Dataset Management

Create a dataset, add columns, enter rows manually or via CSV, run evaluations, and export results from the Future AGI dashboard. No code required.

📝
TL;DR

Create a dataset, add columns, populate rows manually and via CSV import, run an evaluation on the data, and export the results. All from the Future AGI dashboard, no code required.

TimeDifficultyPackage
10 minBeginnerDashboard only
Prerequisites

Tutorial

Create a dataset

  1. Go to app.futureagi.com, then Dataset, then Add Dataset
  2. Select Add datasets Manually
  3. Fill in:
    • Dataset Name: rag-eval-dataset
    • No. of Rows: 3
    • No. of Columns: 4
  4. Click Save

You should see an empty rag-eval-dataset grid with 3 rows and 4 unnamed columns.

Add columns

Click Add Column then Static Columns then Text. Add these four columns one at a time:

  1. input
  2. output
  3. context
  4. expected_answer

You should see all four columns named across the top of the grid, replacing the unnamed placeholders.

Enter rows manually

Click any cell and type to fill it in.

Row 1:

inputoutputcontextexpected_answer
What is the return window for electronics?Electronics can be returned within 30 days of purchase with a receipt, and shipping is free on all returns.Our return policy allows 30-day returns on all electronics with original receipt. Items must be in original packaging.30 days with receipt and original packaging.

Row 2:

inputoutputcontextexpected_answer
Do you offer international shipping?Yes, we ship to over 50 countries. Delivery takes 7-14 business days.International shipping is available to 50+ countries. Standard delivery is 7-14 business days. Express shipping is available for select destinations.Yes, to 50+ countries, 7-14 business days standard.

Row 3:

inputoutputcontextexpected_answer
How do I reset my account password?Click “Forgot password” on the login page and follow the email link.To reset your password, go to the login page and click “Forgot password”. You will receive an email with a reset link valid for 24 hours.Use the “Forgot password” link on the login page; the reset link expires in 24 hours.

You should see all 3 rows filled in across the 4 columns.

Import more rows from CSV

Save as rag-eval-rows.csv:

input,output,context,expected_answer
"What payment methods do you accept?","We accept Visa, Mastercard, PayPal, and bank transfers.","Accepted payment methods include Visa, Mastercard, American Express, PayPal, and direct bank transfer.","Visa, Mastercard, PayPal, and bank transfers."
"Can I cancel an order after placing it?","Orders can be cancelled within 2 hours of placement.","Orders are eligible for cancellation within 2 hours of being placed. After this window, the order enters processing and cannot be cancelled.","Yes, within 2 hours of placement."
"Is there a loyalty rewards program?","Yes, earn 1 point per dollar spent. Points expire after 12 months.","Our loyalty program awards 1 point per $1 spent. 100 points equals $1 in rewards. Points expire 12 months after being earned.","Yes, 1 point per dollar. 100 points = $1. Points expire after 12 months."

Note

CSV column headers must match your dataset column names exactly (case-sensitive). Unmatched headers create new columns.

  1. Click Add Row then Upload a file (JSONl/ JSON/ CSV)
  2. Drop or browse for rag-eval-rows.csv
  3. Click Done

You should see the dataset grow from 3 rows to 6, with the 3 new rows matched into the existing input/output/context/expected_answer columns.

Run evals on the dataset

  1. Click Evaluate then Add Evaluations
  2. Select groundedness
  3. Map keys: output to output, context to context, input to input
  4. Click Add & Run

You should see a new groundedness column appear with a score for each of the 6 rows. Open Row 1 (the return window question): its output claims shipping is free on all returns, but the context never mentions shipping cost, so the groundedness eval marks it as not grounded. That’s the fix to make: either remove the unsupported shipping claim from the output cell or add a sentence to context that actually supports it, then rerun the evaluation on that row.

Download the dataset

Click the download icon in the dataset toolbar to export as CSV.

You should get rag-eval-dataset.csv with all original columns plus the groundedness score column.

Troubleshooting

SymptomCauseFix
CSV import adds new columns instead of filling existing onesCSV header spelling or case doesn’t match the dataset column name exactlyRename the CSV header to match the dataset column exactly, then re-upload
A column stays empty after CSV importThe matched CSV column had blank cells for those rowsOpen the cell and type the value in manually, or fix the CSV and re-import
Add Evaluations has no metric that fitsThe dataset’s columns don’t cover the metric’s required keys (for example no context column for a groundedness check)Add the missing column first, then open Add Evaluations again
Eval run finishes but some scores are blankThe row was missing a value in one of the mapped columns when the run startedFill in the missing cell and rerun the evaluation on that row
Row count after CSV import is lower than expectedThe CSV has a malformed row (unescaped comma or unmatched quote) that fails to parseOpen the CSV in a spreadsheet tool, fix the row, and re-upload
Downloaded CSV is missing the eval score columnThe download was triggered while the evaluation was still runningWait until every row shows a score, then click download again

Next

Manage the same dataset from code, including batch evaluation, in Dataset SDK: Batch Evaluation.

Was this page helpful?

Questions & Discussion