Tracing And Runs Api

Agent wraps a multi-step pipeline with @traceable on the orchestrator and child @traceable on each step. A trace tree is expected with the orchestrator as parent_run and three child runs under it.

Decorate the orchestrator and each step with @traceable; rely on LANGSMITH_TRACING=true and Python contextvars (or run_tree propagation) to auto-attach child runs to the active parent. Do NOT pass parent_run_id manually unless crossing a thread/process boundary — incorrect manual ids create orphane…

LLM call streams tokens via SSE. The client aborts mid-stream after 12 of an expected 200 tokens. Operator wants the span to land in LangSmith with the partial output and an error marker.

Finalize the run on cancellation with end_time set, the partial outputs captured, and error set to a descriptive message (e.g., 'client disconnect after N tokens'). Do not leave the run in 'running' state — orphan runs skew P50/P99 latency and inflate active-run counts.

Operator wants to export only top-level production traces (no nested child runs) for batch analysis via client.list_runs.

Pass is_root=True to client.list_runs along with a project_name filter. The API returns only runs where parent_run_id is null. Iterating list_runs without is_root pulls every child span and inflates result size by an order of magnitude in deep traces.