Trigger Remote Custom Experiments from UI
Marlies MayerhoferEnable non-technical users to trigger external experiment pipelines directly from the Langfuse UI on datasets.
You can now trigger external experiment pipelines directly from the Langfuse UI. This feature enables non-technical users to run experiments on Datasets without requiring code changes or manual setup.
Configure a remote experiment URL at the dataset level, and team members can trigger experiments with a single click. Langfuse will send the dataset name, ID, and any custom payload to your experiment service, then wait up to 10 seconds for a response.
Features
- Dataset-level configuration: Set up experiment URLs and custom payloads per dataset
- Simple triggering: One-click experiment execution from the UI
- Custom payloads: Send additional context or parameters to your experiment service
- Response handling: 10-second timeout with status feedback
Setup
- Navigate to your Dataset in Langfuse. In the top right corner of your Dataset Runs page, click New Experiment
- Configure the Remote Experiment URL where your experiment service is hosted
- Add any Custom Payload (JSON format) needed by your experiment runner
Triggering an experiment
- Click New Experiment in the top right corner of the dataset runs page. Click the Run button
- Confirm your payload, and click Run again to send dataset details to your service
Integration
Your remote experiment service will receive:
- Dataset name and ID
- Custom payload (if configured)
Ideally, your experiment runner uses Langfuse SDKs/API to push the dataset run and evaluation results back to Langfuse for complete experiment tracking.
Questions or Feedback?
Share your remote experiment workflows and use cases in our GitHub Discussions.
Please open a GitHub issue if you encounter any problems or have suggestions for improvements.