33-TruthGuard

Appearance

Step 6: System Architecture & Data Flow Design ​

1. High-Level System Architecture ​

TruthGuard implements a federated micro-frontend topology combined with serverless Edge Functions and client-side forensic engines. The system unifies decentralized forensic tools under an NVIDIA-inspired dark glassmorphism dashboard.

2. Request Lifecycles (Top 3 Features) ​

1. Camera Tamper & EXIF Forensics Lifecycle (F2) ​

2. Multi-Modal Deepfake & Fact-Checking Lifecycle (F1) ​

3. Scam Call Transcript Triage Lifecycle (F3) ​

3. Database Schema & Data Modeling ​

TruthGuard leverages PostgreSQL on Supabase Cloud, governed by Row Level Security (RLS) to enforce strict role-based access.

Database Indexes & Optimization ​

  • Index on profiles.user_id:
    • CREATE UNIQUE INDEX idx_profiles_user_id ON public.profiles(user_id);
    • Reasoning: Accelerates point lookups during authentication profile retrieval in [AuthContext.tsx](file:///Users/avishjain/avishjain9-repos/33-codealchemist/mainpage/src/contexts/AuthContext.tsx#L41) to under 1ms.

4. Authentication & Authorization Flow ​

5. AI Integration & Inference Pipeline ​

  • Prompt Construction: Modality-specific envelopes encapsulate raw inputs with strict safety guardrails.
  • Structured Output Parsing: The edge function formats outputs into typed JSON schemas (AnalysisData and TelecomResult), guaranteeing that client components receive predictable numerical scores and array indicators.
  • Cost & Token Control: Text submissions are capped at 5,000 characters ([truthlence/src/pages/Index.tsx:L73](file:///Users/avishjain/avishjain9-repos/33-codealchemist/truthlence/src/pages/Index.tsx#L73)) to prevent token exhaustion and control inference costs.
  • Fallback Strategy: If an AI provider experiences downtime, the edge function falls back to deterministic heuristic regex pattern matching, ensuring the client UI never crashes.

6. Scaling to 100x Traffic & The First Three Bottlenecks ​

  1. Bottleneck 1: Edge Function Concurrency & AI Provider Quotas:
    • Issue: 100x concurrent users uploading images will trigger rate-limit throttling (HTTP 429) from inference APIs.
    • Mitigation: Introduce an asynchronous queue (BullMQ or AWS SQS) to level load peaks, combined with a Redis semantic cache that returns identical cached verdicts for duplicate SHA-256 hashes.
  2. Bottleneck 2: Browser LocalStorage 5MB Capacity:
    • Issue: MediaVault's localStorage persistence crashes with QuotaExceededError if hundreds of Base64 files are stored locally.
    • Mitigation: Transition client storage to IndexedDB via idb-keyval (supporting gigabytes of offline storage) and offload primary asset storage to AWS S3 or Supabase Storage.
  3. Bottleneck 3: Port Prober Socket Flooding:
    • Issue: Running HEAD requests across 24 local ports per user causes socket contention.
    • Mitigation: In production, replace multi-port local development with a single unified NGINX reverse proxy routing path-based micro-frontends (/truthlence, /camera, /vault).