How aashmitmckenzie/aura-ai-powered-interview-assistant works
Project Type and Purpose
• This is a full-stack AI‐powered interview platform (“Aura 7.0”) with a React/TypeScript frontend and a FastAPI/Python backend.
• Key use cases: live audio transcription, sentiment & bias analysis, question generation, admin user management, PDF reporting.
File Structure
• Top‐level:
• Questions_Data/ – CSV files for 20+ role‐specific question banks
• backend/ – FastAPI service under app/ with routers, models, schemas, security, middleware
• frontend/ – React components, contexts, Vite build config
• requirements.txt / package.json – dependency manifests
• backend/app/routers/ contains modular endpoints (auth, admin, ai, sentiment, realtime_bias, question_selector, etc.) suggesting a micro‐modular service design.
• security_middleware.py, security_utils.py, and config.py encapsulate auth and configuration concerns.
• SQLite (app_data.db) plus CSVs form the data layer.
• frontend/src/components and contexts follow a clean React Context + component hierarchy.
Extract Architecture Insights from README
• Real‐time speech: Web Speech API in browser + local Whisper tiny model
• Hybrid AI analysis: VADER (rule‐based) + Transformers (ML‐based)
• JWT and RBAC for secure auth, audit logging middleware
• Modular FastAPI (microservice style) with SQLAlchemy ORM
• React state management via Context API, Vite optimizations
Main Components
a. Client Layer
• Browser UI (React + Web Speech API), includes Admin Panel & Candidate Interview Screen
b. API Gateway / Backend Layer
• FastAPI server—grouped by functional modules:
• Authentication & Security (auth.py, security_middleware, JWT service)
• Question Service (question_selector.py, seed_questions)
• Real‐time Analysis (realtime_bias.py, continuous_ai.py, sentiment.py)
• Reporting (reports.py)
• User/Admin Management (users.py, admin.py)
c. AI Processing Layer
• Whisper transcription engine
• VADER sentiment analyzer
• Transformers bias detector
d. Data Layer
• SQLite database (app_data.db) via SQLAlchemy ORM
• Questions_Data CSV question stores
e. External Services
• OpenAI Whisper (if using cloud API)
• Hugging Face Transformers repos (downloaded models)
Relationships & Data Flows
• UI → HTTP/WS calls → FastAPI endpoints
• Audio stream: Browser → FastAPI → Whisper → transcribed text → AI modules
• AI modules publish analysis results → FastAPI → WebSocket or HTTP responses → UI
• Question generation: UI requests → question_selector → CSV/DB → session payload
• Reporting: FastAPI generates PDF → UI download
• Auth flows: UI login/signup → auth router → JWT issuance → UI stores token → all backend calls include token
Architectural Patterns & Principles
• Modular “micro‐router” approach in FastAPI (each feature a separate router)
• API‐First design (clear RESTful endpoints, documented via Swagger)
• Separation of concerns: security, business logic, data access, AI modules
• React Context for global state, component‐based UI
• Lazy loading of AI models, caching strategies
Technologies / Frameworks
• Frontend: React 18, TypeScript, Vite, Axios, Web Speech API
• Backend: FastAPI, Python 3.8+, SQLAlchemy, Pydantic, Uvicorn
• AI/ML: OpenAI Whisper, NLTK/VADER, Hugging Face Transformers
• Database: SQLite + CSV files
• Security: JWT, RBAC, CORS, middleware audit logging