Generated-text validation
A successful HTTP response must contain generated text. Empty, blocked, malformed, and error responses fail the check.
Real inference
Know when your model can return an answer.
Check OpenRouter, OpenAI, Anthropic, and Gemini with real inference. Require generated text and record latency and token usage using configured provider credentials.
LLM Uptime Monitoring
Real inference
What Telesis inspects
- Generated-text validation
- Bounded token usage
- Scoped credentials
- Model-specific evidence
What you get
Bounded token usage
A small fixed prompt, at most 64 output tokens, no executor retries, and hourly schedule defaults keep probe scope explicit. Provider credits are consumed.
Scoped credentials
Keys stay on the executor. Scheduled checks must match the configured organization and allowed model; public quick checks cannot use runner keys.
Model-specific evidence
Read model, provider, latency, generated-text presence, and token count. Results measure that credential and region, not every provider endpoint.
Why teams pick it
- OpenRouter and three direct provider integrations
- Keys and generated response text excluded from evidence
- CLI and MCP support for local checks
- Scheduled checks use configured runners and exact provider endpoints
Ready to try LLM Uptime Monitoring?
Check OpenRouter, OpenAI, Anthropic, and Gemini with real inference. Require generated text and record latency and token usage using configured provider credentials.