Skip to content
Tina Diet Docs

Troubleshooting

Quick reference for common production issues.

Bot not responding

Symptoms

User sends a message in LINE, no reply.

Diagnose

  1. Check LINE webhook delivery:

    • LINE Developers Console → Messaging API channel → Webhook
    • “Last delivery” should show recent attempts with HTTP status
    • If 200 but no reply → backend bug (check logs)
    • If non-200 → backend not reachable or returning error

  2. Check backend /healthz:

    Terminal window
    curl https://api.tinadiet.com/healthz

    Should return {"status":"ok",...}. If timeout or 500 → backend down.

  3. Check Railway logs:

    • Railway dashboard → backend → Logs tab
    • Filter for webhook.line or level":"error"
    • Look for stack traces around the time of the message

Common causes

  • LINE_CHANNEL_SECRET / ACCESS_TOKEN rotated in console but Railway env not updated → signature verification fails → 401
  • OpenAI API key invalid (key revoked, budget hit, etc.) → food parsing throws → handler crashes → no reply
  • DB locked (rare with better-sqlite3, but possible if a previous write hung) → restart backend to release

Fix

  • Rotate offending secret, update Railway env, redeploy
  • If DB issue: restart backend service via Railway dashboard

Payment doesn’t apply

Symptoms

User completed payment in PromptPay/TrueMoney but premium not active in LIFF.

Diagnose

  1. Check payments table for the charge:

    Terminal window
    node -e 'const db=require("better-sqlite3")("/data/app.db");const rows=db.prepare("SELECT * FROM payments WHERE user_id = <USER_ID> ORDER BY id DESC LIMIT 5").all();console.table(rows);'
    • Status pending → webhook didn’t arrive or failed
    • Status successful but no premium → bug in grant application
    • No row → charge never created

  2. Check Omise dashboard Recent Deliveries:

    • dashboard.omise.co/test/webhooks (or /webhooks for LIVE)
    • Look for the most recent delivery, check status
    • Status 200 → backend received OK; bug somewhere downstream
    • Status 400 BAD_SIGNATURE → OMISE_WEBHOOK_SECRET mismatch
    • Status 4xx/5xx other → backend logged the error

  3. Check Railway logs for webhook.omise:

    • webhook.omise.handled shows event processed
    • webhook.omise.handle_failed shows handler error
    • webhook.omise.bad_signature shows sig mismatch

Common causes

  • OMISE_WEBHOOK_SECRET was rolled in dashboard but Railway not updated
  • Webhook URL changed (now wrong) — verify https://api.tinadiet.com/webhooks/omise is the registered URL
  • tinadiet_user_id metadata missing from charge → backend can’t find user
  • charge metadata.tinadiet_user_id doesn’t match an existing user (created in TEST mode with placeholder, etc.)

Fix

  • Verify OMISE_WEBHOOK_SECRET matches dashboard (re-roll if uncertain)
  • Manually sync via API: curl -X GET -H "Authorization: Bearer <JWT>" https://api.tinadiet.com/api/v1/billing/omise/charge/<charge_id> — forces sync and applies grant if eligible
  • Manual grant via Railway Console (last resort) — see Manual grants

LIFF blank/stuck screen

Symptoms

User opens LIFF, sees mascot loading forever, or blank white screen.

Diagnose

  1. In LINE webview, open dev tools:

    • LINE Android: tap ︙ (top right) → “Webview tools” → Inspect
    • LINE iOS: connect to Safari debugger via USB cable
    • Check Console for errors

  2. Common error patterns:

    • LiffInitError → LIFF SDK init failed → likely wrong VITE_LIFF_ID in Cloudflare env, or LIFF channel “Status” is “Developing” (must be Public for non-admin users)
    • Failed to fetch → backend CORS or network issue
    • 401 Unauthorized → session JWT exchange failed; backend may not have valid LINE_LOGIN_CHANNEL_ID

  3. Check Cloudflare deploy status:

    • Dashboard → Workers → tinadiet-liff → Deployments
    • Most recent deploy succeeded?

Common causes

  • LIFF channel set to “Developing” instead of “Public” (only admin users can use Developing-status LIFF)
  • VITE_API_BASE_URL build env wrong → LIFF can’t reach backend
  • Backend session JWT secret rotated but old JWTs cached in LIFF SDK → exchange fails until next session

Fix

  • Set LIFF Status to Public in LINE Developers Console
  • Verify Cloudflare env vars on Workers Build
  • Rotate SESSION_JWT_SECRET and push to Railway → all sessions invalidate, next LIFF init re-exchanges

Premium expired but still showing premium UI

Symptoms

User’s premium expired (date passed) but LIFF still shows Premium badge.

Diagnose

  1. Check users.plan and premium_expires_at:

    Terminal window
    node -e 'const db=require("better-sqlite3")("/data/app.db");const row=db.prepare("SELECT id, plan, premium_expires_at FROM users WHERE id = ?").get(<USER_ID>);console.log(row);'

  2. Verify expire_premium cron is running:

    • Railway logs → search jobs.expire_premium.tick
    • Should fire daily at 02:00 ICT
    • If never fired → cron not scheduled (env CRON_ENABLED=false?)

Fix

  • Run expire_premium manually:
    Terminal window
    curl -X POST -H "x-jobs-secret: $JOBS_TRIGGER_SECRET" \
      https://api.tinadiet.com/internal/jobs/expire-premium
    Returns count of users reverted to free.
  • If cron disabled: set CRON_ENABLED=true in Railway env

OpenAI errors

Symptoms

Bot replies with “ขออภัยค่ะ Tina มีปัญหาเล็กน้อย” or similar fallback.

Diagnose

  1. Check Railway logs for food_parser or coach errors
  2. Check OpenAI dashboard usage:
    • platform.openai.com → Usage tab
    • Hit monthly budget cap?
    • API key revoked?

Common causes

  • Budget cap hit (project key has $35/month limit)
  • API key rotated but Railway env not updated
  • OpenAI outage (rare but does happen — status.openai.com)
  • gpt-4o-mini deprecated/renamed — model env points to wrong name

Fix

  • Increase budget cap in OpenAI dashboard, or rotate key, or wait for upstream recovery

Cloudflare DNS issues

Symptoms

tinadiet.com or api.tinadiet.com not resolving, or wrong target.

Diagnose

Terminal window
dig tinadiet.com
dig api.tinadiet.com
dig app.tinadiet.com

Common causes

  • DNS record deleted accidentally
  • TTL changed but cache hasn’t refreshed
  • Cloudflare proxy disabled (gray cloud instead of orange) for api — TLS may not work

Fix

  • Re-add missing DNS record in Cloudflare dashboard
  • Ensure orange cloud (proxy on) for all records

See also