🧩 LangSmith Traces

LangSmith is LangChain’s observability platform for LLM applications. The LangSmith Traces data source connector reads the run history and feedback of your LangSmith projects and writes them to two incremental tables in Storage.

Experimental

This connector is experimental. Its configuration and output may still change. Send feedback using the feedback button in your project.

Prerequisites

• A LangSmith user API key. The connector stores it encrypted.
• For organizations on the EU instance, the EU API URL https://eu.api.smith.langchain.com.

Configuration

Create a new configuration of the LangSmith Traces data source connector. The connector has no rows; one configuration is one extraction.

Credentials

Field	Description
LangSmith API key	Your user API key.
LangSmith API URL	Optional. Leave empty for the default US instance (https://api.smith.langchain.com), or enter https://eu.api.smith.langchain.com for the EU instance.

Click Test connection.

Scope

Field	Description
Projects	Leave empty to sync every project the API key can see, or click Reload projects and pick specific ones.
Fetch from	Start of the interval: an ISO 8601 timestamp such as 2024-01-01T00:00:00Z, or a relative phrase such as 7 days ago (default).
Fetch to	End of the interval. Default Now.

Advanced parameters

Field	Description
Max retries	Retry attempts when the LangSmith API throttles requests. Default 5.
Feedback batch size	Number of runs processed together when fetching feedback. Default 200; lower it if you hit memory limits.

Output

Two tables, both loaded incrementally:

Table	Primary key	Contents
runs	id	The run history.
langsmith_feedback	id, run_id	The feedback on the runs, including evaluator scores and corrections.