Análise de Desempenho Mobile e Função Edge do Supabase

### 🔴 Critical Issues 1. **Edge Function Import Weight & Cold Start Latency** - Functions importing heavy Deno modules (e.g., `axios`, `puppeteer`, or large JSON configs) suffer from multi-second cold starts. - During app launch or session restore, this stacks with the mobile network RTT (200–600ms on 4G, higher on 3G) and becomes **UX-breaking**. 2. **Blocking UI on Edge Responses** - Screens that await Edge responses without optimistic UI or skeleton loaders leave users staring at a blank/spinner during cold start + network latency. - This is visible in session-restore and user-triggered actions (e.g., payment, data refresh). 3. **Global-Scope Side Effects** - Any Edge Function that initializes SDK clients, loads env vars, or warms DB connections at module top-level delays every invocation, even hot ones. 4. **Synchronous Orchestration of External APIs** - If an Edge Function awaits multiple external APIs sequentially (e.g., Stripe + OpenAI + CRM), the mobile client waits for the sum of all latencies, risking timeout on unstable networks. ### 🟠 Moderate Risks 1. **Edge Used for Heavy AI or Long-Running Tasks** - Running AI inference or batch processing in Edge Functions risks hitting Deno’s execution limits (e.g., 15s) and wastes mobile battery/network while waiting. - Mobile users lose patience; retries become frequent. 2. **No Background Execution for Non-Critical Calls** - The app may fetch non-essential data (e.g., recommendations) in the foreground, blocking interaction instead of deferring to background sync. 3. **Misplacement of Validation Logic** - If Edge Functions perform complex policy validation (e.g., business rules, role checks) that could run client-side or in a Postgres RPC, this adds unnecessary round trips. 4. **Streaming Responses in Edge** - Streaming (e.g., chat or progress) from Edge to mobile is fragile on flaky networks; connection drops cause incomplete data delivery. ### 🟢 Acceptable / Well-Designed Areas 1. **Auth / Guard Functions** - Edge Functions for JWT validation or session hydration are appropriate; they’re lightweight and fast to cold start. 2. **Simple Proxies** - Using Edge as a thin proxy to third-party APIs (with caching) reduces client CORS and secrets exposure-good practice. 3. **Optimistic UI for User Actions** - If some flows already apply optimistic updates and fire Edge calls in the background, this is correct for mobile perceived performance. --- ### Recommendations (Specific to This Codebase) 1. **Reduce Import Footprint in Critical Path Functions** - **What**: Strip unused imports; replace heavy HTTP clients with native `fetch` (Deno-native). - **Why**: Smaller bundle → faster cold start (from ~1–3s to <500ms). - **Impact**: UX improves from “UX-breaking” to “noticeable” for launch and session-restore calls. 2. **Add Client-Side Latency Budget & Background Queue** - **What**: For non-critical Edge calls (e.g., recommendations, analytics), defer them until after the primary UI is interactive or until app goes to background (using `AppState`). - **Why**: Removes synchronous waits from initial render. - **Impact**: “UX-breaking” becomes “invisible” for those flows. 3. **Move Heavy Compute to Server or Queue** - **What**: If AI/ML calls or batch jobs exist in Edge, move them to a Supabase background worker or external queue (e.g., pg_cron + server). - **Why**: Edge is for short, synchronous, lightweight tasks; offloading prevents timeout and battery drain. - **Impact**: Reliability increases; mobile users no longer wait 10+ seconds. 4. **Parallelize External API Calls Inside Edge** - **What**: Where Edge orchestrates multiple APIs, use `Promise.all` instead of sequential awaits. - **Why**: Reduces total response time from sum(D1+ D2 + D3) to max(D1, D2, D3). - **Impact**: Cuts latency by 2–3x on mobile. 5. **Use Deno’s `--no-check` and Tree-Shaking in Build** - **What**: Configure Supabase to deploy Edge Functions with `--no-check` and ensure esbuild tree-shaking is enabled. - **Why**: Avoids type-checking overhead at cold start and removes dead code. - **Impact**: Cold start drops further; bundle size shrinks. --- ### Final Verdict - **Mobile-appropriate?** No, not fully. Current architecture risks stacking cold starts and network latency in critical UX flows. - **Edge overused?** Yes, for heavy compute and orchestration; underused for auth/light proxy where it shines. - **Single highest-impact improvement?** **Strip Edge Function imports and eagerly loaded SDKs; use native fetch + tree-shaking.** This alone can halve cold start time and turn a “spinner-of-death” into a smooth first paint.