# Google GenAI (Gemini) Generation API Schemas

GeminiGenerationRequest:
  type: object
  properties:
    model:
      type: string
      description: Model field for explicit model specification
    contents:
      type: array
      items:
        $ref: '#/GeminiContent'
      description: Content for the model to process
    systemInstruction:
      $ref: '#/GeminiContent'
      description: System instruction for the model
    generationConfig:
      $ref: '#/GeminiGenerationConfig'
    safetySettings:
      type: array
      items:
        $ref: '#/GeminiSafetySetting'
    tools:
      type: array
      items:
        $ref: '#/GeminiTool'
    toolConfig:
      $ref: '#/GeminiToolConfig'
    cachedContent:
      type: string
      description: Cached content resource name
    labels:
      type: object
      additionalProperties:
        type: string
      description: Labels for the request
    requests:
      type: array
      items:
        $ref: '#/GeminiEmbeddingRequest'
      description: Batch embedding requests
    # Bifrost-specific
    fallbacks:
      type: array
      items:
        type: string

GeminiContent:
  type: object
  properties:
    role:
      type: string
      enum: [user, model]
      description: The producer of the content. Must be either 'user' or 'model'
    parts:
      type: array
      items:
        $ref: '#/GeminiPart'
      description: List of parts that constitute a single message

GeminiPart:
  type: object
  properties:
    text:
      type: string
      description: Text part (can be code)
    thought:
      type: boolean
      description: Indicates if the part is thought from the model
    thoughtSignature:
      type: string
      format: byte
      description: Opaque signature for thought that can be reused in subsequent requests
    inlineData:
      $ref: '#/GeminiBlob'
    fileData:
      $ref: '#/GeminiFileData'
    functionCall:
      $ref: '#/GeminiFunctionCall'
    functionResponse:
      $ref: '#/GeminiFunctionResponse'
    executableCode:
      $ref: '#/GeminiExecutableCode'
    codeExecutionResult:
      $ref: '#/GeminiCodeExecutionResult'
    videoMetadata:
      $ref: '#/GeminiVideoMetadata'

GeminiBlob:
  type: object
  properties:
    mimeType:
      type: string
      description: The IANA standard MIME type of the source data
    data:
      type: string
      format: byte
      description: Base64-encoded raw bytes
    displayName:
      type: string
      description: Display name of the blob (not currently used in GenerateContent calls)

GeminiFileData:
  type: object
  properties:
    mimeType:
      type: string
      description: The IANA standard MIME type of the source data
    fileUri:
      type: string
      description: URI of the file
    displayName:
      type: string
      description: Display name of the file data

GeminiFunctionCall:
  type: object
  properties:
    id:
      type: string
      description: Unique ID of the function call. If populated, client should return response with matching id
    name:
      type: string
      description: The name of the function to call. Matches FunctionDeclaration.name
    args:
      type: object
      description: Function parameters and values in JSON object format

GeminiFunctionResponse:
  type: object
  properties:
    id:
      type: string
      description: ID of the function call this response is for. Matches FunctionCall.id
    name:
      type: string
      description: The name of the function. Matches FunctionDeclaration.name and FunctionCall.name
    response:
      type: object
      description: Function response in JSON object format. Use "output" key for output and "error" key for error details
    willContinue:
      type: boolean
      description: Signals that function call continues (NON_BLOCKING only). If false, future responses will not be considered
    scheduling:
      type: string
      description: How the response should be scheduled (NON_BLOCKING only). Defaults to WHEN_IDLE

GeminiExecutableCode:
  type: object
  properties:
    language:
      type: string
      description: Programming language of the code
    code:
      type: string
      description: The code to be executed

GeminiCodeExecutionResult:
  type: object
  properties:
    outcome:
      type: string
      enum: [OUTCOME_UNSPECIFIED, OUTCOME_OK, OUTCOME_FAILED, OUTCOME_DEADLINE_EXCEEDED]
      description: Outcome of the code execution
    output:
      type: string
      description: Contains stdout when successful, stderr or other description otherwise

GeminiVideoMetadata:
  type: object
  properties:
    fps:
      type: number
      description: Frame rate of the video. Range is (0.0, 24.0]
    startOffset:
      type: string
      description: Start offset of the video
    endOffset:
      type: string
      description: End offset of the video

GeminiGenerationConfig:
  type: object
  properties:
    temperature:
      type: number
      description: Controls the randomness of predictions
    topP:
      type: number
      description: Nucleus sampling parameter
    topK:
      type: integer
      description: Top-k sampling parameter
    candidateCount:
      type: integer
      description: Number of candidates to generate. Defaults to 1
    maxOutputTokens:
      type: integer
      description: Maximum number of output tokens to generate per message
    stopSequences:
      type: array
      items:
        type: string
      description: Stop sequences
    responseMimeType:
      type: string
      description: Output response mimetype (text/plain, application/json)
    responseSchema:
      $ref: '#/GeminiSchema'
      description: Schema for JSON response (OpenAPI 3.0 subset)
    responseJsonSchema:
      type: object
      description: Alternative to responseSchema using JSON Schema format
    responseModalities:
      type: array
      items:
        type: string
        enum: [MODALITY_UNSPECIFIED, TEXT, IMAGE, AUDIO]
      description: The modalities of the response
    speechConfig:
      $ref: '#/GeminiSpeechConfig'
    thinkingConfig:
      $ref: '#/GeminiThinkingConfig'
    frequencyPenalty:
      type: number
      description: Frequency penalty for token generation
    presencePenalty:
      type: number
      description: Presence penalty for token generation
    seed:
      type: integer
      description: Seed for deterministic generation
    logprobs:
      type: integer
      description: Number of log probabilities to return
    responseLogprobs:
      type: boolean
      description: If true, export logprobs results in response
    audioTimestamp:
      type: boolean
      description: If enabled, audio timestamp will be included in request
    mediaResolution:
      type: string
      description: Media resolution specification
    routingConfig:
      $ref: '#/GeminiRoutingConfig'
    modelSelectionConfig:
      $ref: '#/GeminiModelSelectionConfig'
    enableAffectiveDialog:
      type: boolean
      description: If enabled, model will detect emotions and adapt responses

GeminiSchema:
  type: object
  description: Schema object for defining input/output data types (OpenAPI 3.0 subset)
  properties:
    type:
      type: string
      enum: [TYPE_UNSPECIFIED, STRING, NUMBER, INTEGER, BOOLEAN, ARRAY, OBJECT, 'NULL']
      description: The type of the data
    format:
      type: string
      description: Format of the data (e.g., float, double, int32, int64, email, byte)
    title:
      type: string
      description: The title of the Schema
    description:
      type: string
      description: The description of the data
    nullable:
      type: boolean
      description: Indicates if the value may be null
    enum:
      type: array
      items:
        type: string
      description: Possible values for primitive types with enum format
    properties:
      type: object
      additionalProperties:
        $ref: '#/GeminiSchema'
      description: Properties of Type.OBJECT
    required:
      type: array
      items:
        type: string
      description: Required properties of Type.OBJECT
    items:
      $ref: '#/GeminiSchema'
      description: Schema of the elements of Type.ARRAY
    minItems:
      type: integer
      description: Minimum number of elements for Type.ARRAY
    maxItems:
      type: integer
      description: Maximum number of elements for Type.ARRAY
    minLength:
      type: integer
      description: Minimum length of Type.STRING
    maxLength:
      type: integer
      description: Maximum length of Type.STRING
    minimum:
      type: number
      description: Minimum value of Type.INTEGER and Type.NUMBER
    maximum:
      type: number
      description: Maximum value of Type.INTEGER and Type.NUMBER
    pattern:
      type: string
      description: Pattern to restrict a string to a regular expression
    default:
      description: Default value of the data
    example:
      description: Example of the object (only populated when object is root)
    anyOf:
      type: array
      items:
        $ref: '#/GeminiSchema'
      description: Value should be validated against any of the subschemas
    propertyOrdering:
      type: array
      items:
        type: string
      description: Order of the properties (not standard OpenAPI)
    minProperties:
      type: integer
      description: Minimum number of properties for Type.OBJECT
    maxProperties:
      type: integer
      description: Maximum number of properties for Type.OBJECT

GeminiThinkingConfig:
  type: object
  properties:
    includeThoughts:
      type: boolean
      description: Whether to include thoughts in the response
    thinkingBudget:
      type: integer
      description: Thinking budget in tokens
    thinkingLevel:
      type: string
      enum: [THINKING_LEVEL_UNSPECIFIED, LOW, HIGH]
      description: Thinking level preset

GeminiRoutingConfig:
  type: object
  properties:
    autoMode:
      type: object
      properties:
        modelRoutingPreference:
          type: string
          description: Model routing preference
    manualMode:
      type: object
      properties:
        modelName:
          type: string
          description: Model name to use

GeminiModelSelectionConfig:
  type: object
  properties:
    featureSelectionPreference:
      type: string
      description: Options for feature selection preference

GeminiSpeechConfig:
  type: object
  properties:
    voiceConfig:
      $ref: '#/GeminiVoiceConfig'
    multiSpeakerVoiceConfig:
      $ref: '#/GeminiMultiSpeakerVoiceConfig'
    languageCode:
      type: string
      description: Language code (ISO 639) for speech synthesis. Only available for Live API

GeminiVoiceConfig:
  type: object
  properties:
    prebuiltVoiceConfig:
      type: object
      properties:
        voiceName:
          type: string
          description: The name of the prebuilt voice to use

GeminiMultiSpeakerVoiceConfig:
  type: object
  properties:
    speakerVoiceConfigs:
      type: array
      items:
        type: object
        properties:
          speaker:
            type: string
            description: The name of the speaker (should match prompt)
          voiceConfig:
            $ref: '#/GeminiVoiceConfig'

GeminiSafetySetting:
  type: object
  properties:
    category:
      type: string
      description: Harm category
    threshold:
      type: string
      description: The harm block threshold
    method:
      type: string
      description: Determines if harm block uses probability or probability and severity scores

GeminiTool:
  type: object
  properties:
    functionDeclarations:
      type: array
      items:
        $ref: '#/GeminiFunctionDeclaration'
    googleSearch:
      $ref: '#/GeminiGoogleSearch'
    googleSearchRetrieval:
      $ref: '#/GeminiGoogleSearchRetrieval'
    retrieval:
      $ref: '#/GeminiRetrieval'
    codeExecution:
      type: object
      description: Enables code execution by the model
    enterpriseWebSearch:
      $ref: '#/GeminiEnterpriseWebSearch'
    googleMaps:
      $ref: '#/GeminiGoogleMaps'
    urlContext:
      type: object
      description: Tool to support URL context retrieval
    computerUse:
      $ref: '#/GeminiComputerUse'

GeminiGoogleSearch:
  type: object
  properties:
    timeRangeFilter:
      type: object
      properties:
        startTime:
          type: string
          format: date-time
        endTime:
          type: string
          format: date-time
    excludeDomains:
      type: array
      items:
        type: string
      description: List of domains to exclude from search results (max 2000)

GeminiGoogleSearchRetrieval:
  type: object
  properties:
    dynamicRetrievalConfig:
      type: object
      properties:
        mode:
          type: string
          description: The mode of the predictor for dynamic retrieval
        dynamicThreshold:
          type: number
          description: Threshold for dynamic retrieval

GeminiRetrieval:
  type: object
  properties:
    disableAttribution:
      type: boolean
      deprecated: true
      description: Deprecated. This option is no longer supported
    externalApi:
      $ref: '#/GeminiExternalAPI'
    vertexAiSearch:
      $ref: '#/GeminiVertexAISearch'
    vertexRagStore:
      $ref: '#/GeminiVertexRAGStore'

GeminiExternalAPI:
  type: object
  properties:
    endpoint:
      type: string
      description: The endpoint of the external API
    apiSpec:
      type: string
      enum: [API_SPEC_UNSPECIFIED, SIMPLE_SEARCH, ELASTIC_SEARCH]
    authConfig:
      $ref: '#/GeminiAuthConfig'
    elasticSearchParams:
      type: object
      properties:
        index:
          type: string
          description: The ElasticSearch index to use
        numHits:
          type: integer
          description: Number of hits (chunks) to request
        searchTemplate:
          type: string
          description: The ElasticSearch search template to use

GeminiVertexAISearch:
  type: object
  properties:
    datastore:
      type: string
      description: Fully-qualified Vertex AI Search data store resource ID
    engine:
      type: string
      description: Fully-qualified Vertex AI Search engine resource ID
    filter:
      type: string
      description: Filter strings to be passed to the search API
    maxResults:
      type: integer
      description: Number of search results to return (max 10, default 10)
    dataStoreSpecs:
      type: array
      items:
        type: object
        properties:
          dataStore:
            type: string
            description: Full resource name of DataStore
          filter:
            type: string
            description: Filter specification for documents in the data store

GeminiVertexRAGStore:
  type: object
  properties:
    ragCorpora:
      type: array
      items:
        type: string
      deprecated: true
      description: Deprecated. Use ragResources instead
    ragResources:
      type: array
      items:
        type: object
        properties:
          ragCorpus:
            type: string
            description: RAGCorpora resource name
          ragFileIds:
            type: array
            items:
              type: string
            description: rag_file_id. Files should be in the same rag_corpus
    ragRetrievalConfig:
      type: object
      properties:
        topK:
          type: integer
          description: The number of contexts to retrieve
        filter:
          type: object
          properties:
            metadataFilter:
              type: string
              description: String for metadata filtering
            vectorDistanceThreshold:
              type: number
              description: Only returns contexts with vector distance smaller than threshold
            vectorSimilarityThreshold:
              type: number
              description: Only returns contexts with vector similarity larger than threshold
        hybridSearch:
          type: object
          properties:
            alpha:
              type: number
              description: Weight between dense and sparse vector search (0-1). 0 = sparse only, 1 = dense only
        ranking:
          type: object
          properties:
            llmRanker:
              type: object
              properties:
                modelName:
                  type: string
            rankService:
              type: object
              properties:
                modelName:
                  type: string
    similarityTopK:
      type: integer
      description: Number of top k results to return from selected corpora
    storeContext:
      type: boolean
      description: For Gemini Multimodal Live API - memorize interactions
    vectorDistanceThreshold:
      type: number
      description: Only return results with vector distance smaller than threshold

GeminiAuthConfig:
  type: object
  properties:
    authType:
      type: string
      enum: [AUTH_TYPE_UNSPECIFIED, NO_AUTH, API_KEY_AUTH, HTTP_BASIC_AUTH, GOOGLE_SERVICE_ACCOUNT_AUTH, OAUTH, OIDC_AUTH]
    apiKeyConfig:
      type: object
      properties:
        apiKeyString:
          type: string
    googleServiceAccountConfig:
      type: object
      properties:
        serviceAccount:
          type: string
    httpBasicAuthConfig:
      type: object
      properties:
        credentialSecret:
          type: string
    oauthConfig:
      type: object
      properties:
        accessToken:
          type: string
        serviceAccount:
          type: string
    oidcConfig:
      type: object
      properties:
        idToken:
          type: string
        serviceAccount:
          type: string

GeminiEnterpriseWebSearch:
  type: object
  properties:
    excludeDomains:
      type: array
      items:
        type: string
      description: List of domains to exclude (max 2000)

GeminiGoogleMaps:
  type: object
  properties:
    authConfig:
      $ref: '#/GeminiAuthConfig'

GeminiComputerUse:
  type: object
  properties:
    environment:
      type: string
      enum: [ENVIRONMENT_UNSPECIFIED, ENVIRONMENT_BROWSER]
      description: The environment being operated

GeminiFunctionDeclaration:
  type: object
  properties:
    name:
      type: string
      description: Function name. Must start with letter/underscore, a-z, A-Z, 0-9, underscores, dots, dashes. Max 64 chars
    description:
      type: string
      description: Description and purpose of the function
    parameters:
      $ref: '#/GeminiSchema'
    parametersJsonSchema:
      type: object
      description: Alternative to parameters using JSON Schema format
    response:
      $ref: '#/GeminiSchema'
      description: Output schema for the function
    responseJsonSchema:
      type: object
      description: Alternative to response using JSON Schema format
    behavior:
      type: string
      enum: [UNSPECIFIED, BLOCKING, NON_BLOCKING]
      description: Function behavior mode. BLOCKING waits for response, NON_BLOCKING continues conversation

GeminiToolConfig:
  type: object
  properties:
    functionCallingConfig:
      type: object
      properties:
        mode:
          type: string
          enum: [MODE_UNSPECIFIED, AUTO, ANY, NONE, VALIDATED]
          description: Function calling mode
        allowedFunctionNames:
          type: array
          items:
            type: string
          description: Function names to call when mode is ANY
    retrievalConfig:
      type: object
      properties:
        latLng:
          type: object
          properties:
            latitude:
              type: number
              description: Latitude in degrees [-90.0, +90.0]
            longitude:
              type: number
              description: Longitude in degrees [-180.0, +180.0]
        languageCode:
          type: string

# Response types
GeminiGenerationResponse:
  type: object
  properties:
    candidates:
      type: array
      items:
        $ref: '#/GeminiCandidate'
    promptFeedback:
      $ref: '#/GeminiPromptFeedback'
    usageMetadata:
      $ref: '#/GeminiUsageMetadata'
    modelVersion:
      type: string
      description: The model version used to generate the response
    responseId:
      type: string
      description: Response ID for identifying each response (encoding of event_id)
    createTime:
      type: string
      format: date-time
      description: Timestamp when the request was made to the server

GeminiCandidate:
  type: object
  properties:
    content:
      $ref: '#/GeminiContent'
    finishReason:
      type: string
      enum: [FINISH_REASON_UNSPECIFIED, STOP, MAX_TOKENS, SAFETY, RECITATION, LANGUAGE, OTHER, BLOCKLIST, PROHIBITED_CONTENT, SPII, MALFORMED_FUNCTION_CALL, IMAGE_SAFETY, UNEXPECTED_TOOL_CALL]
    finishMessage:
      type: string
      description: Human-readable finish message
    tokenCount:
      type: integer
      description: Number of tokens for this candidate
    safetyRatings:
      type: array
      items:
        $ref: '#/GeminiSafetyRating'
    citationMetadata:
      type: object
      description: Source attribution of the generated content
    index:
      type: integer
      description: Index of the candidate
    groundingMetadata:
      type: object
      description: Metadata specifying sources used to ground generated content
    urlContextMetadata:
      $ref: '#/GeminiURLContextMetadata'
    avgLogprobs:
      type: number
      description: Average log probability score of the candidate
    logprobsResult:
      $ref: '#/GeminiLogprobsResult'

GeminiURLContextMetadata:
  type: object
  properties:
    urlMetadata:
      type: array
      items:
        type: object
        properties:
          retrievedUrl:
            type: string
          urlRetrievalStatus:
            type: string

GeminiLogprobsResult:
  type: object
  properties:
    chosenCandidates:
      type: array
      items:
        $ref: '#/GeminiLogprobsCandidate'
    topCandidates:
      type: array
      items:
        type: object
        properties:
          candidates:
            type: array
            items:
              $ref: '#/GeminiLogprobsCandidate'

GeminiLogprobsCandidate:
  type: object
  properties:
    token:
      type: string
      description: The candidate's token string value
    tokenId:
      type: integer
      description: The candidate's token ID value
    logProbability:
      type: number
      description: The candidate's log probability

GeminiPromptFeedback:
  type: object
  properties:
    blockReason:
      type: string
    blockReasonMessage:
      type: string
      description: Human-readable block reason message
    safetyRatings:
      type: array
      items:
        $ref: '#/GeminiSafetyRating'

GeminiSafetyRating:
  type: object
  properties:
    category:
      type: string
      description: Harm category
    probability:
      type: string
      description: Harm probability level
    probabilityScore:
      type: number
      description: Harm probability score
    severity:
      type: string
      description: Harm severity level
    severityScore:
      type: number
      description: Harm severity score
    blocked:
      type: boolean
      description: Whether content was filtered
    overwrittenThreshold:
      type: string
      description: Overwritten threshold for safety category (for Gemini 2.0 image output with minors detected)

GeminiUsageMetadata:
  type: object
  properties:
    promptTokenCount:
      type: integer
      description: Number of tokens in the prompt (includes cached content)
    candidatesTokenCount:
      type: integer
      description: Number of tokens in the response(s)
    totalTokenCount:
      type: integer
      description: Total token count for prompt, response candidates, and tool-use prompts
    cachedContentTokenCount:
      type: integer
      description: Number of tokens in the cached part of the input
    thoughtsTokenCount:
      type: integer
      description: Number of tokens in thoughts output
    toolUsePromptTokenCount:
      type: integer
      description: Number of tokens in tool-use prompts
    trafficType:
      type: string
      description: Traffic type (Pay-As-You-Go or Provisioned Throughput)
    cacheTokensDetails:
      type: array
      items:
        $ref: '#/GeminiModalityTokenCount'
      description: Modalities of the cached content in the request input
    candidatesTokensDetails:
      type: array
      items:
        $ref: '#/GeminiModalityTokenCount'
      description: Modalities returned in the response
    promptTokensDetails:
      type: array
      items:
        $ref: '#/GeminiModalityTokenCount'
      description: Modalities processed in the request input
    toolUsePromptTokensDetails:
      type: array
      items:
        $ref: '#/GeminiModalityTokenCount'
      description: Modalities processed for tool-use request inputs

GeminiModalityTokenCount:
  type: object
  properties:
    modality:
      type: string
      description: The modality (TEXT, IMAGE, AUDIO, etc.)
    tokenCount:
      type: integer

# Embedding types
GeminiEmbeddingRequest:
  type: object
  properties:
    model:
      type: string
    content:
      $ref: '#/GeminiContent'
    taskType:
      type: string
    title:
      type: string
    outputDimensionality:
      type: integer

GeminiBatchEmbeddingRequest:
  type: object
  properties:
    requests:
      type: array
      items:
        $ref: '#/GeminiEmbeddingRequest'

GeminiEmbeddingResponse:
  type: object
  properties:
    embeddings:
      type: array
      items:
        $ref: '#/GeminiEmbedding'
    metadata:
      $ref: '#/GeminiEmbedContentMetadata'

GeminiEmbedding:
  type: object
  properties:
    values:
      type: array
      items:
        type: number
    statistics:
      $ref: '#/GeminiContentEmbeddingStatistics'

GeminiEmbedContentMetadata:
  type: object
  properties:
    billableCharacterCount:
      type: integer

GeminiContentEmbeddingStatistics:
  type: object
  properties:
    tokenCount:
      type: integer

GeminiBatchEmbeddingResponse:
  type: object
  properties:
    embeddings:
      type: array
      items:
        $ref: '#/GeminiEmbedding'

# Count tokens types
GeminiCountTokensRequest:
  type: object
  properties:
    contents:
      type: array
      items:
        $ref: '#/GeminiContent'
    generateContentRequest:
      $ref: '#/GeminiGenerationRequest'

GeminiCountTokensResponse:
  type: object
  properties:
    totalTokens:
      type: integer
      description: Number of tokens that the model tokenizes the prompt into
    cachedContentTokenCount:
      type: integer
      description: Number of tokens in the cached part of the prompt
    promptTokensDetails:
      type: array
      items:
        $ref: '#/GeminiModalityTokenCount'
      description: Modalities processed in the request input
    cacheTokensDetails:
      type: array
      items:
        $ref: '#/GeminiModalityTokenCount'
      description: Modalities in the cached content