FeaturesEvals

Datasets

Create and manage datasets for evaluations and experiments.
  1. Sign up — Create an account at platform.respan.ai
  2. Create an API key — Generate one on the API keys page
  3. Add credits or a provider key — Add credits on the Credits page or connect your own provider key on the Integrations page

Add the Docs MCP to your AI coding tool to get help building with Respan. No API key needed.

1{
2  "mcpServers": {
3    "respan-docs": {
4      "url": "https://mcp.respan.ai/mcp/docs"
5    }
6  }
7}

What is a dataset?

A dataset is a collection of test inputs (and optionally outputs) that you use to run experiments. Each row represents one test case. You can build datasets from production spans or upload a CSV. Datasets power both offline experiments and evaluator testing.

In the current UI, clicking Create dataset always creates an empty dataset first. After you open the dataset detail table page, click Insert rows to populate it in one of two ways:

  1. Insert by sampling
  2. Insert from CSV

Create a dataset

Start by creating an empty dataset. Then use Insert rows inside the dataset detail table page to add rows.

1

Go to Datasets and click Create dataset

Navigate to Datasets and click Create dataset. It is located under the Evals section.

2

Open the dataset detail table page

After the dataset is created, open it to access the rows table and the Insert rows menu.

Insert by sampling

Sample request logs and add them as rows to an existing dataset. Use filters, a time range, and a sampling percentage to control what gets inserted.

1

Choose Insert rows > Insert by sampling

In the dataset detail table page, click Insert rows and select Insert by sampling.

2

Add filters and choose a time range

In the modal, click + Filter to narrow down which request logs to sample. You can filter by fields such as status, model, or metadata, then choose the time range to pull from.

3

Set the sampling percentage and review estimated rows

Adjust the Sampling percentage to control how many matching logs are inserted. The modal also shows Estimated rows so you can preview how many rows will be added before confirming.

5

Insert rows and verify

Click Insert to add the sampled rows to the existing dataset. Once ingestion finishes, confirm the new rows appear in the dataset table.

Insert from CSV

Upload rows from a CSV file into an existing dataset.

1

Step 1: Insert the CSV file

In the dataset detail table page, click Insert rows and select Insert from CSV. Then select the file from your computer.

Once the file is uploaded, Respan takes you to the field mapping step.

2

Step 2: Map all fields from the CSV

After the file is uploaded, review every field under Incoming fields and map it to the correct Destination field in the dataset.

For each CSV column, choose the matching dataset field from the dropdown. Use the preview panel on the right to confirm the selected field contains the expected values. If a column should not be imported, leave it as Not mapped.

You can map multiple CSV columns to input. When you do, Respan wraps those columns into a single input object, using the original CSV column names as JSON keys and the CSV cell values as the corresponding values.

3

Review the data preview and import

Use the data preview to verify the mapped values look correct, then click Import to add the rows to the dataset. You can import up to 500 rows at a time, and the imported rows are created as request logs in the dataset.

CSV template (Google Drive)