fix(ui): actionable 413/504/401 toasts + VPS nginx deploy template

Production QA reported "Submit failed 413" on /generation and plain
"Upload failed" (no detail) on resource upload. Root cause is the
outer reverse-proxy nginx on the VPS — still on default
`client_max_body_size 1m` — rejecting requests before they reach
Odoo. The inner docker-nginx and Odoo limits were already raised in
earlier commits; only the VPS proxy was left unpatched.

UI side
- Add `describeApiError(err, fallback, context)` helper in
  lib/api-client.ts. Maps common HTTP codes to human-readable,
  actionable toast descriptions (413 → "ask ops to raise nginx
  client_max_body_size", 504 → "server took too long, retry", 502
  → "Odoo is restarting", 401 → "sign in again", 4xx/5xx → server
  error body + status code).
- Use it in GenerationPage submit, ResourceManager upload,
  TeacherLibrary upload, CustomExamCreate save/publish. Replaces
  four slightly-different ad-hoc mappings with one source of truth.

Deploy side
- Add deploy/nginx-vps.conf.example: full TLS reverse-proxy template
  with the body/timeout knobs that must match Odoo's limit_request
  (128 MB) and limit_time_real (900 s).
- Add docs/DEPLOY_413_504_FIX.md: step-by-step remediation guide for
  the team lead covering the 3 layers that can block large bodies
  (outer nginx, Cloudflare, Odoo worker) and how to verify each.

Made-with: Cursor

deploy/nginx-vps.conf.example

@@ -0,0 +1,85 @@
+# EnCoach — example VPS reverse-proxy config
+#
+# Sits in front of the docker-compose stack (frontend:80 + backend:8069).
+# Copy to /etc/nginx/sites-available/encoach, edit server_name / TLS paths,
+# symlink to sites-enabled, then `nginx -t && systemctl reload nginx`.
+#
+# The inner frontend container already raises its own nginx limits
+# (see frontend/Dockerfile). If THIS outer proxy isn't raised too, the
+# outer layer returns 413/504 before traffic ever reaches Odoo, and users
+# see "Submit failed 413" or "Upload failed" with no server-side log entry.
+
+# ─────────────────────────────────────────────────────────────────
+# HTTP → HTTPS redirect (remove if you're not terminating TLS here)
+server {
+    listen 80;
+    listen [::]:80;
+    server_name encoach.example.com;
+    return 301 https://$host$request_uri;
+}
+
+# ─────────────────────────────────────────────────────────────────
+# Main TLS vhost
+server {
+    listen 443 ssl http2;
+    listen [::]:443 ssl http2;
+    server_name encoach.example.com;
+
+    ssl_certificate     /etc/letsencrypt/live/encoach.example.com/fullchain.pem;
+    ssl_certificate_key /etc/letsencrypt/live/encoach.example.com/privkey.pem;
+    ssl_protocols       TLSv1.2 TLSv1.3;
+    ssl_ciphers         HIGH:!aNULL:!MD5;
+
+    # ── CRITICAL ────────────────────────────────────────────────
+    # Default nginx caps request bodies at 1 MB, which triggers
+    # HTTP 413 on:
+    #   * /api/resources          (PDF / audio / image uploads)
+    #   * /api/exam/generation/submit (full module payload — can
+    #     easily be 2–10 MB when all tasks are filled)
+    #   * /api/approval-requests  (attachments on approval flow)
+    # Match this value to Odoo's `limit_request` (128 MB).
+    client_max_body_size       128m;
+    client_body_buffer_size    1m;
+    client_body_timeout        900s;
+    large_client_header_buffers 4 16k;
+
+    # ── Static / SPA served by the frontend container ───────────
+    location / {
+        proxy_pass         http://127.0.0.1:8080;   # docker-compose: frontend:80
+        proxy_set_header   Host              $host;
+        proxy_set_header   X-Real-IP         $remote_addr;
+        proxy_set_header   X-Forwarded-For   $proxy_add_x_forwarded_for;
+        proxy_set_header   X-Forwarded-Proto $scheme;
+    }
+
+    # ── Odoo API: AI calls can take minutes, needs long timeouts ─
+    # Must exceed Odoo's limit_time_real (900s in odoo-docker.conf).
+    # proxy_request_buffering off streams the body to Odoo instead of
+    # buffering it entirely on disk → less chance of hitting the 413
+    # when uploading big resources.
+    location /api/ {
+        proxy_pass                 http://127.0.0.1:8069/api/;
+        proxy_set_header           Host              $host;
+        proxy_set_header           X-Real-IP         $remote_addr;
+        proxy_set_header           X-Forwarded-For   $proxy_add_x_forwarded_for;
+        proxy_set_header           X-Forwarded-Proto $scheme;
+        proxy_http_version         1.1;
+        proxy_read_timeout         900s;
+        proxy_send_timeout         900s;
+        proxy_connect_timeout      60s;
+        proxy_request_buffering    off;
+        proxy_buffering            off;    # stream AI responses to the client
+    }
+
+    # ── Odoo web client + longpolling (optional, for /web/* URLs) ─
+    location /longpolling/ {
+        proxy_pass                 http://127.0.0.1:8069/longpolling/;
+        proxy_http_version         1.1;
+        proxy_read_timeout         3600s;
+        proxy_send_timeout         3600s;
+    }
+
+    # Access / error logs — keep them separate so 413s are easy to spot
+    access_log /var/log/nginx/encoach.access.log;
+    error_log  /var/log/nginx/encoach.error.log warn;
+}