Netra Custom Spans And Actions

Use this skill when auto/decorator tracing is not enough and you need explicit span lifecycle control.

When To Use

• You need spans around external APIs, tool calls, or batch steps.
• You need usage/cost tracking per operation.
• You need structured action records for DB/API side effects.

Procedure

1. Wrap each critical operation in Netra.start_span(...).
2. Set span attributes (model, llm_system, operation keys).
3. Attach usage via UsageModel.
4. Attach actions via ActionModel.
5. Add events for checkpoints and failures.

Python Pattern

from netra import Netra, UsageModel, ActionModel