Skip to main content
Datasets are collections of input/output pairs used to evaluate your agents systematically. You can build datasets from real traces, upload them as CSV, or populate them manually. Once created, run any evaluation template against the dataset to measure quality at scale.
Zespan datasets view showing dataset list with item counts and evaluation history

​
Creating a dataset

​
From traces

The fastest way to build a dataset is from existing traces:
  1. Open Traces and filter to the runs you want to evaluate
  2. Select one or more trace rows using the checkboxes
  3. Click Add to dataset → choose an existing dataset or create a new one
The trace’s input (prompt or agent instruction) and output (completion or agent response) are added as a row.

​
By uploading CSV

Upload a CSV file with columns matching the dataset schema. Required columns:
ColumnDescription
inputThe prompt or instruction sent to the agent
outputThe agent’s response to evaluate
expected(Optional) The ground truth answer for comparison evaluators
Go to Datasets → New dataset → Upload CSV and select your file.

​
Manually

Add rows one at a time using the Add row button. Useful for small curated datasets of known edge cases.

​
Running evaluations

Select a dataset and click Run evaluation. Choose an evaluation template and configure:
  • Evaluator — the LLM judge and prompt template to use
  • Sample size — evaluate all rows or a random sample
  • Concurrency — how many rows to evaluate in parallel
Results appear in the Evaluations section linked to this dataset run.

​
Dataset versioning

Each dataset has a version history. When you add or remove rows, the previous version is preserved. Evaluation runs are tied to a specific dataset version so results remain reproducible.

​
Next steps

  • Evaluations — run and review evaluation results
  • Simulations — test prompt changes against a dataset before deploying