How sharadgup/ai-research-assistant-system works
Main System Components
a. Client (Web Browser)
• Static assets: HTML templates (Jinja2), CSS, JavaScript
• Pages: Landing, Login/Register, Dashboard, Domain‑agent interfaces (Education, Healthcare, Construction, News, PDF Analyzer, Data Analyzer, Data Cleaner, Voice Agent)
• Socket.IO client for real‑time chat and voice events
b. Backend Application
• Entry point: app.py (Flask app, Flask‑SocketIO, Eventlet)
• Blueprints / Routes:
• Auth routes (auth_routes.py)
• Chat & PDF chat routes (chat_routes.py, pdf_routes.py)
• Report routes (report_routes.py)
• Socket handlers: voice_handlers.py, pdf_chat_handlers.py
• Utilities: pdf_utils.py, chart_parser.py, voice_utils.py
• Models: user_model.py, report_model.py
• Configuration loader (config.py)
c. Data & File Storage
• MongoDB (user sessions, metadata, transcripts)
• File system: uploads/ (PDFs), uploads/analysis_data/ (CSV exports)
d. External Services
• Google Generative AI (Gemini) for text semantics, chat, TTS
• Google OAuth 2.0 (via Flask‑Dance)
Component Relationships & Data Flows
• Web Browser → HTTP/HTTPS → Flask Routes (Auth, Chat, PDF, Report)
• Web Browser ↔ WebSocket → Socket Handlers (real‑time chat & voice)
• Flask Application → MongoDB → CRUD for users, reports, transcripts
• Flask Utilities → File System → PDF ingestion & storage
• Flask Utilities → Google Gemini API → AI processing (analysis, chat responses, TTS audio)
• Chart Parser Utility → Generates chart data → Frontend charts via JSON over HTTP
Architectural Patterns & Principles
• Layered architecture:
• Presentation (templates + static JS/CSS)
• Application logic (Flask routes + SocketIO handlers)
• Domain logic (utils + models)
• Infrastructure (MongoDB, file storage, external AI service)
• MVC‑style separation via Flask Blueprints and Jinja2 templates
• Event‑driven real‑time interactions using WebSockets (Flask‑SocketIO + Eventlet)
• Asynchronous, non‑blocking I/O for low‑latency voice & chat
• Extensible “multi‑agent” pattern: each domain agent (education, healthcare, construction, news, PDF, data) implemented as distinct route + template + JS module
Technologies & Frameworks
• Backend: Python, Flask, Flask‑SocketIO, Eventlet, Flask‑Dance, PyMuPDF (fitz)
• AI Engine: google‑generativeai (Gemini 1.5 Flash)
• Database: MongoDB
• Frontend: HTML/CSS, vanilla JS (Fetch API), Socket.IO client
• Deployment helpers: devserver.sh, Procfile (Heroku), Nix (dev.nix)
Final Checklist
[ ] Detailed path of a representative flow (e.g., PDF → chat → AI → response)