Run bulk prompts

Submit a CSV of prompts to a Text-to-Text model and watch the job finish in Async History.

Before you begin

• A Text-to-Text model’s page open in GenAI API. Bulk Prompts starts from there.
• An existing dataset to hold the model’s responses, other than a file-system dataset that has already finished copying.
• A CSV file with a column headed prompts, one prompt per row, saved as UTF-8 and up to 10 MB in size. The file can have other columns too.

Prepare your CSV file

Save the file as plain comma-separated values, with one prompt per row in the prompts column. GenAI API accepts only a valid CSV file with a prompts column. No published limit applies to how many prompts a file can hold.

Upload the file

1. On the model’s page, open Playground, and select the link to upload Bulk Prompts.
2. Under Select Dataset, pick the existing dataset that will hold the model’s responses. This step offers existing datasets only.
3. Upload your CSV file, one file at a time. The upload accepts only .csv files.
4. Submit the job. When the submission succeeds, Async History opens so you can watch its progress.

If the dataset is not valid for your project, or GenAI API cannot find your file, a message names what went wrong. Troubleshooting TIR GenAI API explains each one.

Track progress in Async History

Each entry’s status is one of Queued, Processing, Completed, or Failed. An entry in any other state shows as Unknown.

• Download is available once an entry’s status is Completed.
• Delete is available once an entry’s status is Completed or Failed, and asks you to confirm before it goes through.
• Clear All removes every Completed and Failed entry at once, after you confirm. It is always available.

Download the responses

Once an entry’s status reaches Completed, select Download to get the model’s responses for every prompt in your file.

To send one prompt at a time instead, see Use the Playground. For help with any message, see Troubleshooting TIR GenAI API.