Modern Gemini Integration via OpenAI-Compatible Endpoint

Stop writing custom wrapper logic every time you want to switch or add an LLM provider. This skill generates a production-ready Python integration layer for Google Gemini, built entirely around a shared AIProvider abstraction. By leveraging Gemini's OpenAI-compatible endpoint, it ensures your backend architecture remains clean, decoupled, and consistent with other providers like Claude or Qwen.

Why use this skill

• Drop-In Compatibility: Uses the exact same interface as Nex Claude and Qwen integrations, allowing for one-line provider swaps.
• Optimized Model Split: Automatically configures a dual-model factory—routing high-quality generation to Gemini Pro and high-volume scoring to Gemini Flash.
• Production-Grade Resilience: Includes a shared error taxonomy with built-in retry policies for timeouts and transient API errors.
• Full Observability: Native token accounting is baked into every call, returning consistent result objects for logging and billing.

Supported Tools & Frameworks

• Python: 3.8+ backend environments.
• OpenAI Python Client: Uses the familiar SDK idiom for Google's endpoints.
• Environment Management: Pre-configured for python-dotenv and secure API key handling.

The Output

You receive a structured provider package including base.py (the interface), gemini_provider.py (the implementation), and a factory.py for easy dependency injection.