Structured Extractor

Structured Extractor

Extract structured data from unstructured text (emails, PDFs, logs, transcripts, scraped HTML) into a validated JSON object matching a target schema.

How to invoke

Give the agent:

1. The raw input text to extract from.
2. The target JSON schema (inline, as a file path, or described in plain language).

Extraction rules

1. Read the schema first. Note required vs optional fields, enums, and format constraints (dates, currencies, IDs). The schema is the contract — never emit a key it does not define.
2. Scan the input for each field. Prefer explicit values over inferred ones. If a required field is genuinely absent, use null rather than guessing.
3. Normalize as you extract:
   • Trim whitespace
   • Coerce dates to ISO 8601 (YYYY-MM-DD or YYYY-MM-DDTHH:mm:ssZ)
   • Strip currency symbols → numeric value + ISO 4217 code (e.g. {"amount": 1500.00, "currency": "USD"})
   • Collapse enum synonyms to their canonical value
4. Emit a single JSON object (or array if the schema is a list). No prose, no markdown fences — just the JSON.
5. Ambiguity handling: pick the most conservative interpretation. If the schema allows additionalProperties, note ambiguities in a top-level "_extraction_notes" field.

Output format

Pure JSON only. No explanation before or after unless the user explicitly asks for one.

Common use cases

• Parse invoice PDFs → structured billing records
• Extract contact info from email signatures
• Convert log lines → typed event objects
• Scrape product data from HTML → catalog entries
• Parse meeting transcripts → action items with owners and due dates