SOMA docs
Ops

Production state

Current prod topology, endpoints, and smoke-test expectations.

Snapshot as of 2026-04-19.

Surfaces

SurfaceURLHostStatus
Web apphttps://soma-ai.ccVercel Pro (happ-ai/soma)✅ SSL, CSP, auto-rollback ready
Telegram bot@happ_soma_botSame Vercel deploy, /api/telegram/webhook✅ HMAC-guarded, pending=0
Agent workflowshttps://soma-ai.cc/api/inngestSame Vercel deploy✅ 8 functions, Inngest Cloud synced
DBdiftdhvsmcnxeygapgdd.supabase.coSupabase Pro, EU-west-1✅ 7 tables, RLS on all, pgvector
DNSsoma-ai.ccCloudflare zone✅ A → 76.76.21.21 (Vercel)
SecretsDoppler soma/prod23/23 filled
Docshttps://docs.soma-ai.ccCloudflare Pages

Observability

StackProject
Langfusecloud.langfuse.com
Sentryhapp-de/soma-api (EU region)
PostHogEU cloud (eu.i.posthog.com)
Axiomdataset soma-logs

Smoke-test matrix

Everything below is green on current prod.

curl https://soma-ai.cc/api/health
# {"ok":true,"service":"web","db":true,"ts":"..."}

curl https://soma-ai.cc/api/inngest
# has_event_key:true, has_signing_key:true, mode:cloud, function_count:8

curl -I https://soma-ai.cc/login
# HTTP/2 200

curl -sI -X POST https://soma-ai.cc/api/chat
# HTTP/2 307 → /login (auth middleware working)

Telegram webhook:

curl "https://api.telegram.org/bot$TG/getWebhookInfo"
# url: https://soma-ai.cc/api/telegram/webhook
# pending_update_count: 0

End-to-end check

  1. Open https://t.me/happ_soma_bot, send /start.
  2. Send "I just read Atomic Habits by James Clear."
  3. Expect reply within 3-5s + a new row in entities with type=book, name='Atomic Habits'.
  4. Send "What books have I read?" — should recall Atomic Habits via memory_recall.
  5. Langfuse shows two traces tagged ['bot', 'message'].