payload_too_large

​

When this happens

• Raw byte cap. The request body exceeds the API’s configured MAX_REQUEST_SIZE (10 MB by default).
• Per-analyzer token cap. The hero POST /api/v1/analyze/ rejects prompts that exceed the per-analyzer token limits (e.g. 100,000 tokens for the safety / prompt-injection / vector analyzers, 1,000,000 for YARA / SDP / URL).

​

Example response

{
  "detail": "Request body exceeds 10 MB.",
  "code": "payload_too_large",
  "request_id": "5b3f6c7e-7d24-4d40-9b12-3a59c01c6e91",
  "doc_url": "https://docs.lasscyber.com/errors/payload_too_large"
}

{
  "detail": "Prompt exceeds 100,000 token limit for safety_moderation_analyzer.",
  "code": "payload_too_large",
  "request_id": "...",
  "doc_url": "https://docs.lasscyber.com/errors/payload_too_large"
}

​

How to fix

• For prompt analysis: trim the prompt or split it into chunks yourself before sending to Agnes. The classifier already chunks at 400 tokens internally for prompt-injection; the cap is on the input you submit.
• For uploads / threat-intel ingestion: split the payload across multiple requests.

​

Per-analyzer token caps

​

SDK behaviour

​

Related

• Combined analyzer — token limits and chunking.