Current inputs
Excel workbooks `.xlsx/.xls`, `.csv`, `.tsv` and `.parquet` files. They are the simplest starting point for business datasets you already have.
Platform
From spreadsheet to answer, without turning everything into indistinct text.
TalkToMyExcel treats tabular data as tabular data: rows, columns, filters, counts and descriptive text remain separate and controllable elements.
01
Upload the file
Excel, CSV, TSV and Parquet files are saved in staging and profiled before replacing or adding to the active dataset.
02
Choose what to index
The system suggests the descriptive columns most useful for semantic search, but the final choice remains explicit.
03
Ask the data
Questions about counts and status go through DuckDB; similar cases and descriptions go through Chroma.
04
Read a cited answer
The model generates the answer from compact evidence, not from the entire file pasted into a prompt.
Formats and scenarios
One entry point for exports, reports and spreadsheets.
Business use
After-sales, maintenance, tickets, quality, price lists, spare parts, CRM exports and periodic reports: different data, same question experience.
Hybrid engine
Precise numbers when numbers matter. Semantics when meaning matters.
An LLM alone can be brilliant but approximate. TalkToMyExcel uses it as a reasoning and synthesis layer, while filters, rows and counts stay handled by dedicated data components.
See the security boundaryDemo
Bring a concrete use case.
A real spreadsheet, a hard question, a result that can be checked.