srirammurugesan/fastapiapp

Purpose and entry points

TalentSpark is a full-stack AI job portal: authenticated users manage companies/jobs, upload resumes, perform semantic job discovery and matching, analyze resumes, and use a career chatbot. Backend startup is backend/app/main.py (uvicorn app.main:app); frontend startup is frontend/talentspark/src/main.tsx, which mounts frontend/talentspark/src/App.tsx.

Client application

The React/TypeScript Vite client provides login/registration, job matching, resume analysis, and chat views. Pages call the FastAPI REST API through domain-specific service clients, keeping UI concerns separate from API transport and domain types.

• frontend/talentspark/src/App.tsx

• frontend/talentspark/src/Services/api.ts

• frontend/talentspark/src/pages/JobMatch.tsx

API composition and request boundary

FastAPI’s application entry point registers routers and cross-cutting logging/middleware. Router modules are the HTTP boundary: they validate requests with Pydantic schemas, apply authorization dependencies where required, invoke persistence/AI/storage services, and return API responses. Swagger is exposed by FastAPI.

• backend/app/main.py

• backend/routers/auth.py

• backend/routers/rag.py

Identity and authorization

Authentication registers users with hashed passwords, logs them in to receive JWTs, and protects selected routes through OAuth2/JWT dependencies and RBAC (admin, hr, student). Authorization is primarily enforced at backend route boundaries rather than trusted to the frontend.

• backend/routers/auth.py

• backend/utils/oauth2.py

• backend/utils/token.py

Relational job-domain persistence

SQLAlchemy async models represent users, companies, jobs, and resume metadata. API CRUD routers operate against PostgreSQL—documented as Supabase-hosted locally—with Alembic managing schema evolution. Jobs and companies are the source data for vector indexing; resume records link uploaded objects to users.

• backend/database.py

• backend/models/job.py

• backend/alembic/versions/250d9ca9979c_initial_migrations.py

Resume object-storage flow

A logged-in user uploads a resume through the S3 router. The storage service sends the file to AWS S3 when configured, with a documented local fallback, then resume-related persistence can retain filename, object URL, and owning user. This isolates cloud-storage mechanics from request handlers.

• backend/routers/s3_demo.py

• backend/services/s3_service.py

• backend/models/resume.py

AI, retrieval, and conversational flows

RAG endpoints index PostgreSQL job content into Qdrant using embeddings, perform semantic retrieval, and use retrieved job context for answers, resume analysis, and job-match scoring. Qdrant is an external vector-store boundary. The chat router delegates session-memory career conversations to LangChain and Groq Llama; resume-oriented generation uses Gemini. These API-key-backed services are isolated from routers.

• backend/services/rag_service.py

• backend/services/qdrant_service.py

• backend/services/langchai_service.py

Runtime and deployment

The system runs as separate backend and frontend containers coordinated by Compose. Backend runtime is Python/FastAPI with async PostgreSQL access; frontend is Node-built React/Vite. External infrastructure comprises PostgreSQL/Supabase, Qdrant Cloud, AWS S3, Groq, and Google Gemini; deployment guidance targets AWS Elastic Beanstalk.

• docker-compose.yml

• backend/Dockerfile

• frontend/talentspark/Dockerfile