Ai platforms

This event will be published when conversations have been received from the LLM provider.

Payload

object

idstring
Required

Identifies the event.

minLength = 1
timestring

Timestamp of when the occurrence happened. Must adhere to RFC 3339.

minLength = 1, format = "date-time"
typestring
Required

Unique identifier for the event type.

default = "com.qlik.core.ai-platform.conversations-job.finished"
sourcestring
Required

Identifies the context in which an event happened.

minLength = 1, format = "uri-reference"
specversionstring
Required

The version of the CloudEvents specification which the event uses.

minLength = 1
datacontenttypestring

Content type of the data value. Must adhere to RFC 2046 format.

minLength = 1, default = "application/json"
hoststring

Unique identifier for the host related to the event.
reasonstring

Reason associated with the occurrence, when applicable.
useridstring

Unique identifier for the user related to the event.
owneridstring

Unique identifier for the owner related to the event.
spaceidstring

Unique identifier for the space related to the event.
authtypestring

Representing the type of principal that triggered the occurrence.
clientidstring

Unique identifier for the client related to the event.
tenantidstring
Required

Unique identifier for the tenant related to the event.
sessionidstring

Unique identifier for the session related to the event.
authclaimsstring

A JSON string representing the impersonating entity (the JWT "act" claim), if any.
tracestatestring

A comma-delimited list of key-value pairs.
traceparentstring

Contains a version, trace ID, span ID, and trace options.
toplevelresourceidstring

If the event originated from a sub resource the toplevelresourceid MUST contain the id of the top level resource associated with the sub resource. If the event originated from a top level resource it MUST NOT be defined.
dataobject

Conversations finished event specific data.
Show data properties
- idstring
  Required
  
  A unique identifier for the completion.
- costobject
  
  Estimated token cost in USD. When per-token pricing is configured for the model, the object contains input, output, and total cost values. When pricing is not configured for the model, this property is omitted entirely. Prompt-cache token pricing is not included.
  
  Show cost properties
  
  inputCoststring
  
  Cost of prompt/input tokens in USD.
  
  totalCoststring
  
  Sum of inputCost and outputCost.
  
  outputCoststring
  
  Cost of completion/output tokens in USD.
- _metaobject
  
  Internal metadata associated with the request.
  
  Show _meta properties
  
  actobject
  
  Details of the impersonating entity.
  
  Show act properties
  
  issstring
  
  Issuer of the impersonating entity's identity.
  
  substring
  
  Subject of the impersonating entity.
  
  subTypestring
  
  Type of the impersonating entity's subject.
  
  audarray of strings
  
  Audience(s) the JWT was issued for.
  
  issstring
  
  Issuer of the principal's identity.
  
  sidstring
  
  ID of the session associated with this JWT, if any.
  
  substring
  
  Subject of the principal's identity.
  
  levelstring
  
  Access level of the principal.
  
  userIdstring
  
  Internal ID of the user making the request. Present when subType is "user".
  
  subTypestring
  
  Type of the principal's subject.
  
  clientIdstring
  
  ID of the client making the request.
  
  tenantIdstring
  
  Internal ID of the tenant to restrict access to.
  
  licenseIdstring
  
  License ID associated with the principal.
  
  parentLicenseIdstring
  
  Parent license ID associated with the principal.
- jobIdstring
  
  Optional RFS job ID forwarded from the X-Qlik-RFS-JobID request header to correlate triggered and finished events.
- modelstring
  
  The model used for conversation.
- usageobject
  
  Token usage for a conversations response, including optional prompt-cache token counts.
  
  Show usage properties
  
  totalTokensinteger
  Required
  
  Addition of the completion and prompt tokens.
  
  promptTokensinteger
  Required
  
  Tokens explicitly provided as input to the model.
  
  completionTokensinteger
  Required
  
  Tokens generated by the model in response to the input.
  
  cacheReadInputTokensinteger
  
  Tokens read from the prompt cache.
  
  cacheWriteInputTokensinteger
  
  Tokens written to the prompt cache.
- callIdstring
  
  The call ID associated with this request.
- objectstring
  Required
  
  The object type. Always set to conversation.
- choicesarray of objects
  Required
  
  A list of conversation choices.
  
  Show choices properties
  
  indexinteger
  Required
  
  Position of a specific token within the generated output. Each token in the model's response has an associated index, starting from 0.
  
  messageobject
  
  Individual communication within a conversation. This field (along with toolCalls/functionCall and any other generated content) is redacted and is never present on the published event; it is documented here only for historical/reference purposes.
  
  Show message properties
  
  rolestring
  Required
  
  The role of the message sender, either "user" or "assistant".
  
  Can be one of: "user""assistant"
  
  finishReasonstring
  Required
  
  Reason why a model's response generation process ended.
- traceIdstring
  
  The trace ID associated with this request.
- createdAtinteger
  Required
  
  Unix timestamp of when the response was generated.
- systemFingerprintstring|null
  
  This fingerprint represents the backend configuration that the model runs with.
- performanceMetricsobject
  
  Only present on streaming requests.
  
  Show performanceMetrics properties
  
  ttftstring|null
  Required
  
  Time to first token (TTFT) in seconds. Null when timing data is unavailable.
  
  endTimestring
  Required
  
  Finishing time of the task.
  
  startTimestring
  Required
  
  Starting time of the task.
  
  totalTimestring|null
  Required
  
  Total time of the task in seconds. Null when timing data is unavailable.
  
  inferenceSpeedstring|null
  Required
  
  Tokens per second generated by the model in response to the input. Null when timing data is unavailable.

com.qlik.core.ai-platform.conversations-job.finished

Example payload of the "Conversations finished" event

{
  "id": "A234-1234-1234",
  "time": "2023-01-01T12:00:00Z",
  "type": "com.qlik.core.ai-platform.conversations-job.finished",
  "source": "com.qlik/ai-platform",
  "specversion": "1.0",
  "datacontenttype": "application/json",
  "host": "ai-platform-llm-gateway-844db7d4c7-4glrb",
  "reason": "string",
  "userid": "00000000-0000-0000-0000-000000000000",
  "ownerid": "00000000-0000-0000-0000-000000000000",
  "spaceid": "00000000-0000-0000-0000-000000000000",
  "authtype": "user",
  "clientid": "00000000-0000-0000-0000-000000000000",
  "tenantid": "00000000-0000-0000-0000-000000000000",
  "sessionid": "00000000-0000-0000-0000-000000000000",
  "authclaims": "{\"sub\":\"1234567890\",\"subType\":\"user\",\"iss\":\"qlik.api.internal\"}",
  "tracestate": "congo=t61rcWkgMzE",
  "traceparent": "00-4bf92f3577b34da6a3ce929d0e0e4736-00f067aa0ba902b7-01",
  "toplevelresourceid": "id123",
  "data": {
    "id": "converse-393592ed-ca78-4740-aa87-0ca97480706f",
    "cost": {
      "inputCost": "0.000300",
      "totalCost": "0.003300",
      "outputCost": "0.003000"
    },
    "model": "anthropic.claude-3-sonnet-20240229-v1:0",
    "usage": {
      "totalTokens": 238,
      "promptTokens": 16,
      "completionTokens": 222
    },
    "object": "conversation",
    "createdAt": 1728909071,
    "systemFingerprint": null,
    "performanceMetrics": {
      "ttft": "0.57",
      "endTime": "2024-07-10T15:48:30.416262",
      "startTime": "2024-07-10T15:48:21.598495",
      "totalTime": "8.82",
      "inferenceSpeed": "34.5"
    }
  }
}