New Relic data dictionary

This data dictionary defines some of our default-reported data stored in NRDB and queryable via NRQL. This includes events like Transaction, MobileRequest, and BrowserInteraction, and our Metric, Log, and Span data types. These definitions are also available from our query builder UI by mousing over applicable data type and attribute names.

This dictionary doesn't contain:

  • Data from our infrastructure integrations (for that, see docs for a specific integration)
  • Custom data
  • Detailed integration-specific attributes for Metric data

LlmChatCompletionMessage

Data source:

An event that corresponds to each message (sent and received) from a chat completion call. These messages include those created by the user, assistant, and system.

Attribute nameDefinitionData types
completion_id
ID

The ID of the LlmChatCompletionSummary event that a message event is connected to.

content

A string value that describes the contents of a message.

id
ID

An agent-generated identifier for the event.

ingest_source

The name of the APM agent that collected the data.

is_response
Boolean

Boolean set to True if a message is the result of a chat completion. The attribute is omitted if False, like in cases where the message is the result of an input.

response.model

The model that returned the response. May or may not match the value for request.model.

role

A string value that describes the source of the message. This varies depending on the LLM vendor, but could have values that represent prompt engineering, the user, or a response from the LLM itself.

sequence

An index associated with each message, including the prompt and responses. Begins at 0.

span_id
ID

The identifier for the span.

token_count

The integer value returned from llm_token_count_callback. Omitted if no value is returned.

trace_id
ID

The identifier for the trace.

vendor

The name of the vendor that provided the model.