Is your feature request related to a problem?
We lack visibility into the health of our most used endpoints (llm/call, evaluation/, llm/chain). When an endpoint goes down, we only discover it through failures. We currently do have monitoring and alerts in place to track PENDING jobs which remain in that state for longer than expected, but this is retroactive behavior.
Describe the solution you'd like
Include:
- Multiple endpoints (llm/call, evaluation/, llm/chain)
- Multiple providers (Google/Vertex, OpenAI, Sarvam, Elevenlabs etc.)
- All supported modalities (text, STT, TTS)
Ensure it remains minimal and low-cost:
- Use a short sentence/phrase as probe input to minimize API costs
- Focus on endpoint health signals rather than a complete monitoring framework
Is your feature request related to a problem?
We lack visibility into the health of our most used endpoints (llm/call, evaluation/, llm/chain). When an endpoint goes down, we only discover it through failures. We currently do have monitoring and alerts in place to track
PENDINGjobs which remain in that state for longer than expected, but this is retroactive behavior.Describe the solution you'd like
Include:
Ensure it remains minimal and low-cost: