Verify before you go live.
Track what is actually known. No Telegram or external-provider test is marked successful without execution.
Integration test checklist
Telegram commands
/start, /help, /clear, and /status
Real AI response
Valid model credentials and a genuine question
Conversation memory
Follow-up context, reply references, and /clear
Web search & source links
Tavily query and inspected, clickable sources
Invalid credentials
Safe errors with no secret leakage
Timeouts & rate limits
Bounded retries and deferred delivery
User isolation
Two Telegram accounts with distinct context
Telegram message delivery
Actual received messages and threaded replies
Duplicate update handling
Replay an accepted webhook update
Deployed health check
Request the real worker /health endpoint
Browser AI playground
NOT TESTEDNo browser AI request has been executed in this session. Use the playground on Overview to run a real request.
Open playgroundRun the supplied automated tests
Use npm run test:bot from the exported project. Tests cover command text, namespace separation, history limits, source validation, Unicode message splitting, invalid keys, timeout handling, and rate-limit handling. These have not been run here; fixture tests do not prove real integration success.