Tender Intake Evidence Extractor

Use this Actor for tender document extraction from supplied English text into a bounded intake row with source evidence. It extracts project, issuer, site, submission-date text, and required documents, while retaining document IDs, page references, and source quotes.

Missing values stay missing. Conflicting values remain flagged for human review; the Actor does not select a deadline or provide legal advice.

Quick start

1. Supply one to five English text documents.
2. Include an id and text for each document.
3. Add page_count, page_number, and source_url when available.
4. Run the Actor and review the extracted values, evidence, and diagnostics.

{
  "documents": [
    {
      "id": "invitation",
      "text": "Project: Riverside Library\nIssuer: Example Council\nSite: 1 Main Street\nReturn by 12 September 2026 at noon.\nRequired documents: tender form, safety plan",
      "page_count": 1,
      "page_number": 1
    }
  ]
}

What it returns

The output contains an intake row with extracted fields and evidence for each value: document ID, page number, and source quote. It also reports missing fields, conflicting values, and input diagnostics so a reviewer can check the source material before using the result.

Limits

• Accepts supplied English text; it does not fetch source files or perform PDF parsing or OCR.
• Each package is limited to 5 documents and 100 declared pages.
• This release is limited to synthetic or public non-confidential text until a separate data-handling decision is made for customer documents.
• It does not accept credentials, inbox or CRM connections, drawings, or quantities.
• It provides no legal, bid, compliance, or contract advice.

Responsible use

Keep a human reviewer in the loop. Check every important extracted value and its source quote against the original tender material before acting on it.

Trial and support

Start with the synthetic invitation above. Expect fields.project.value to be Riverside Library and needs_review to include timezone_not_stated. Each field has an extraction status and source evidence. Extraction uses supported text labels and date phrases; other layouts can leave fields missing. It does not infer arbitrary tender requirements. page_number is a reference supplied by you, not a page detected by a PDF reader. Check the Pricing tab and run cost estimate before starting; platform usage may cost money. Report unexpected results through the Issues tab using a synthetic or public non-confidential example.