While experimental OTEL support exists for LLM calls, a production-grade backend integration requires deeper observability to track agent behavior, performance, and costs.
Requirements:
- Add spans for tool execution (e.g., tracking bash or read duration).
- Trace internal engine logic (planning and execution loops).
- Export standardized metrics (token usage, provider latency, cost) via OTLP.
- Allow configuration of OTEL exporters directly through aictrl.json.
While experimental OTEL support exists for LLM calls, a production-grade backend integration requires deeper observability to track agent behavior, performance, and costs.
Requirements: