Deploy document-to-JSON extraction APIs and ETL pipelines with Unstract
Use Unstract when an operator needs agents or automation pipelines to turn recurring PDFs, scans, and document batches into structured JSON through prompt-defined extraction workflows.
npx skills add agentskillexchange/skills --skill deploy-document-to-json-extraction-apis-and-etl-pipelines-with-unstract
Use Unstract when an operator needs a repeatable document extraction workflow: define the fields to extract in Prompt Studio, connect an LLM provider and document source, then expose the extraction as a REST API, ETL pipeline, MCP server, or automation node that returns structured JSON for downstream systems.
What this skill actually does
This is skill-shaped because the agent/operator job is bounded: build, test, and run a document-to-JSON extraction workflow for known document families, then hand off validated structured output to an API consumer, warehouse, or workflow runner. Invoke it when documents need repeatable schema extraction and deployment, not when the user only wants to browse the Unstract product or use a generic LLM SDK.
Inputs and prerequisites: Unstract platform; Docker and Docker Compose for local runs; Git; access to the target documents; configured LLM provider credentials; optional API deployment, ETL, MCP, or n8n integration.
Setup notes: Clone https://github.com/Zipstack/unstract, run ./run-platform.sh for a local platform, open the local frontend, configure provider credentials, define the extraction prompt/schema in Prompt Studio, and deploy the workflow as an API or ETL pipeline according to the upstream documentation.
Source and verification boundary: use https://github.com/Zipstack/unstract and https://docs.unstract.com as the canonical references before running the workflow; keep extraction prompts, schema changes, provider credentials, and produced JSON reviewable against those upstream docs.
Framework fit: publish this as a Multi-Framework workflow because upstream intentionally exposes the same bounded extraction workflow through REST APIs, ETL pipelines, MCP, and automation integrations rather than requiring one agent runtime.