Docs
Datasets
Overview

Datasets

Datasets in Langfuse are a collection of inputs (and expected outputs) of an LLM application. They are used to benchmark new releases before deployment to production. Datasets can be incrementally created from new edge cases found in production.

For an end-to-end example, check out the Datasets Notebook (Python).

Creating a dataset

Datasets have a name which is unique within a project.

langfuse.create_dataset(name="<dataset_name>")

Create new dataset items

Individual items can be added to a dataset by providing the input and optionally the expected output.

langfuse.create_dataset_item(
    dataset_name="<dataset_name>",
    # any python object or value
    input={
        "text": "hello world"
    },
    # any python object or value, optional
    expected_output={
        "text": "hello world"
    }
)

Create items from production data

In the UI, use + Add to dataset on any observation (span, event, generation) of a production trace.

Edit/archive items

In the UI, you can edit or archive items by clicking on the item in the table. Archiving items will remove them from future experiment runs.

Run experiment on a dataset

When running an experiment on a dataset, the application that shall be tested is executed for each item in the dataset. The execution trace is then linked to the dataset item. This allows to compare different runs of the same application on the same dataset. Each experiment is identified by a run_name.

Optionally, the output of the application can be evaluated to compare different runs more easily. Use any evaluation function and add a score to the observation. More details on scores/evals here.

dataset = langfuse.get_dataset("<dataset_name>")
 
for item in dataset.items:
    # execute application function and get Langfuse parent observation (span/generation/event)
    # output also returned as it is used to evaluate the run
    generation, output = my_llm_application.run(item.input)
 
    # link the execution trace to the dataset item and give it a run_name
    item.link(generation, "<run_name>")
 
    # optionally, evaluate the output to compare different runs more easily
    generation.score(
        name="<example_eval>",
        # any float value
        value=my_eval_fn(
            item.input,
            output,
            item.expected_output
        )
    )

Evaluate dataset runs

After each experiment run on a dataset, you can check the aggregated score in the dataset runs table.

Dataset runs

Conceptually

Was this page useful?

Questions? We're here to help

Subscribe to updates