Skip to content

Inference Endpoint

POST /v1/chat/completions

Run an inference and receive a cryptographically signed receipt proving execution integrity, routing, and settlement.

Request Format

{
  "model": "default",
  "messages": [
    {"role": "system", "content": "You are a compliance assistant."},
    {"role": "user", "content": "Summarize HIPAA Section 164.312."}
  ],
  "privacy_tier": "sovereign"
}

Required Fields

Field Type Description
model string Model identifier. Use "default" or a specific model ID from your provisioned models.
messages array Array of message objects, each with role and content.

Message Roles

Role Description
system Sets context and behavior constraints
user The user's input
assistant Previous model responses (for multi-turn)

Optional Fields

Field Type Default Description
privacy_tier string "standard" Routing privacy level (see below)
temperature float 1.0 Sampling temperature (0.0 -- 2.0)
max_tokens integer Model default Maximum tokens in the response
stream boolean false Enable streaming responses

Privacy Tiers

SIX supports three privacy tiers that control how your request is routed and isolated.

Tier Routing Data Boundary Use Case
standard Shared compute Provider boundary General workloads, development
sovereign Dedicated compute Your boundary Regulated data (HIPAA, SOX)
confidential Isolated compute Hardware-enforced Maximum isolation requirements

Default Tier

If privacy_tier is omitted, requests route to "standard". For regulated workloads, always specify "sovereign" or "confidential".

Privacy Tier Details
Full routing architecture, isolation guarantees, and hardware attestation documentation for each privacy tier are available to NDA partners.

Request documentation →

Response Format

{
  "id": "six-abc123",
  "object": "chat.completion",
  "created": 1706900000,
  "model": "default",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "HIPAA Section 164.312 establishes technical safeguards..."
      },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 42,
    "completion_tokens": 156,
    "total_tokens": 198
  },
  "receipt": {
    "receipt_id": "rcpt_f8c505ce3a...",
    "routing": "sovereign",
    "commitment": "04a7f3c2...e9b1d0f",
    "signature": "0x69c8...f98c8",
    "verified": true
  }
}

Standard Fields

These match the OpenAI chat-completion response format:

Field Type Description
id string Unique inference identifier
object string Always "chat.completion"
created integer Unix timestamp of execution
model string Model that processed the request
choices array Array of response choices
usage object Token usage counts

Receipt Object

The receipt object is the cryptographic proof. See Receipt Format for field-by-field documentation.

Field Type Description
receipt_id string Unique receipt identifier for later verification
routing string Confirmed routing path (sovereign, standard, confidential)
commitment string Cryptographic commitment binding all receipt fields
signature string Server signature over the commitment
verified boolean Server-side verification result

Examples

curl -X POST https://six-sov.com/v1/chat/completions \
  -H "Authorization: Bearer $SIX_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "default",
    "messages": [
      {"role": "user", "content": "Summarize HIPAA Section 164.312"}
    ],
    "privacy_tier": "sovereign"
  }'

import requests

response = requests.post(
    "https://six-sov.com/v1/chat/completions",
    headers={
        "Authorization": f"Bearer {SIX_API_KEY}",
        "Content-Type": "application/json"
    },
    json={
        "model": "default",
        "messages": [
            {"role": "user", "content": "Summarize HIPAA Section 164.312"}
        ],
        "privacy_tier": "sovereign"
    }
)

data = response.json()
receipt = data["receipt"]
print(f"Receipt ID: {receipt['receipt_id']}")
print(f"Verified: {receipt['verified']}")

from openai import OpenAI

client = OpenAI(
    api_key="YOUR_SIX_API_KEY",
    base_url="https://six-sov.com/v1"
)

response = client.chat.completions.create(
    model="default",
    messages=[
        {"role": "user", "content": "Summarize HIPAA Section 164.312"}
    ],
    extra_body={"privacy_tier": "sovereign"}
)

print(response.choices[0].message.content)
# Receipt available in response metadata

Streaming

When stream: true, the response is delivered as server-sent events (SSE). The receipt is included in the final [DONE] event:

data: {"choices":[{"delta":{"content":"HIPAA"}}]}
data: {"choices":[{"delta":{"content":" Section"}}]}
...
data: {"choices":[{"delta":{"content":"."}}],"finish_reason":"stop"}
data: {"receipt":{"receipt_id":"rcpt_...","verified":true}}
data: [DONE]

Error Handling

See Error Codes for the full error reference.

Common inference-specific errors:

Code Cause Resolution
400 Empty messages array Provide at least one message
422 Invalid privacy_tier value Use standard, sovereign, or confidential
422 temperature out of range Use a value between 0.0 and 2.0
503 Model temporarily unavailable Retry with exponential backoff

Next Steps