sqlexplore 0.6.0

SQL exploration over data files with your CLI and LLM

sqlexplore

Get answers from messy data files fast, without building pipelines first.

With sqlexplore, you can:

• Inspect new datasets in minutes instead of writing setup scripts.
• Ask ad hoc questions with SQL and get immediate feedback.
• Move from quick profiling to exportable results in one terminal workflow.

sqlexplore is a terminal SQL workbench for flat files. Point it at a local or remote dataset, then explore with DuckDB SQL in an interactive TUI or run one-shot queries in standard CLI mode.

Useful features

• Works with .csv, .tsv, .txt, .parquet, and .pq.
• Accepts local paths and http(s) URLs for supported file types.
• Accepts piped stdin text (omit data when piping) and opens in TUI by default. If no controlling tty is available, it falls back to --no-ui.
• Interactive TUI with query editor, results grid, preview pane, and activity log.
• SQL + helper commands for common analysis tasks: /summary, /profile, /hist, /corr, /dupes, /top, /group, and more.
• Context-aware autocomplete for SQL clauses and helper command arguments.
• Query history and rerun support (/history, /rerun), plus editor helpers (/last, /clear).
• Export last result to .csv, .parquet/.pq, or .json with /save.
• JSON-aware rendering in result cells and clickable links in preview.
• Image bytes in BLOB or STRUCT{bytes,path} cells render as compact [img ...] tags with metadata in preview.
• Non-interactive mode via --no-ui (optionally with --execute or --file).
• Remote download controls: custom directory (--download-dir) and overwrite behavior (--overwrite); existing local downloads are reused by default.
• .txt files are ingested line-by-line with derived metrics (line_number, word_count, line_hash, etc).

Usage examples

Open a local file in the TUI:

sqlexplore ./data/example.parquet

Open a remote dataset URL (downloaded first, then loaded):

sqlexplore https://github.com/dylanhogg/awesome-python/raw/refs/heads/main/github_data.parquet

Run one SQL query and exit:

sqlexplore ./data/example.parquet --execute "SELECT COUNT(*) AS n FROM data" --no-ui

Run SQL from a file and exit:

sqlexplore ./data/example.parquet --file ./queries/report.sql --no-ui

Run default sample query in plain terminal output (no TUI):

sqlexplore ./data/example.csv --no-ui

Analyze piped terminal text:

ls -lha | sqlexplore

Analyze piped terminal text in plain terminal output (no TUI):

ls -lha | sqlexplore --no-ui

Pipe text and open TUI with a startup SQL query:

ps aux | sqlexplore - --execute "SELECT line FROM data WHERE line ILIKE '%python%'"

Control remote download location and overwrite:

sqlexplore https://github.com/dylanhogg/awesome-python/raw/refs/heads/main/github_data.parquet --download-dir ./data/cache --overwrite

Typical helper commands in the editor:

/limit 5000
/rows 5000
/summary
/profile amount
/top category 10
/hist amount 20 | amount > 0
/corr tip_amount total_amount
/dupes order_id
/save ./out/results.parquet

/limit <n> sets both helper query limit and row display limit.
/rows <n> only sets row display limit.

Show installed version:

sqlexplore --version

Install

Requires Python 3.13+.

pip install sqlexplore

or:

uv tool install sqlexplore

Links

• GitHub
• PyPI