Text Extract API gives agents a repeatable document-ingestion workflow for files that need to become reviewable Markdown or structured JSON. The operator runs the FastAPI service with Celery, Redis, OCR strategies, and optional Ollama-backed model passes, then has the agent submit documents, poll results, and preserve the extracted artifacts for downstream review, RAG, or compliance workflows.

## When to use it

Use this instead of opening a document-conversion product manually when an agent needs to process multiple PDFs, Office files, scans, invoices, medical reports, or pictures in a controlled local or server-side pipeline. It is especially useful when source documents should stay inside the operator environment, when OCR output needs model cleanup, or when the workflow includes PII removal before another system consumes the text.

## Scope boundary

This is a document extraction and parsing workflow, not a generic OCR product listing or a broad FastAPI template. The skill starts with a bounded file or batch, runs the selected extraction strategy, optionally asks an Ollama-supported model to normalize or redact the result, and returns Markdown or JSON plus enough task context for human review. Do not use it as a general document management system, storage platform, or unreviewed compliance decision engine.

## Operator workflow

1. Clone the upstream repository and configure the local environment from the provided examples.
2. Start the API stack, Redis/Celery workers, and the chosen OCR or model backend.
3. Submit files through the CLI or API with explicit strategy, prompt, cache, and redaction settings.
4. Poll for task completion and store Markdown or JSON outputs with source filenames and prompts.
5. Review low-confidence OCR, tables, handwriting, math, and any PII redaction before forwarding the result to RAG, analytics, or casework systems.