document-extract

Turn a batch of documents into a structured data file you can check.

One-time batch — $249. Up to 200 documents. Delivered within two business days.

Built for operations and finance leads at small businesses and agencies who have a pile of documents and need clean fields out of them.

What you get

• A data file in a schema agreed before work starts
• A per-field audit trail showing where each value came from
• A verification certificate for the batch

Fields that cannot be verified are marked needs_review instead of guessed.

See a real sample

sample/ holds actual pipeline output for a messy synthetic CRM export, so you can judge the deliverable before requesting anything:

• crm_export.csv — the messy synthetic input
• output.csv — the normalized data with a row-level _status
• audit-trail.json — per-field source location, value, method, gate results, and needs-review reasons
• certificate.md — jobspec and source crm_export.csv hashes, aggregate counts, and a human spot-check

13 of 16 fields came back verified. The other 3 are named with typed reasons — a blank email, TBD as a deal value, soon as a close date — rather than guessed.

That is the same package a paid batch ships: one-time batch — $249, up to 200 documents, delivered within two business days. Start a batch request

Not sure whether the batch is scoped tightly enough? Use the document extraction batch readiness checklist before opening a request.

Opening an issue does not purchase anything or start work, and issues here are public and permanent. Do not include credentials, confidential or regulated data, customer data, or document contents — the document type and the fields you need are enough to scope the batch.

How it works

1. Inspect the proof. See what the verification certificate contains and how the audit trail works: Verified document extraction certificate
2. Submit a public request. Open a request issue: Start a batch request
3. Get a private draft. After the request is qualified, a human replies with a private jobspec and payment-link draft.

This is a request form, not checkout

Opening an issue does not purchase anything, does not charge anything, and does not start work. A human reads it and follows up.

Do not put sensitive material in the issue

Issues in this repository are public and permanent. Do not include:

• credentials, API keys, or passwords
• confidential or contract-restricted information
• regulated data (health, personal, or financial account data, or anything under similar rules)
• customer data
• document contents, excerpts, screenshots, or links to private files

Describe the document type and the fields you need. That is enough to scope the batch.