Gemini API Skill

Build AI applications with Google's Gemini models and tools.

Quick Start

Installation

# Python
pip install google-genai

# JavaScript/Node.js
npm install @google/genai

# Go
go get google.golang.org/genai

Environment Setup

export GEMINI_API_KEY="your-api-key"

Basic Usage

Python:

from google import genai

client = genai.Client()
response = client.models.generate_content(
    model="gemini-2.5-flash",
    contents="Your prompt here"
)
print(response.text)

JavaScript:

import { GoogleGenAI } from "@google/genai";

const ai = new GoogleGenAI({});
const response = await ai.models.generateContent({
    model: "gemini-2.5-flash",
    contents: "Your prompt here"
});
console.log(response.text);

REST:

curl "https://generativelanguage.googleapis.com/v1beta/models/gemini-2.5-flash:generateContent" \
  -H "x-goog-api-key: $GEMINI_API_KEY" \
  -H 'Content-Type: application/json' \
  -d '{"contents": [{"parts": [{"text": "Your prompt here"}]}]}'

Model Selection

Reference Documentation Index

Getting Started

Models & Pricing

Core Capabilities

Advanced Features

Tools

Live API (Real-time Streaming)

Guides

Troubleshooting & Migration

API Reference (Technical Endpoints)

Note: These are technical endpoint specifications with schemas and parameter details. For usage guides and code examples, see the guide files above.

Large Files - Search Patterns

For large reference files (>30K), use grep to find specific sections:

image-generation-gemini.md (174K):

grep -n "## " references/image-generation-gemini.md  # List sections
grep -n "edit" references/image-generation-gemini.md  # Find editing info
grep -n "style" references/image-generation-gemini.md  # Find style transfer

api-generate-content.md (166K):

grep -n "## " references/api-generate-content.md  # List sections
grep -n "GenerationConfig" references/api-generate-content.md  # Config options
grep -n "SafetySetting" references/api-generate-content.md  # Safety types

api-caching-reference.md (89K):

grep -n "## " references/api-caching-reference.md  # List sections
grep -n "CachedContent" references/api-caching-reference.md  # Cache types

veo.md (69K):

grep -n "## " references/veo.md  # List sections
grep -n "audio" references/veo.md  # Find audio generation info

models-overview.md (67K):

grep -n "gemini-3" references/models-overview.md
grep -n "context" references/models-overview.md

function-calling.md (54K):

grep -n "## " references/function-calling.md
grep -n "parallel" references/function-calling.md  # Parallel function calls

Common Patterns

Multimodal Input (Image + Text)

from google import genai
from google.genai import types

client = genai.Client()
response = client.models.generate_content(
    model="gemini-2.5-flash",
    contents=[
        types.Part.from_image(image_path),
        types.Part.from_text("Describe this image")
    ]
)

Function Calling

tools = [
    types.Tool(function_declarations=[{
        "name": "get_weather",
        "description": "Get weather for a location",
        "parameters": {
            "type": "object",
            "properties": {"location": {"type": "string"}},
            "required": ["location"]
        }
    }])
]

response = client.models.generate_content(
    model="gemini-2.5-flash",
    contents="What's the weather in Paris?",
    config=types.GenerateContentConfig(tools=tools)
)

Google Search Grounding

response = client.models.generate_content(
    model="gemini-2.5-flash",
    contents="What are the latest AI developments?",
    config=types.GenerateContentConfig(
        tools=[types.Tool(google_search=types.GoogleSearch())]
    )
)

Thinking Mode

response = client.models.generate_content(
    model="gemini-2.5-pro",
    contents="Solve this complex problem...",
    config=types.GenerateContentConfig(
        thinking_config=types.ThinkingConfig(thinking_budget_tokens=10000)
    )
)

Streaming

for chunk in client.models.generate_content_stream(
    model="gemini-2.5-flash",
    contents="Write a story"
):
    print(chunk.text, end="")

Key Concepts

Tool Execution Flow

Built-in tools (Google Search, Code Execution): Executed by Google

1. Send prompt with tool config → Model executes tool → Response with grounded results

Custom tools (Function Calling): You execute

1. Send prompt with function declarations → Model returns function call JSON
2. You execute function, send result back → Model generates final response

Thought Signatures (Important)

• If using official SDKs with chat feature: Thought signatures are handled automatically. No action needed.
• If manually managing conversation history: Read thought-signatures.md for Gemini 3 Pro function calling requirements.

API Endpoints

/v1beta/models/{model}:generateContent

/v1beta/models/{model}:streamGenerateContent

/v1beta/models/{model}:embedContent

/v1beta/models/{model}:countTokens

Base URL:

https://generativelanguage.googleapis.com