🐛 BugBuddy

An AI-powered developer tool that turns a raw stack trace into a plain-English explanation and a concrete suggested fix. Paste an error, pick the language, and BugBuddy returns a diagnosis backed by Google's Gemini 2.5 Flash — with a MySQL result cache in front so the same error never pays for the AI twice.

• Frontend: React 19 + TanStack Start on Vercel
• Backend: Spring Boot 3.5 / Java 21 on Railway, MySQL

---

What it does

1. You paste a Java, Python, or JavaScript stack trace and submit.
2. The backend checks a MySQL cache for an exact match on that error text.
   • Cache hit → the stored analysis is returned in under ~50ms, with no AI call.
   • Cache miss → the error is sent to Gemini 2.5 Flash, the response is parsed, persisted, and returned.
3. Every triage is saved to a searchable history.

Why it's built this way

• Cache ahead of the model. Gemini calls cost money and latency. An exact-match lookup on the error text means a repeated error is answered from the database, not the API — most real error streams are highly repetitive.
• Servlet-level CORS, not @CrossOrigin. The cross-origin config lives in a CorsFilter bean (WebConfig) that runs early in the filter chain, ahead of Spring MVC. A controller-level @CrossOrigin annotation sat too late in the chain to resolve the preflight conflict this app hit.
• Two-layer payload guard. A 5,000-character limit is enforced independently on both the client and the server, so the server never trusts the client to have checked.
• Cost protection on the one endpoint that spends money. POST /api/bugs/analyze is the only endpoint that can reach Gemini, so it is the only one that is rate limited (per-IP) and can be gated behind an optional API key.

Tech stack

Layer	Choice
Backend	Java 21, Spring Boot 3.5 (Web, Data JPA, Validation)
Database	MySQL (H2 in-memory for tests)
AI	Google Gemini 2.5 Flash via REST
Frontend	React 19, TanStack Start & Router, TanStack Query, TypeScript
Styling	Tailwind CSS v4, Radix UI
Build	Maven (backend), Vite (frontend)
Hosting	Railway (backend), Vercel (frontend)

API

Base path: /api

Method	Path	Description
POST	/bugs/analyze	Triage an error. Rate-limited; optionally API-key gated. 201 new, 200 cache hit, 400 invalid, 429 rate limited, 401 missing key
GET	/bugs	Paginated triage history
GET	/bugs/search?q=	Full-text search over past errors
GET	/bugs/{id}	Fetch one triage
DELETE	/bugs/{id}	Delete a triage

Running locally

Backend (port 8080)

Requires Java 21, Maven, and a MySQL instance. Configuration is entirely environment-driven — no credential is ever committed.

export MYSQLHOST=localhost MYSQLPORT=3306 MYSQLDATABASE=bugbuddy
export MYSQLUSER=root MYSQLPASSWORD=yourpassword
export GEMINI_API_KEY=your_gemini_key
mvn spring-boot:run

Variable	Purpose	Default
MYSQLHOST / MYSQLPORT / MYSQLDATABASE	Database connection	localhost / 3306 / bugbuddy
MYSQLUSER / MYSQLPASSWORD	Database credentials	root / password
GEMINI_API_KEY	Google Gemini API key (required for live analysis)	—
BUGBUDDY_API_KEY	If set, POST /bugs/analyze requires X-API-Key	disabled
bugbuddy.rate-limit.requests-per-minute	Per-IP limit on analyze	10
PORT	Server port	8080

Frontend (port 5173)

cd frontend
npm install
npm run dev

Point the frontend at a backend with VITE_API_BASE_URL (defaults to the deployed Railway URL). The production build targets Vercel's Build Output API:

npm run build      # emits .vercel/output

Tests

mvn test

17 tests across a @WebMvcTest controller slice and Mockito service unit tests, covering the cache hit/miss paths, the server-side payload-size rejection, request validation, and 404 handling.

License

MIT