Endpoint Details

POST /v1/chat/completions

Send an OpenAI-compatible chat-completions request through Seal.

bash
curl https://api.aqta.ai/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer aqta_your_key" \
  -d '{
    "model": "gpt-4o",
    "messages": [
      {"role": "user", "content": "Summarise this support case in one sentence."}
    ]
  }'

Required fields

Field	Type	Description
`model`	string	A model available through your configured provider credentials
`messages`	array	OpenAI-compatible message objects

Additional OpenAI-compatible fields are forwarded only when supported by the configured provider and pilot configuration.

Response and receipt

The normal completion response includes an extra top-level aqta object. The aqta.attestation field is the signed ATTESTATION-v1 record for the routed decision.

When using an SDK that drops unknown fields, capture the raw HTTP body to obtain this object. The Quick Start shows one Python approach.

GET /v1/attestation/public-key

Retrieve the current public Ed25519 signing key.

bash
curl https://api.aqta.ai/v1/attestation/public-key

The response contains the encoded key, its identifier, algorithm, and format. Trust comes from pinning a verified copy of this key in your own process.

GET /v1/attestation/keys

Retrieve active and historical published key material for handling a key transition.

bash
curl https://api.aqta.ai/v1/attestation/keys

Use this endpoint to learn the issuer's declared key history. Your verifier should still enforce the keys and revocation policy it trusts.

Errors

Expect standard HTTP errors. Use bounded exponential backoff for temporary failures such as 429 or an upstream provider error. Avoid automatic retries when repeating a business action could produce a duplicate side effect.