Google Gemini Integration

Google Gemini steps provide direct integration with Google's advanced AI models for text generation, image generation, file processing, and document analysis. These steps are optimized for long-context understanding and multi-modal processing.

Available Steps (5)

gemini_prompt

Generate text responses using Gemini models with customizable prompts.

Activity Name: gemini_prompt

Use Cases: Content generation, Q&A, creative writing, code generation

gemini_text_reader

Process and analyze text content with customizable analysis prompts.

Activity Name: gemini_text_reader

Use Cases: Text summarization, content analysis, document processing

gemini_json_reader

Read and analyze JSON data or files with intelligent parsing.

Activity Name: gemini_json_reader

Use Cases: Data analysis, JSON summarization, structured data processing

gemini_file_reader

Upload and analyze files directly using Gemini's file processing capabilities.

Activity Name: gemini_file_reader

Use Cases: Document analysis, file content extraction, multi-format processing

gemini_image_generator

Generate images from prompts, optionally conditioned on input images.

Activity Name: gemini_image_generator

Use Cases: Image generation, image editing, visual content creation

Configuration

Authentication

All Gemini steps use unified API key management:

{
  "api_key_secret": "GEMINI_API_KEY"
}

Authentication Patterns:

• Direct: "api_key": "your-api-key"
• Secrets Manager: "api_key_secret": "GEMINI_API_KEY" (recommended)
• Environment: Falls back to GEMINI_API_KEY environment variable

Model Selection

All steps support model configuration:

{
  "model": "gemini-2.0-flash-001"
}

Available Models:

• gemini-2.0-flash-001 (default) - Latest fast model
• gemini-1.5-pro - High-quality analysis model
• gemini-1.5-flash - Balanced speed and quality

Step Documentation

gemini_prompt

Generate text using Gemini models with direct prompts.

Configuration

{
  "activity": "gemini_prompt",
  "model": "gemini-2.0-flash-001",
  "prompt": "Your prompt here"
}

Parameters

• prompt (string, required) - The text prompt for generation
• model (string, default: gemini-2.0-flash-001) - Gemini model to use
• api_key_secret (string, default: GEMINI_API_KEY) - Secret containing API key

Example

{
  "name": "generate_content",
  "activity": "gemini_prompt",
  "config": {
    "prompt": "Write a professional email about project updates",
    "model": "gemini-2.0-flash-001"
  }
}

gemini_text_reader

Analyze text content with customizable analysis prompts.

Configuration

{
  "activity": "gemini_text_reader",
  "text_path": "previous_step.outputs.text",
  "prompt": "Summarize this text in 100 words"
}

Parameters

• text_path (string, required) - Path to text content from previous steps
• prompt (string, default: summary prompt) - Analysis instruction
• model (string, default: gemini-2.0-flash-001) - Gemini model to use

Example

{
  "name": "analyze_content",
  "activity": "gemini_text_reader",
  "config": {
    "text_path": "fetch_article.outputs.content",
    "prompt": "Extract the main themes and key points from this article"
  }
}

gemini_json_reader

Process JSON data with intelligent analysis and summarization.

Configuration

{
  "activity": "gemini_json_reader",
  "json_path": "previous_step.outputs.data",
  "prompt": "Analyze this JSON structure"
}

Parameters

• json_path (string, required) - Path to JSON data or file path
• prompt (string, default: summary prompt) - Analysis instruction
• model (string, default: gemini-2.0-flash-001) - Gemini model to use

Features

• Automatic JSON validation and parsing
• File path resolution for JSON files
• Intelligent structure analysis

Example

{
  "name": "analyze_data",
  "activity": "gemini_json_reader",
  "config": {
    "json_path": "api_response.outputs.json_data",
    "prompt": "Identify trends and anomalies in this dataset"
  }
}

gemini_file_reader

Upload and analyze files directly using Gemini's native file processing.

Configuration

{
  "activity": "gemini_file_reader",
  "asset_path": "previous_step.outputs.file_path",
  "prompt": "Analyze this document"
}

Parameters

• asset_path (string, required) - Path to file for upload and analysis
• prompt (string, default: generic analysis) - Analysis instruction
• model (string, default: gemini-2.0-flash-001) - Gemini model to use

Supported Formats

• Documents: PDF, DOCX, TXT, MD
• Images: PNG, JPG, GIF, WebP
• Data: CSV, JSON, XML
• Code: Most programming languages

Example

{
  "name": "process_document",
  "activity": "gemini_file_reader",
  "config": {
    "asset_path": "upload_file.outputs.file_path",
    "prompt": "Extract key information and create a summary report"
  }
}

gemini_image_generator

Generate images from prompts using Gemini's native image generation capabilities.

Configuration

{
  "activity": "gemini_image_generator",
  "model": "gemini-2.5-flash-image-preview",
  "prompt": "A serene mountain landscape at sunset"
}

Parameters

• prompt (string, required) - Text description of desired image
• model (string, default: gemini-2.5-flash-image-preview) - Gemini image model
• image_path (string, optional) - Input image path for image-to-image generation

Output

• images (array) - Generated images with path, extension, and content_type
• text (string, optional) - Any text response from the model
• response_length (int, optional) - Length of text response

Example: Text-to-Image

{
  "name": "generate_illustration",
  "activity": "gemini_image_generator",
  "config": {
    "prompt": "A minimalist logo design for a tech startup, clean lines, blue and white color scheme",
    "model": "gemini-2.5-flash-image-preview"
  }
}

Example: Image-to-Image

{
  "name": "edit_image",
  "activity": "gemini_image_generator",
  "config": {
    "prompt": "Transform this photo into a watercolor painting style",
    "image_path": "upload_step.outputs.image_path",
    "model": "gemini-2.5-flash-image-preview"
  }
}

Output Structure

{
  "outputs": {
    "images": [
      {
        "path": "collection/flow/0001/gemini_image_generator_0.png",
        "extension": "png",
        "content_type": "image/png"
      }
    ],
    "text": "I've created a minimalist logo...",
    "response_length": 45
  }
}

Advanced Usage

Chaining Gemini Steps

{
  "steps": [
    {
      "name": "read_document",
      "activity": "gemini_file_reader",
      "config": {
        "asset_path": "document.pdf",
        "prompt": "Extract main topics from this document"
      }
    },
    {
      "name": "expand_content",
      "activity": "gemini_prompt",
      "config": {
        "prompt": "Based on these topics: {{read_document.outputs.text}}, write detailed explanations"
      }
    }
  ]
}

Multi-Format Processing

{
  "steps": [
    {
      "name": "process_json",
      "activity": "gemini_json_reader",
      "config": {
        "json_path": "data.json",
        "prompt": "Create a data quality report"
      }
    },
    {
      "name": "process_image",
      "activity": "gemini_file_reader",
      "config": {
        "asset_path": "chart.png",
        "prompt": "Describe the trends shown in this chart"
      }
    }
  ]
}

Error Handling

Common Issues

• Authentication Error: Verify GEMINI_API_KEY configuration
• File Upload Error: Check file format support and size limits
• Path Resolution Error: Ensure referenced paths exist in trajectory
• Model Not Found: Verify model name is supported

Best Practices

• Use appropriate models for task complexity (Flash for speed, Pro for quality)
• Implement retry logic for rate limiting scenarios
• Cache file uploads when processing multiple times
• Use specific prompts for better results

Performance Tips

Model Selection

• gemini-2.0-flash-001: Fast responses, good for simple tasks
• gemini-1.5-pro: High quality analysis, longer processing time
• gemini-1.5-flash: Balanced performance for most use cases

Rate Limiting

• Gemini has generous rate limits compared to other providers
• Automatic retry with exponential backoff handled internally
• Consider batch processing for multiple documents

Related Steps

• LiteLLM Chat - Alternative chat interface with multi-provider support
• Replicate Text Stream - Streaming text generation
• File Processing Tools - Complement Gemini's file analysis capabilities

Integration Examples

View complete workflow examples in the Flow Library:

• Document analysis workflows
• Content generation pipelines
• Multi-modal processing examples