Log user feedback using the SDK

LangSmith makes it easy to attach feedback to traces. This feedback can come from users, annotators, automated evaluators, and so on, which is crucial for monitoring and evaluating agents. This page details how to log feedback using the SDK. For the structure of feedback objects, refer to Feedback data format.

In an agent-based workspace, create an agent once before sending traces. Use agent addressing to select its identifier and environment. If an integration requires project_name, use the selected environment’s tracer_session_name as a compatibility value.

Use `create_feedback()` / `createFeedback`

Child runs You can attach user feedback to any child run of a trace, not just the trace (root run) itself. This is useful for critiquing specific steps of the agent, such as the retrieval step or generation step of a RAG pipeline.

Non-blocking creation (Python only) The Python client will automatically background feedback creation if you pass trace_id= to create_feedback(). This is essential for low-latency environments, where you want to make sure your agent isn’t blocked on feedback creation.

The following example creates a trace with two child runs, then logs feedback against the root run and against one of the child runs. The TypeScript snippet shows the equivalent createFeedback call shape, assuming a runId is already available from your agent.

import asyncio

from langsmith import Client, trace, traceable

@traceable
def foo(x):
    return {"y": x * 2}

@traceable
def bar(y):
    return {"z": y - 1}

client = Client()

inputs = {"x": 1}
with trace(name="foobar", inputs=inputs) as root_run:
    result = foo(**inputs)
    result = bar(**result)
    root_run.outputs = result
    trace_id = root_run.id
    child_runs = root_run.child_runs

# Resolve the session_id for the agent environment
if root_run.address is not None:
    resolved = asyncio.run(client.sessions.resolve(**root_run.address))
    session_id = resolved.session_id
else:
    session_id = client.read_project(project_name=root_run.session_name).id

# Provide feedback for a trace (a.k.a. a root run)
client.create_feedback(
    key="user_feedback",
    score=1,
    trace_id=trace_id,
    session_id=session_id,
    comment="the user said that ..."
)

# Provide feedback for a child run
foo_run_id = [run for run in child_runs if run.name == "foo"][0].id
client.create_feedback(
    key="correctness",
    score=0,
    run_id=foo_run_id,
    # trace_id= is optional but recommended to enable batched and backgrounded
    # feedback ingestion.
    trace_id=trace_id,
    session_id=session_id,
)

import { AgentAddress, Client } from "langsmith";
const client = new Client();

    // ... Run your agent and get the run_id...
    // This information can be the result of a user-facing feedback form

// Resolve the session_id for the agent environment
const address = new AgentAddress("<your-agent-id>", "production");
const { session_id: sessionId } = await client.sessions.resolve(address);

await client.createFeedback({
    runId,
    sessionId,
    key: "feedback-key",
    score: 1.0,
    comment: "comment",
});

For project-based traces, use readProject with the existing project name to get its ID instead of resolving an agent address. For resolution requirements, see Resolve an agent environment. You can even log feedback for in-progress runs using create_feedback() / createFeedback. See Access the current run (span) within a traced function for how to get the run ID of an in-progress run.

Collect feedback from a user interface

If you need to collect feedback from a browser or other client-side environment without exposing your API key, use presigned feedback tokens. These generate a URL scoped to a specific run and feedback key that clients can call directly. See Collect feedback with presigned URLs for the full guide. To learn more about how to filter traces based on various attributes, including user feedback, see Filter traces.

Connect these docs to your agent of choice via MCP for real-time answers.

Edit this page on GitHub or file an issue.

​Use create_feedback() / createFeedback

​Collect feedback from a user interface

Use `create_feedback()` / `createFeedback`

Collect feedback from a user interface