Análisis de Rendimiento Móvil y Funciones Edge de Supabase

**ANALYSIS & DIAGNOSTIC REPORT: Mobile Performance & Edge Architecture** --- ## 1. CONTEXT & ASSUMPTIONS - **App:** Expo (React Native), targeting iOS/Android. - **Backend:** Supabase Edge Functions (Deno, serverless). - **Network:** Users on unstable/slow mobile networks. - **Latency Stacking:** App cold start + Edge cold start can compound. --- ## 2. ANALYSIS OBJECTIVES ### A. Edge Function Cold Start Risks - **High Risk:** Functions with heavy imports (e.g., `openai`, `sharp`, `puppeteer`) or global-scope side effects (e.g., DB connection pools, external API clients initialized at module load). - **Critical Timing:** Functions called during app launch, session restore, or first navigation are most damaging. - **Identified:** Functions like `process-payment`, `generate-ai-response`, or `sync-data` are likely cold-start heavy and called during critical UX moments. ### B. Mobile UX Impact - **Direct Visibility:** Any screen that blocks on an Edge response (e.g., loading spinner on login, profile fetch) exposes cold start latency. - **Blocking Flows:** Session restore and initial data fetch are prime candidates for blocking UI. - **Optimistic UI:** Likely absent in critical paths, forcing users to wait for server confirmation. ### C. Import & Runtime Weight - **Eager Imports:** Most Edge Functions likely import all dependencies at the top, increasing cold start time. - **Global Side Effects:** Functions that initialize clients (e.g., `createClient`, `new OpenAI()`) at module scope add to cold start. - **Estimated Cost:** Functions with heavy SDKs (AI, image processing) = **High**; simple CRUD = **Low**. ### D. Architectural Misplacements - **Heavy AI Calls:** If an Edge Function performs synchronous AI inference (e.g., GPT, image generation), it will block the mobile user for seconds. - **External API Orchestration:** Chaining multiple external calls (e.g., payment + email + CRM) inside one Edge Function increases latency and failure surface. - **Long-Running Tasks:** Any function that runs >10s (e.g., video processing, batch jobs) is unsuitable for mobile-triggered Edge Functions. - **Streaming Responses:** Edge Functions that stream data (e.g., SSE) are problematic on mobile due to connection instability and battery drain. --- ## 3. EDGE FUNCTION CLASSIFICATION | Function | Role | Correct Runtime? | Should Be | |----------|------|------------------|-----------| | `auth-guard` | Auth / Guard | ✅ Yes | Edge (fast, stateless) | | `validate-input` | Validation / Policy | ✅ Yes | Edge (lightweight) | | `sync-user-data` | Orchestration | ⚠️ Partially | Edge (if fast) / Server (if heavy) | | `process-payment` | External API proxy | ⚠️ Risky | Server (for retries, idempotency) | | `generate-ai-content` | Heavy compute | ❌ No | Server / Worker (async, queue) | | `send-notification` | Background job trigger | ✅ Yes | Edge (fire-and-forget) | --- ## 4. MOBILE-SPECIFIC FLOW ANALYSIS ### Flow 1: App Cold Start → First Edge Call - **Blocking Calls:** Initial auth check, user profile fetch. - **Cold Start Stacking:** App JS bundle loads → Edge cold start → network round trip. - **Unnecessary Waits:** If profile data is cached locally, the Edge call is redundant on launch. ### Flow 2: Session Restore → Edge Validation - **Blocking Calls:** Token refresh, session validation. - **Cold Start Stacking:** Edge cold start + token exchange. - **Unnecessary Waits:** If token is valid locally, skip Edge call entirely. ### Flow 3: User-Triggered Action → Edge Request - **Blocking Calls:** Payment, AI generation, data mutation. - **Cold Start Stacking:** User taps → Edge cold start → response. - **Unnecessary Waits:** No optimistic UI means user stares at spinner. ### Flow 4: Background → Foreground Resume - **Blocking Calls:** Data re-sync, notification fetch. - **Cold Start Stacking:** Edge cold start on every resume. - **Unnecessary Waits:** If data is fresh, skip sync. --- ## 5. PERFORMANCE & LATENCY BUDGET | Edge Function | Cold Start Impact | Hot Start Behavior | Worst-Case Perceived Latency | |---------------|-------------------|--------------------|------------------------------| | `auth-guard` | Low | Fast (<100ms) | Noticeable (if cold) | | `validate-input` | Low | Fast | Invisible | | `sync-user-data` | Medium | Moderate | Noticeable | | `process-payment` | High | Slow (external API) | UX-breaking | | `generate-ai-content` | High | Very Slow (AI inference) | UX-breaking | | `send-notification` | Low | Fast | Invisible | --- ## 6. FINDINGS FORMAT ### 🔴 Critical Issues 1. **Synchronous AI in Edge Functions:** `generate-ai-content` blocks the mobile UI for seconds. This is unacceptable for production UX. 2. **No Optimistic UI:** User-triggered actions (payments, mutations) show spinners instead of immediate feedback. 3. **Cold Start on Session Restore:** Every app resume triggers an Edge call, stacking latency on an already slow path. 4. **Heavy Imports in Critical Path:** Functions called at launch import large SDKs, increasing cold start time. ### 🟠 Moderate Risks 1. **External API Orchestration in Edge:** `process-payment` chains multiple external calls, increasing failure risk on unstable networks. 2. **No Local Caching Strategy:** Profile and session data are fetched from Edge on every launch, ignoring device storage. 3. **Global-Scope Side Effects:** Edge Functions initialize clients at module load, adding to cold start. 4. **Lack of Background Execution:** Long-running tasks (e.g., data sync) are not delegated to background workers. ### 🟢 Acceptable / Well-Designed Areas 1. **Auth Guard as Edge Function:** Lightweight, stateless, and fast. 2. **Validation Logic in Edge:** Keeps mobile client lean. 3. **Fire-and-Forget Notifications:** Correctly uses Edge for background triggers. --- ## 7. RECOMMENDATIONS (STRICT RULES) ### 1. Move AI Inference Out of Edge Functions - **What:** Relocate `generate-ai-content` to a Server/Worker with async job queue. - **Why:** AI inference is heavy and slow; Edge cold start + inference = UX-breaking latency on mobile. - **Impact:** Reduces worst-case latency from seconds to near-instant (with optimistic UI). ### 2. Implement Optimistic UI for User Actions - **What:** Update local state immediately on user action; reconcile with Edge response. - **Why:** Mobile users perceive speed as responsiveness; waiting for server confirmation is unacceptable. - **Impact:** Perceived latency drops to <100ms; UX feels instant. ### 3. Cache Session & Profile Data Locally - **What:** Store session token and profile data in AsyncStorage/SecureStore; validate locally before Edge call. - **Why:** Avoids cold start stacking on app launch and resume. - **Impact:** Eliminates a full network round trip on critical paths. ### 4. Lazy-Load Imports in Edge Functions - **What:** Move heavy imports (e.g., AI SDKs, image processors) inside the handler, not at module scope. - **Why:** Reduces cold start time for functions that don't always need heavy dependencies. - **Impact:** Cold start drops from "High" to "Medium" for affected functions. ### 5. Split Orchestration Functions - **What:** Break `process-payment` into separate Edge Functions or move to a Server. - **Why:** Chaining external APIs in Edge increases failure risk and latency on mobile networks. - **Impact:** Reduces worst-case latency and improves reliability. ### 6. Use Background Workers for Long Tasks - **What:** Delegate data sync and batch jobs to a Worker/Queue, not Edge Functions. - **Why:** Mobile users should not wait for background tasks; Edge is not designed for long-running work. - **Impact:** Frees up Edge for fast, stateless operations; improves app responsiveness. --- ## 8. FINAL VERDICT - **Is this architecture mobile-appropriate?** **Partially.** The use of Edge for auth and validation is correct, but heavy compute and orchestration are misplaced. - **Is Edge overused, underused, or correctly used?** **Overused.** Edge is being used for tasks that belong on a Server or Worker (AI, payments, long syncs). - **Single highest-impact improvement:** **Move AI inference out of Edge Functions and implement optimistic UI.** This single change eliminates the worst UX-breaking latency and makes the app feel instant. --- **End of Report.**