Method: googleapis.aiplatform.v1.projects.locations.endpoints.generateContent

Generate content with multimodal inputs.

Arguments

Parameters
model

string

Required. The name of the publisher model requested to serve the prediction. Format: projects/{project}/locations/{location}/publishers/*/models/*
region

string

Required. Region of the HTTP endpoint. For example, if region is set to us-central1, the endpoint https://us-central1-integrations.googleapis.com will be used. See service endpoints.
body

object (GoogleCloudAiplatformV1GenerateContentRequest)

Required.

Raised exceptions

Exceptions
ConnectionError In case of a network problem (such as DNS failure or refused connection).
HttpError If the response status is >= 400 (excluding 429 and 503).
TimeoutError If a long-running operation takes longer to finish than the specified timeout limit.
TypeError If an operation or function receives an argument of the wrong type.
ValueError If an operation or function receives an argument of the right type but an inappropriate value. For example, a negative timeout.

Response

If successful, the response contains an instance of GoogleCloudAiplatformV1GenerateContentResponse.

Subworkflow snippet

Some fields might be optional or required. To identify required fields, refer to the API documentation.

Gemini Pro

# Translate `Good morning` from English to French
# Vertex AI API must be enabled
# Workflow service account must be granted
# `Vertex AI User` role
- call_gemini:
    call: googleapis.aiplatform.v1.projects.locations.endpoints.generateContent
    args:
      model: "projects/{PROJECT_ID}/locations/us-central1/publishers/google/models/gemini-pro"
      region: "us-central1"
      body:
        contents:
          role: user
          parts:
            - text: "Translate good morning from English to French"
        safety_settings:  # optional
          category: HARM_CATEGORY_DANGEROUS_CONTENT
          threshold: BLOCK_ONLY_HIGH
        generation_config:  # optional
          temperature: 0.2
          maxOutputTokens: 2000
          topK: 10
          topP: 0.9
    result: gemini_response
- returnStep:
    return: ${gemini_response.candidates[0].content.parts[0].text}

Gemini Pro Vision

# Describe an image
- call_gemini:
    call: googleapis.aiplatform.v1.projects.locations.endpoints.generateContent
    args:
      model: "projects/PROJECT_ID/locations/us-central1/publishers/google/models/gemini-1.0-pro-vision"
      region: "us-central1"
      body:
        contents:
          role: user
          parts:
            - fileData:
                mimeType: "image/jpeg"
                # Only Cloud Storage URIs are supported
                fileUri: IMAGE_URI
            - text: "Describe this picture"
    result: gemini_response
- returnStep:
    return: ${gemini_response.candidates[0].content.parts[0].text}