Implement Google Gemini Application

I’m kicking off a proof-of-concept around Google’s new Gemini multimodal model and need an experienced hand to get the fundamentals right. Your task is to spin up a small yet extensible codebase that:
• Authenticates through Google AI Studio
• Accepts text (and ideally image) prompts
• Returns structured JSON responses
• Stores key metadata in Postgres

I’d like the logic wrapped in a clean REST service so my in-house team can drop it straight into our existing stack. Python is preferred, but Node.js is fine if that’s where you shine. Please containerise everything—one `docker compose up` should bring the service online and fire a successful Gemini request using the API key I’ll supply.

Deliverables
• Dockerised repo on GitHub/GitLab
• Postman (or similar) collection demonstrating sample calls
• Concise setup & usage guide (Markdown)
• Light unit-test suite; code must clear PEP8/ESLint

Acceptance criteria
• Service launches via Docker without manual tweaks
• At least two demo endpoints (plain text and image-to-text) hitting Gemini successfully
• Linting and tests pass cleanly

If you’ve already played with Gemini Pro or Ultra, show me. Real-world examples will move you to the top of the list.