Generate realistic fake data with the same shape and stats as your real data
Intermediate16h1–2
PythonDuckDBLLM
16/20Ask Your CSV in Plain English
Upload a CSV, ask a question and get a chart plus the SQL that produced it.
Intermediate14h1–2 people18/20 fit
Build Ask Your CSV in Plain English for analysts, founders and students in ~14h with 1–2 people. Stack: DuckDB + LLM + Charts + Next.js.
Why it wins: Showing the generated SQL builds trust and teaches at the same time.
Build flow
01 · Problem
People with data but no SQL skills cannot answer simple questions quickly.
02 · Solution
Load the CSV into DuckDB in the browser, have an LLM write SQL from the schema and question, run it locally and pick a chart type for the result.
03 · Build
- 1Load CSV with DuckDB-Wasm
- 2Send schema and question to the LLM
- 3Run the SQL and catch errors
- 4Render a chart and table
04 · Demo
Ask 'which month had the highest refunds?' on a real dataset and show the SQL.
MVP checklist
- Browser-side CSV loading
- Question to SQL
- Auto chart
- Editable SQL
Tech stack
| Tech | Role | |
|---|---|---|
| DuckDB | In-process analytics SQL | Data |
| LLM | Language model | AI |
| Charts | Data visualisation | UI |
| Next.js | Frontend + API routes | Web |
Reference repos
Open-source projects you can build on. Links checked 2026-10-31.
Related ideas
More AIExtract tables from messy PDFs into clean, checked CSV files
Intermediate14h1–2MarkItDownPythonLLM16/20See where in a call the customer got frustrated and why
Intermediate16h1–3WhisperLLMCharts16/20