev auth bypass by default (swap to production JWT/PostgreSQL yourself). This is boilerplate source code, not a finished end-to-end product — each tier is priced for the specific architecture it delivers, not for how much of the others it happens to contain. Pick the tier that matches your actual workflow rather than assuming price tracks completeness. See the FAQ below for the exact breakdown.

What this product is not

This is not a new OCR model — it integrates existing commercial and open-source OCR/VLM providers you bring your own API keys or local models for. It does not claim universal accuracy across every document type out of the box; extraction quality depends on the provider and document shape, which is exactly why the validation engine (Developer) and human review queue exist. It is not a compliance product — teams in regulated domains (legal, healthcare, financial) must add their own controls before relying on it for compliance-sensitive workflows.