Bankwise：基于 Gemini 与 Google Cloud 的 AI 银行产品对比平台

AI-powered banking comparison and financial decision platform.

Problem

Finding financial-product information is easy.

Understanding the actual financial outcome, conditions, penalties, fees, liquidity and trade-offs is difficult.

Solution

Bankwise uses Gemini and Google Cloud to understand a user's financial goal, compare verified banking products, perform deterministic calculations, and explain the trade-offs with source transparency.

Example

I have ₹5 lakh and want to invest for 2 years. I may need the money early.

Bankwise:

1. Understands the requirement.
2. Finds relevant products.
3. Verifies banking data.
4. Calculates expected outcomes.
5. Compares conditions and liquidity.
6. Explains the trade-offs.
7. Shows the sources.

Current MVP

The MVP focuses on Fixed Deposits.

Supported:

• Product comparison
• FD calculations
• Liquidity comparison
• Premature withdrawal comparison
• Source verification
• Conflict detection
• AI explanations

Architecture

User
 ↓
Next.js
 ↓
Cloud Run / FastAPI
 ↓
Gemini + Google ADK
 ↓
Research / Calculation / Verification
 ↓
BigQuery catalogue + Firestore sessions
 ↓
Decision Engine
 ↓
Gemini Explanation

Backend

The Python FastAPI service is in backend/README.md. It reads curated product and source records from BigQuery and stores redacted decision-session state in Firestore. The catalogue starts empty until official product data and source metadata are curated.

Continuous integration

GitHub Actions runs frontend-lint, frontend-build, backend-lint, and backend-test for pull requests and pushes to main. To make these checks block merges, configure the repository's branch protection or ruleset for main and require all four status checks.