Submit an AI output for evaluation

Default mode is async (`sync` omitted or `false`): the log is stored and the endpoint responds immediately with HTTP `202` and `status: pending` while evaluation runs in the background. Pass `sync: true` to evaluate inline and return the full verdict with HTTP `200`. Rate limit: 60 requests per minute per workspace token. Exceeding the limit returns `429` with a `Retry-After: 60` header.

Authorization

webhookToken

AuthorizationBearer <token>

Workspace webhook token from Settings -> Integrations.

In: header

Request Body

application/json

TypeScript Definitions

Use the request body type in TypeScript.

project*string

Project name. Auto-created on first use.

task?|

Segments logs within a project. Used for per-task rule targeting and filtering. Auto-created on first use.

output*string

The AI-generated text or content to evaluate. Up to 100,000 characters.

Lengthlength <= 100000

model?|

The model identifier that produced this output (e.g. gpt-4o, claude-3-5-sonnet). Stored for filtering and display.

latencyMs?|

End-to-end latency of the AI call in milliseconds. Stored for tracking and display.

costUsd?|

Cost of the AI call in US dollars. Stored for cost tracking and display.

input?|array<|||>

The prompt or conversation that produced this output. Accepts plain text or an array of multi-modal content parts (text, image_url, audio_url, file_url).

inputText?|

Plain-text prompt shown alongside the output in the log viewer. Use when input is multi-modal. Up to 100,000 characters.

Lengthlength <= 100000

attachments?array<>|

Files attached to the input. Each file is fetched/validated and durably stored during ingestion. Up to 32 MB per file for url and uploaded sources. Base64 upload is bound by the 4.5 MB request-body cap, so for large files create an upload (POST /api/uploads) and pass { source: "uploaded", key }.

Itemsitems <= 20

callbackUrl?|

URL to POST the evaluation result to when it completes.

Formaturi

callbackEvents?|

Which events trigger the callback. all fires on every completion; flagged_error only on non-pass outcomes; decisions on human approve/reject/change.

Value in"all" | "flagged_error" | "decisions" | "flagged_error_decisions" | "pass"

metadata?

Arbitrary key-value pairs attached to the log. Useful for customer IDs, run metadata, or environment tags.

id?|

Optional caller-supplied log ID for idempotent writes.

revisionOf?|

The id of a log that was sent back via the changed decision. When set, this log is stored as that log's next revision (linked via parentLogId, with revisionNumber incremented). Pass back the logId you received in the decision callback. An unknown id is ignored; the log is still stored, just unlinked.

Lengthlength <= 255

sync?boolean

Evaluation mode. Omit or set false for async (default): the log is stored and the endpoint responds immediately with HTTP 202 and status: pending while evaluation runs in the background. Set true to evaluate inline and return the full verdict with HTTP 200 (evaluation is capped at 30s; a timeout returns 504).

Defaultfalse

confidence?|

The model's self-reported confidence score (0–1). Stored on the log and usable in confidence_threshold rules.

Range0 <= value <= 1

sessionId?|

Optional run or conversation grouping ID. Equivalent context aliases such as conversationId, threadId, and chatId are also accepted.

subjectId?|

Optional impacted entity ID. Equivalent aliases such as customerId, recordId, ticketId, and accountId are also accepted.

actorId?|

Optional actor ID. Equivalent aliases such as userId, endUserId, and agentId are also accepted.

timestamp?|

Optional ISO 8601 timestamp to use as the log date instead of the current server time. Useful for replaying or reprocessing past executions (e.g. 2024-11-15T10:30:00Z).

Formatdate-time

Response Body

`application/json`

curl -X POST "https://example.com/api/logs" \  -H "Content-Type: application/json" \  -d '{    "project": "Support Assistant",    "output": "Thanks for reaching out. We will process your refund within 3-5 business days."  }'

{
  "ok": true,
  "id": "string",
  "verdict": "pass",
  "confidenceScore": 0,
  "explanation": "string",
  "status": "pass"
}

{
  "ok": true,
  "id": "string",
  "status": "pending"
}

{
  "error": "string",
  "code": "string",
  "fields": {
    "property1": "string",
    "property2": "string"
  }
}

{
  "error": "string",
  "code": "string",
  "fields": {
    "property1": "string",
    "property2": "string"
  }
}

{
  "error": "string",
  "code": "string",
  "fields": {
    "property1": "string",
    "property2": "string"
  }
}

{
  "error": "string",
  "code": "string",
  "fields": {
    "property1": "string",
    "property2": "string"
  }
}

{
  "error": "string",
  "code": "string",
  "fields": {
    "property1": "string",
    "property2": "string"
  }
}

{
  "error": "string",
  "code": "string",
  "fields": {
    "property1": "string",
    "property2": "string"
  }
}

{
  "error": "string",
  "code": "string",
  "fields": {
    "property1": "string",
    "property2": "string"
  }
}

{
  "error": "string",
  "code": "string",
  "fields": {
    "property1": "string",
    "property2": "string"
  }
}

{
  "error": "string",
  "code": "string",
  "fields": {
    "property1": "string",
    "property2": "string"
  }
}

Submit an AI output for evaluation

Authorization

Request Body

Response Body

200application/json

202application/json

400application/json

401application/json

402application/json

409application/json

413application/json

422application/json

429application/json

500application/json

504application/json

`application/json`

`application/json`

`application/json`

`application/json`

`application/json`

`application/json`

`application/json`

`application/json`

`application/json`

`application/json`

`application/json`