View Source GoogleApi.AIPlatform.V1.Model.LearningServingLlmMessageMetadata (google_api_ai_platform v0.2.0)

LINT.IfChange This metadata contains additional information required for debugging.

Attributes

  • classifierSummary (type: GoogleApi.AIPlatform.V1.Model.LearningGenaiRootClassifierOutputSummary.t, default: nil) - Summary of classifier output. We attach this to all messages regardless of whether classification rules triggered or not.
  • codeyOutput (type: GoogleApi.AIPlatform.V1.Model.LearningGenaiRootCodeyOutput.t, default: nil) - Contains metadata related to Codey Processors.
  • currentStreamTextLength (type: integer(), default: nil) -
  • deleted (type: boolean(), default: nil) - Whether the corresponding message has been deleted.
  • filterMeta (type: list(GoogleApi.AIPlatform.V1.Model.LearningGenaiRootFilterMetadata.t), default: nil) - Metadata for filters that triggered.
  • finalMessageScore (type: GoogleApi.AIPlatform.V1.Model.LearningGenaiRootScore.t, default: nil) - This score is finally used for ranking the message. This will be same as the score present in Message.score field.
  • finishReason (type: String.t, default: nil) - NOT YET IMPLEMENTED.
  • groundingMetadata (type: GoogleApi.AIPlatform.V1.Model.LearningGenaiRootGroundingMetadata.t, default: nil) -
  • isCode (type: boolean(), default: nil) - Applies to streaming response message only. Whether the message is a code.
  • isFallback (type: boolean(), default: nil) - Applies to Response message only. Indicates whether the message is a fallback and the response would have otherwise been empty.
  • langidResult (type: GoogleApi.AIPlatform.V1.Model.NlpSaftLangIdResult.t, default: nil) - Result from nlp_saft DetectLanguage method. Currently the predicted language code and language probability is used.
  • language (type: String.t, default: nil) - Detected language.
  • lmPrefix (type: String.t, default: nil) - The LM prefix used to generate this response.
  • originalText (type: String.t, default: nil) - The original text generated by LLM. This is the raw output for debugging purposes.
  • perStreamDecodedTokenCount (type: integer(), default: nil) - NOT YET IMPLEMENTED. Applies to streaming only. Number of tokens decoded / emitted by the model as part of this stream. This may be different from token_count, which contains number of tokens returned in this response after any response rewriting / truncation.
  • raiOutputs (type: list(GoogleApi.AIPlatform.V1.Model.LearningGenaiRootRAIOutput.t), default: nil) - Results of running RAI on the query or this response candidate. One output per rai_config. It will be populated regardless of whether the threshold is exceeded or not.
  • recitationResult (type: GoogleApi.AIPlatform.V1.Model.LearningGenaiRecitationRecitationResult.t, default: nil) - Recitation Results. It will be populated as long as Recitation processing is enabled, regardless of recitation outcome.
  • returnTokenCount (type: integer(), default: nil) - NOT YET IMPLEMENTED. Number of tokens returned as part of this candidate.
  • scores (type: list(GoogleApi.AIPlatform.V1.Model.LearningGenaiRootScore.t), default: nil) - All the different scores for a message are logged here.
  • streamTerminated (type: boolean(), default: nil) - Whether the response is terminated during streaming return. Only used for streaming requests.
  • totalDecodedTokenCount (type: integer(), default: nil) - NOT YET IMPLEMENTED. Aggregated number of total tokens decoded so far. For streaming, this is sum of all the tokens decoded so far i.e. aggregated count.
  • translatedUserPrompts (type: list(String.t), default: nil) - Translated user-prompt used for RAI post processing. This is for internal processing only. We will translate in pre-processor and pass the translated text to the post processor using this field. It will be empty if non of the signals requested need translation.
  • vertexRaiResult (type: GoogleApi.AIPlatform.V1.Model.CloudAiNlLlmProtoServiceRaiResult.t, default: nil) - The metadata from Vertex SafetyCat processors

Summary

Functions

Unwrap a decoded JSON object into its complex fields.

Types

@type t() :: %GoogleApi.AIPlatform.V1.Model.LearningServingLlmMessageMetadata{
  classifierSummary:
    GoogleApi.AIPlatform.V1.Model.LearningGenaiRootClassifierOutputSummary.t()
    | nil,
  codeyOutput:
    GoogleApi.AIPlatform.V1.Model.LearningGenaiRootCodeyOutput.t() | nil,
  currentStreamTextLength: integer() | nil,
  deleted: boolean() | nil,
  filterMeta:
    [GoogleApi.AIPlatform.V1.Model.LearningGenaiRootFilterMetadata.t()] | nil,
  finalMessageScore:
    GoogleApi.AIPlatform.V1.Model.LearningGenaiRootScore.t() | nil,
  finishReason: String.t() | nil,
  groundingMetadata:
    GoogleApi.AIPlatform.V1.Model.LearningGenaiRootGroundingMetadata.t() | nil,
  isCode: boolean() | nil,
  isFallback: boolean() | nil,
  langidResult: GoogleApi.AIPlatform.V1.Model.NlpSaftLangIdResult.t() | nil,
  language: String.t() | nil,
  lmPrefix: String.t() | nil,
  originalText: String.t() | nil,
  perStreamDecodedTokenCount: integer() | nil,
  raiOutputs:
    [GoogleApi.AIPlatform.V1.Model.LearningGenaiRootRAIOutput.t()] | nil,
  recitationResult:
    GoogleApi.AIPlatform.V1.Model.LearningGenaiRecitationRecitationResult.t()
    | nil,
  returnTokenCount: integer() | nil,
  scores: [GoogleApi.AIPlatform.V1.Model.LearningGenaiRootScore.t()] | nil,
  streamTerminated: boolean() | nil,
  totalDecodedTokenCount: integer() | nil,
  translatedUserPrompts: [String.t()] | nil,
  vertexRaiResult:
    GoogleApi.AIPlatform.V1.Model.CloudAiNlLlmProtoServiceRaiResult.t() | nil
}

Functions

@spec decode(struct(), keyword()) :: struct()

Unwrap a decoded JSON object into its complex fields.