Google Gemini Api

What it does

This skill provides guidance and assistance for working with the Google Gemini API, specifically focusing on current best practices and migration from deprecated SDKs. It enables AI agents to generate text, process multimodal inputs like images, video, audio, and PDFs, utilize function calling, manage chat interactions, and leverage advanced features such as context caching, code execution (with a Python interpreter), and grounding responses with real-time Google Search results. The skill supports both the Node.js SDK (@google/genai) and fetch approaches for interacting with the API.

When to use it

• You're migrating from the deprecated @google/generative-ai SDK and need a migration guide.
• You want to build applications that utilize Google Gemini’s multimodal capabilities (images, video, audio, PDFs).
• You require code execution within your AI agent using a built-in Python interpreter.
• You need to ground responses with up-to-date information from Google Search and include citations.
• You want to optimize costs by implementing context caching for Gemini API calls.

Key capabilities

• Text Generation (basic & streaming)
• Multimodal Input Processing (images, video, audio, PDFs)
• Function Calling (basic & parallel execution)
• System Instructions and Multi-turn Chat Management
• Thinking Mode Configuration
• Generation Parameter Control (temperature, top-p, top-k, stop sequences)
• Node.js SDK (@google/genai) and Fetch API Support
• Context Caching for Cost Optimization
• Code Execution with a Python Interpreter
• Grounding Responses with Google Search & Citations

Example prompts

• "Show me how to migrate my code from @google/generative-ai to the current SDK."
• "Generate an image description using Gemini and then summarize it."
• "Execute this Python code snippet: print('Hello, world!')"
• "Answer the question 'What is the capital of France?' and cite your source from Google Search."

Tips & gotchas

• SDK Version: Ensure you are using @google/[email protected]. The deprecated @google/generative-ai SDK will be sunset on November 30, 2025.
• Environment Variables: You'll need to set the GEMINI_API_KEY environment variable for authentication.
• Gemini 3 Preview: Be aware that Gemini 3 models (e.g., gemini-3-pro-preview) are in a preview release and may have limited documentation.