Ask questions to your CSV files
Sometimes the data lives in a single CSV. Data Talks lets you upload it, ask questions in plain language, and get charts back — without spinning up a database.
CSV uploads in Data Talks aren’t a toy feature. The agent profiles the file (column types, null rates, distinct values) on upload and exposes it as a queryable source. Behind the scenes the file becomes a DuckDB-backed table, so even multi-million-row CSVs answer in seconds.
This makes Data Talks the fastest way to interrogate ad-hoc exports — vendor reports, support audit dumps, marketing exports — without the back-and-forth of building a one-off database.
Upload a file in 10 seconds
- 1
Drag a .csv or .xlsx file into the source picker.
- 2
Data Talks profiles the file — column types, null rates, sample rows.
- 3
Confirm column types if you want to override anything.
- 4
Open a conversation and start asking questions about the file.
Things you can ask of a CSV
- › What’s the distribution of order_amount, and which 10 rows are outliers?
- › Group by status and count, then show as a pie chart.
- › Filter to rows where customer_country is BR and source = paid; what’s the median ticket size?
- › Plot daily trend of created_at by week.
Why CSV in Data Talks beats Excel filters
Multi-million-row files supported via DuckDB-backed storage.
Auto-profiling: column types, null rates, distinct counts on upload.
Same conversational UX as databases — no syntax to learn.
Pin charts from a CSV to a dashboard you share with your team.
Combine CSV with database sources in one conversation.
Files stay in your self-hosted instance — no third-party upload.
CSV FAQ
What’s the file size limit?
Practically: hundreds of MBs to a few GBs depending on host RAM and disk. DuckDB handles the heavy lifting.
Does it support XLSX with multiple sheets?
Yes. Each sheet becomes a separate table inside the same source.
Can I refresh a CSV after re-uploading?
Yes. Replace-and-refresh keeps the source ID, so existing dashboards keep working with new data.
What about CSVs in S3 or GCS?
Connect S3 / GCS as a source and the agent treats matching CSVs as queryable tables.
Drop a CSV and start asking
Open source, self-hosted, free. The fastest path from spreadsheet to insight.
Get Started