Users, feedback & signals

Identifying users

identify_user associates a stable user ID — and optional traits — with the current run. Traits are free-form: a plan, a company, a cohort.

client.identify_user("cus_123", traits={"plan": "pro", "company": "Acme"})

Scoped user context

When you handle many runs for the same user, the user(...) context manager scopes identity to a block — every trace opened inside it inherits the user automatically.

with client.user(user_id="cus_123", traits={"plan": "pro"}):    with client.trace("agent_run") as trace:        ...  # this trace is attributed to cus_123

Feedback vs signals

Both record an outcome on a trace, but they mean different things:

• feedback — explicit reactions from the user: a thumbs-up, a star rating, a CSAT score.
• signal— implicit behavioral or business outcomes the user didn't explicitly give: an order was placed, a ticket was reopened, a session converted.

Both can be called on the trace, the client, or the module:

trace.feedback("thumbs_up", True)trace.signal("order_placed", {"order_id": "ord_123"})

Attaching to a known trace

Feedback and signals often arrive after the run has finished — a rating submitted minutes later. When you only have the trace ID, pass it explicitly and the event attaches to the right trace.

client.feedback("csat", 5, trace_id="trace_abc")client.signal("ticket_reopened", True, trace_id="trace_abc")