While the model is working, the formula cell shows #GETTING_DATA. When the result is ready, the table spills into the cells below and to the right. You can keep using Excel while you wait.
EXTRACT returns a table with the columns you ask for. Use it to pull details from receipts, notes or customer feedback.
=XLLAMA.EXTRACT("Extract expenses", A2:A5, "vendor, date, amount")Syntax: =XLLAMA.EXTRACT(prompt, context, fields, [model])
- prompt: what records to extract.
- context: the source cells.
- fields: comma-separated column names, or a range containing those names.
- model: optional local model name.
Leave the cells below and to the right empty. The result spills into them, with headers in the first row. This requires Excel with dynamic arrays, such as Microsoft 365 or Excel 2021 or later. If Excel shows #SPILL!, clear the cells blocking the result.
Use up to 12 field names. An extraction returns at most 500 records. Check dates, amounts and missing values; the model can make mistakes. Answers are cached in the same way as PROMPT.
Use commas between arguments, or semicolons if your Excel uses them. The optional final model argument overrides the selected model. Select at most 1,000 input cells; large tasks work best in smaller batches. Review AI results before relying on them.
To regenerate a table, select its top-left EXTRACT formula cell and choose Settings → Refresh selected formula answers. Select the formula cell, not only its spilled results.