DexL Agents – Extract Data

Choose this endpoint when you need to extract a known set of named fields from unstructured or semi-structured text without writing custom parsing logic. It is ideal for documents like invoices, emails, job postings, support tickets, and product descriptions where the schema is known in advance. It is preferable to raw LLM calls when you want a pay-per-call, no-account-required extraction with typed output and a structured response schema. Use alternatives if you need to extract entities without a predefined schema (open-ended NER) or if the input exceeds 8000 characters.

• Text exceeds 8000 character limit — request rejected with validation error
• More than 25 fields requested — schema validation failure
• Field types mismatched to actual content — may return null or empty values for unrecognized fields
• Ambiguous or vague field descriptions — extracted values may be incorrect or missing
• Payment failure via x402 — request not processed
• Model hallucination — fields may be filled with plausible but incorrect values if the text is ambiguous