How to Handle Rate Limits When Processing Large Datasets with OpenAI API | Clay

How to Handle Rate Limits When Processing Large Datasets with OpenAI API

I'm using Claygent with chatGPT API on a table with 5k rows. I want to run 2k rows at a time but I am getting error in few columns because I'm hitting rate limits on OpenAI API calls. Is there a way to keep running all the rows that have errors? Using "Run all rows that haven't run or have errors" is out of questions since I only want to run 2000 out of the 5k total records.

11 comments

Channeled
APP
·
·
You can run specific rows with errors by using the "Run Starting from a Specific Row" option: 1. Right-click on your Claygent column 2. Hover over "Run Column" > "Choose Number of rows to run" 3. Specify 2000 as the number of rows and set the starting row to target your error rows For OpenAI API rate limits, consider adding a rate limit in your column settings to prevent errors. In the column settings, define a request limit and duration (in milliseconds) that aligns with your API tier limits.
Sources: HTTP API integration overview | Documentation | Clay University Run Options
Jaimin B.
·
·
Choose Number of rows to run wouldn't work since a few random columns have already run, and I woulnd't want to rerun all the columns from a cost point of view.
Channeled
APP
·
·
You can target only the rows with errors by: 1. Creating a new column with a conditional formula that identifies error rows 2. Using "Choose Number of Rows to Run" with a filter: • Add a Row Number column using a number generator • Apply a filter to target only rows with errors • Right-click the column, select "Run Column" > "Choose Number of Rows to Run" • Specify 2000 as your limit Alternatively, try Sandbox mode to test on a smaller subset of your data before running on all error rows.
Sources: Sandbox mode | Documentation | Clay University Conditional Formulas
Jaimin B.
·
·
I don't understand. Can you elaborate?
Channeled
APP
·
·
To run Claygent only on specific rows with errors: 1. Add a Row Number column using a number generator 2. Create a filter to identify only rows with errors 3. Right-click on your Claygent column 4. Select "Run Column" > "Choose Number of Rows to Run" 5. Specify 2000 as your limit Alternatively, try Sandbox mode to test your approach on a small subset before running on your error rows.
Sources: Sandbox mode | Documentation | Clay University Run Options Conditional Formulas
Jaimin B.
·
·
What is "Row Number" column?
Channeled
APP
·
·
A "Row Number" column is a column you can add to your table that assigns sequential numbers to each row. To create one: 1. Add a new column (scroll right and select "Add column") 2. Choose "Number" as the data type 3. Use a number generator formula to assign sequential numbers Once created, you can use this column to filter and target specific rows with errors by: 1. Applying a filter to show only rows with errors 2. Right-click your Claygent column 3. Select "Run Column" > "Choose Number of Rows to Run" 4. Specify your desired row count (2000) This lets you process only the error rows without rerunning successful ones.
Source: Table columns | Documentation | Clay University
Jaimin B.
·
·
Elaborate on step 3
Channeled
APP
·
·
Here's how to use a Row Number column to target error rows: 1. Add a new column and select "Number" as the data type 2. Use a number generator formula to assign sequential numbers to each row 3. Create a view to identify error rows: • Click the filters button at the top of your table • Add a filter to show only rows with errors 4. With error rows filtered, right-click on your Claygent column 5. Select "Run Column" > "Choose Number of Rows to Run" 6. Specify 2000 as your limit and set your starting row This runs Claygent only on the filtered error rows without reprocessing successful ones.
Sources: Run Options Customize your table view | Documentation | Clay University
Jaimin B.
·
·
https://app.clay.com/workspaces/3181/workbooks/wb_0sx7lb7WCrGPApHcFEP/tables/t_0sx7lu6g2SDnvjCzCBA/views/gv_0sx7lu7PfYUUnYmVMYT
Bo (.
·
·
Hey, the best way to handle this is to filter the column for rows with errors. Click the column header, apply a filter to show only rows with errors, then use Run > Save and run on rows that haven’t run or have errors. As long as you haven’t changed the column, it’ll only run on those filtered rows — not the full 5,000. Let me know if you need help setting it up.