Google Gemini API Guide

The /gemini-pc command enables you to access Google Gemini AI models through PinkConnect Proxy. Perfect for:

• Text generation and creative writing
• Image analysis and description
• Multi-turn conversations
• Code generation and analysis
• Function calling capabilities
• Multimodal AI interactions

​

Basic Usage

Use the command to interact with Gemini models:

/gemini-pc generate a short story about space exploration

/gemini-pc analyze this image and describe what you see

/gemini-pc help me write a Python function for data processing

​

Key Features

​

Text Generation

• Creative writing
• Content creation
• Code generation
• Technical documentation
• Conversational AI

​

Multimodal Capabilities

• Image understanding
• Video analysis
• Audio processing
• Visual reasoning
• Cross-modal tasks

​

Advanced Features

• Function calling
• Tool integration
• JSON mode support
• Multi-turn conversations
• Structured outputs

​

Example Commands

​

Creative Writing

/gemini-pc write a poem about artificial intelligence

​

Code Generation

/gemini-pc create a Python class for managing user accounts

​

Image Analysis

/gemini-pc describe the contents of this uploaded image

​

Conversation

/gemini-pc start a conversation about machine learning concepts

​

Function Calling

/gemini-pc get weather information for San Francisco using available tools

​

Available Models

​

Latest Generation

• gemini-2.0-flash: Next-generation features, speed, and thinking
• gemini-2.0-flash-lite: Cost-efficient with low latency
• gemini-2.5-flash: Fast and versatile for general tasks
• gemini-2.5-pro: Advanced reasoning and coding capabilities

​

Established Models

• gemini-1.5-pro: High intelligence for complex tasks
• gemini-1.5-flash: Fast and versatile performance

​

Text Generation

​

Basic Request

{
  "contents": [
    {
      "parts": [
        {
          "text": "Write a short story about a robot learning to paint"
        }
      ]
    }
  ],
  "generationConfig": {
    "temperature": 0.7,
    "topK": 40,
    "topP": 0.95,
    "maxOutputTokens": 1024
  }
}

​

Configuration Options

• temperature: Creativity level (0.0-1.0)
• topK: Token selection diversity
• topP: Nucleus sampling threshold
• maxOutputTokens: Response length limit

​

Image Analysis

​

Image Input

{
  "contents": [
    {
      "parts": [
        {
          "text": "Describe what you see in this image"
        },
        {
          "inline_data": {
            "mime_type": "image/jpeg",
            "data": "base64_encoded_image_data"
          }
        }
      ]
    }
  ]
}

​

Supported Formats

• JPEG images
• PNG images
• WebP images
• Base64 encoding
• Inline data format

​

Multi-turn Conversations

​

Chat Format

{
  "contents": [
    {
      "role": "user",
      "parts": [{"text": "Hello! Can you help me with coding?"}]
    },
    {
      "role": "model",
      "parts": [{"text": "Hello! I'd be happy to help you with coding..."}]
    },
    {
      "role": "user",
      "parts": [{"text": "I need help with Python functions"}]
    }
  ]
}

​

Conversation Flow

• Maintain context across turns
• Use role-based messaging
• Build on previous responses
• Handle conversation history

​

Function Calling

​

Function Declaration

{
  "tools": [
    {
      "function_declarations": [
        {
          "name": "get_weather",
          "description": "Get the current weather for a location",
          "parameters": {
            "type": "object",
            "properties": {
              "location": {
                "type": "string",
                "description": "The city and state, e.g. San Francisco, CA"
              }
            },
            "required": ["location"]
          }
        }
      ]
    }
  ]
}

​

Tool Integration

• Define custom functions
• Structured parameter schemas
• Automatic function calling
• Result integration

​

Response Structure

​

Standard Response

{
  "candidates": [
    {
      "content": {
        "parts": [
          {
            "text": "Generated response text..."
          }
        ],
        "role": "model"
      },
      "finishReason": "STOP",
      "safetyRatings": [...]
    }
  ],
  "usageMetadata": {
    "promptTokenCount": 12,
    "candidatesTokenCount": 150,
    "totalTokenCount": 162
  }
}

​

Response Elements

• candidates: Generated responses
• finishReason: Completion status
• safetyRatings: Content safety scores
• usageMetadata: Token usage information

​

Best Practices

1. Model Selection

   • Choose appropriate model for task
   • Consider speed vs. capability trade-offs
   • Use flash models for quick responses
   • Use pro models for complex reasoning

2. Prompt Engineering

   • Be specific and clear
   • Provide context and examples
   • Use structured formatting
   • Test different approaches

3. Configuration Tuning

   • Adjust temperature for creativity
   • Set appropriate token limits
   • Use sampling parameters wisely
   • Monitor response quality

4. Multimodal Usage

   • Combine text and images effectively
   • Use appropriate mime types
   • Optimize image sizes
   • Provide clear instructions

​

Common Use Cases

​

Content Creation

/gemini-pc create marketing copy for a new product launch

​

Code Analysis

/gemini-pc review this code and suggest improvements

​

Data Processing

/gemini-pc analyze this dataset and provide insights

​

Creative Tasks

/gemini-pc generate ideas for a mobile app interface

​

Error Handling

​

Common Issues

• Invalid API keys
• Model not found
• Rate limiting
• Content safety blocks

​

Response Codes

• 200: Success
• 400: Bad request
• 401: Unauthorized
• 429: Rate limited
• 500: Server error

​

Safety Features

​

Content Filtering

• Automatic safety checks
• Harmful content detection
• Bias mitigation
• Responsible AI practices

​

Safety Categories

• Sexually explicit content
• Hate speech
• Harassment
• Dangerous content

​

Performance Optimization

​

Speed Considerations

• Use flash models for quick responses
• Optimize prompt length
• Batch similar requests
• Cache common responses

​

Cost Management

• Monitor token usage
• Use appropriate models
• Optimize request frequency
• Track usage metrics

​

Tips

• Choose the right model for your specific task requirements
• Use multimodal capabilities by combining text with images
• Leverage function calling for structured interactions
• Adjust temperature settings based on desired creativity level
• Monitor token usage to optimize costs
• Test different prompt strategies for best results