Google Gemini Integration

[kaseywright]

Implement a reusable Google Gemini client integration using the AI Studio API (google-generativeai SDK). This is foundational infrastructure for translation features, supporting configurable model selection with gemini-2.5-flash-lite as the default.

We can also use the gemini-3.1-flash-lite model if the above is not available.

---

Acceptance Criteria

Configuration & Setup

• Add google-generativeai to pyproject.toml dependencies
• Add GOOGLE_AI_API_KEY to Settings class in src/app/config.py
• Add GOOGLE_AI_MODEL setting with default "gemini-2.5-flash-lite"
• Add both vars to .env.example with empty placeholder values
• Re-run ./fai.sh build to verify dependency installation

Client Implementation

• Create src/app/core/ai_clients/google_gemini.py with:
  • GoogleGeminiClient class wrapping google.generativeai.GenerativeModel
  • Async initialization using genai.configure(api_key=...)
  • generate_content(prompt: str) -> str async method
  • Basic error handling: catch SDK exceptions and raise HTTPException(status.HTTP_502_BAD_GATEWAY) with detail message
• Add singleton/dependency factory in src/app/dependencies.py:

  async def get_google_gemini_client() -> GoogleGeminiClient:
      # Returns cached singleton instance

Tests

• Create src/tests/test_google_gemini.py with:
  • Test client initializes with valid API key from config
  • Test generate_content with mocked SDK response (patch genai.GenerativeModel.generate_content_async)
  • Test error handling: SDK exception → HTTPException with 502 status

Contract

# src/app/core/ai_clients/google_gemini.py
class GoogleGeminiClient:
    async def generate_content(self, prompt: str) -> str: ...

---

Out of Scope (Future Tickets)

• Actual translation endpoints/routers
• Prompt templates for translation
• Response streaming
• Rate limiting / retry logic
• Structured logging integration (separate ticket in progress)

---

Testing Commands

./fai.sh test src/tests/test_google_gemini.py -v
./fai.sh lint
./fai.sh typecheck

---

References

• SDK docs: https://ai.google.dev/gemini-api/docs
• Model: gemini-2.5-flash-lite (AI Studio tier)
• Project patterns: Follow existing OpenAI/Anthropic config patterns in src/app/config.py