google-gemini/gemini-skills

Overview

This skill helps developers build applications with Gemini models and the Gemini API, covering multimodal capabilities, function calling, structured outputs, and SDK usage. It summarizes current model choices, SDK packages for major languages, and where to find authoritative API specs and docs. Use it to pick models, implement function-calling/structured outputs, and integrate SDKs correctly.

How this skill works

The skill inspects available Gemini models and maps common use cases (text, image, audio, video, embeddings) to the recommended model families. It outlines SDK packages and quick-start code snippets for Python, JavaScript/TypeScript, Go, and Java. It points to the canonical REST discovery spec and documentation index so implementations use the latest API definitions and features.

When to use it

• Building chat, completion, summarization, or code generation features
• Processing or understanding multimodal content (images, audio, video, documents)
• Implementing function calling or enforcing structured JSON outputs from the model
• Selecting the right Gemini model for latency, context window, or capability needs
• Migrating from legacy SDKs or integrating new SDKs in Python/JS/Go/Java
• Referencing the source-of-truth API discovery spec when debugging or implementing endpoints

Best practices

• Prefer current Gemini models (gemini-3 variants) and avoid legacy/deprecated model families
• Use official SDKs: google-genai (Python), @google/genai (JS/TS), google.golang.org/genai (Go), com.google.genai:google-genai (Java)
• Consult the v1beta REST discovery spec as the default source of truth; use v1 only if pinned to that version
• Design clear schemas for structured outputs and validate model responses against them
• Leverage function calling for deterministic actions and sandboxed code execution when available
• Cache large contexts or use explicit context management to control cost and latency

Example use cases

• Multimodal assistant that accepts images and text, extracts entities, and calls backend functions
• Automated summarization pipeline for documents and meeting audio using embeddings and text generation
• Image editing pipeline using image-capable Gemini models and structured output for edit instructions
• Developer tooling that generates and executes small Python snippets in a sandbox for data analysis
• Semantic search: embed content with Gemini embeddings and perform similarity queries