tracing-observability-and-configuration

# Tracing, Observability, and Configuration Use this sub-skill when a task needs any of the following: - `.env` loading and runtime bootstrap with `setup_env` - logging helpers such as `get_logger` and `printc` - generator state or call logs, including the tracing decorators - trace providers, spans, and local callback ordering for debugging - optional MLflow tracing setup and local troubleshooting - config-file driven observability settings and artifact hygiene Do **not** use this sub-skill for: - evaluation metrics or optimization workflows - agent streaming semantics or runner event plumbing - provider-specific `model_kwargs`, prompt shaping, or generator behavior Start here: - [Tracing and config notes](references/tracing-and-config.md) - [Verified API reference](references/api-reference.md) - [Troubleshooting guide](references/troubleshooting.md) - [Bundled smoke script](scripts/tracing_smoke.py) Practical routing hints: - Use the environment and logging helpers first when traces are missing or noisy. - Use generator state and call loggers when you need prompt history or failed-call artifacts. - Use `trace(...)` and nested spans when you need structured execution evidence. - Use the bundled smoke script when you need a safe, service-free sanity check.