google-gemini-api

Use the @google/genai SDK (v1.27+) to create a Gemini client and call generateContent with text or multimodal inputs. The SDK handles authentication, streaming, and model routing across Gemini 2.5/3 series models like gemini-2.5-flash and gemini-2.5-pro for production text generation and processing.

Yes, function calling in @google/genai supports parallel tool execution and real-time grounding with Google Search and code execution in Phase 2. Define tool schemas, pass them to generateContent, and handle tool responses to build dynamic chat workflows and autonomous agents.

Yes, the @google/genai SDK supports both Node.js and fetch-based implementations, enabling Gemini API deployment on edge platforms like Cloudflare Workers. Use fetch endpoints for edge compatibility while maintaining streaming, multimodal, and function-calling capabilities.

The current @google/genai (v1.27+) is the supported SDK for Gemini API access. The older @google/generative-ai is deprecated; migrate to @google/genai to access Phase 2 features like function calling, grounding, and context caching with explicit model version suffixes.

Context caching in @google/genai requires explicit model version suffixes and cache configuration in generateContent requests. Cache long contexts to reduce latency and cost for repeated queries, enabling efficient long-context reasoning and Phase 2 advanced reasoning modes.

Explicit model version suffixes (e.g., gemini-2.5-flash) are required by @google/genai for caches, resources, and consistent behavior across SDK versions. Suffixes ensure correct model selection, enable versioning, and allow proper allocation of advanced features like thinking mode and function calling.