User guideAI Teammates
Insights (LLM & agents)
Insights shows live OpenTelemetry GenAI / LLM span traffic: cost, tokens, traces, p95 latency, and error rate. It is separate from AI Teammates (hireable SRE employees).
| Console | Path |
|---|---|
| Insights (Agent / Stream) | /observability/llm |
| Sessions | /observability/llm/sessions |
| Quality (evals / scorers) | /observability/llm/evals |
| AI Teammates | /observability/ai-teammates |
Instrument customer traffic
Emit spans with (minimum):
gen_ai.request.modelgen_ai.usage.input_tokens/gen_ai.usage.output_tokens- Optional:
gen_ai.system,gen_ai.usage.cost,gen_ai.conversation.id(orsession.id/thread.id)
Use OpenTelemetry OpenAI instrumentation (or compatible providers), or libraries that emit GenAI semantic conventions. Prompt and completion text are not stored or shown in Owlpane.
Hire AI employees for SRE
Use AI Teammates to hire role templates, bind Ask with agentId, and use the approve queue. Platform LLM or BYOK must be configured so teammates can summarize evidence. See AI Teammates and operator flags in operations.
Last updated