JSON
JSON Explorer
Expandable structure, search, paths, related values, inline edits, diff context, and focused inspection.
Real data handling, made repeatable.
Data WorkBench
Open the payload. Shape it. Prove what changed.
Data WorkBench is one bench for people working with the files they actually touch: JSON, XML, CSV, XLSX, graphics, exports, diagnostics, conversions, and the awkward parts between systems. It will not do everything. It is here to make most of the work less stupid.
- 100% your controlYour files, edits, exports, and save-back choices stay in your hands.
- Complete code transparencyThe public workbench is static code you can inspect in the repo.
- No harvestingAnalytics may measure site traffic, usage, and performance; WorkBench source data is not shared.
Try ANA WorkBench
Why WorkBench?
I needed this for real data handling, so I took Jigs I was already using and brought that repeatable workflow into the WorkBench.
A Jig is a saved ruleset for interpreting, shaping, validating, converting, and exporting a source. For an XLSX workbook, a Jig can remember the sheet, header row, data start, field names, included columns, and export shape.
The point is not novelty. It is cleaner routine work: open the payload, understand it, change it carefully, convert it when needed, and leave proof of what happened.
AI can still help where judgment matters: reasoning, mapping, explaining, reviewing, generating rules, helping design Jigs, and catching weirdness. The WorkBench handles the local routine parts directly.
| The usual mess | WorkBench approach |
|---|---|
| Ugly payload viewer | Cleaner Explorer, paths, related values, and readable structure. |
| Random formatter site | Local browser workflow without the paste-to-mystery-box feeling. |
| Spreadsheet chaos | Workbook, sheet, header, column rules, and reusable Jigs. |
| Silent conversion | Receipts, assumptions, warnings, and review prompts. |
| AI used for tiny formatting chores | Don't use AI to format. My god, it is using a particle accelerator to toast bread. Save a tree, damnit. We are not inventing fire here. Just use a better kitchen. |
| One-off cleanup | Saved rulesets and Jigs for repeatable work. |
| Human-only slog | Human and AI collaboration with clearer context. |
Free: SQL Lab2 beta
Play with your data right away.
Why load an XML into one thing, save a file, go to a prompt, ask it to create SQLite, save that, then go to another prompt just to ask basic questions? That is not a workflow. That is hand-cramp cosplay.
WorkBench still generates SQL and SQLite review artifacts. SQL Lab2 is
the free browser-local place to open generated .sql or
.db files, or bring in CSV, JSON, XML, TSV, and workbook
data directly, inspect tables, map schema, build joins, run review
queries, prepare DB scripts, and export results without uploading the
file.
Find the right doorway without leaving the bench.
Data to SQL
Files and archives
Proof and repeatability
The old way still works, if your hands hate you:
sqlite3 myhandsarecrampimg.sqlite <<'SQL'
.read generated-export.sql
PRAGMA integrity_check;
.tables
SELECT name
FROM sqlite_schema
WHERE type = 'table'
ORDER BY name;
SQL
sqlite3 myhandsarecrampimg.sqlite
Or try SQL Lab2 and review the same kind of artifact locally in the browser.
What Is Here So Far
Data WorkBench is built around work people already do every day: opening payloads, checking exports, fixing files, converting formats, reviewing changes, and trying not to leave a surprise for the next person.
Source trust cautions
Use known and trusted sources, but be precise about what trusted means. It is not just "I used it before" or "they are big." Trust should come from evidence: the official domain, direct navigation, a known owner, current documentation, and clear data handling.
Authorities give the same shape of advice: avoid links that arrive unexpectedly, verify contact details yourself, and prefer official websites or addresses you typed directly.
- USA guidance from FTC and CISA: familiar names and trusted-looking sources can still be impersonation.
- UK NCSC guidance: use details from official websites, not the suspicious message.
- Australian Cyber Security Centre guidance: type official web addresses and use legitimate contact details.
- Government of Canada guidance: check official domains and avoid convincing look-alikes.
Delimited
CSV / TSV / OtherSV
Delimiter choices, header review, row and cell editing, ambiguity warnings, formula-looking cells, and export.
XML
XML
Tree view, raw edit, attributes, namespaces, mixed content, entities, and no pretending XML is JSON.
Workbook
XLSX / Workbook Path
Workbook intake, sheet logic, header rules, column shaping, Jigs, export recipes, and Tab Tamer-style repeatability.
Output
Conversion / Export
Convert between formats, review assumptions and risks, then copy, download, or save back to a browser-approved local file when available.
Trust
Trust / Diagnostics / Receipts
Diffs, diagnostics, fidelity receipts, what changed, what may be risky, and review before export.
AI
Human / AI Workflow
Easier paths for humans, clean outputs for AI, and less wasting AI on tiny formatting chores.
Text
Text / Escape Tools
Inspect JSON strings, XML/HTML entities, URL encoding, invisible characters, line endings, and Unicode weirdness.
What It Is Becoming
Data WorkBench is growing into a place where user actions, source data, exports, and AI-assisted workflows can line up without adding theater. AIs love structure and clarity. So do people trying to ship clean data before lunch.
Local-first by default. Site analytics stay separate from WorkBench source data, and live connectors come later with explicit review.
-
Better working surfaces
Cleaner paths for the source data people already inspect, fix, reshape, and hand off.
-
Representation clarity
Raw text, semantic values, escaping, Unicode, and display behavior stay visible instead of getting politely mangled.
-
Receipts for change
Exports explain assumptions, warnings, loss notes, and review state before output leaves the bench.
-
Source-aware workflows
Files, browser captures, request/response payloads, and future connectors with explicit trust boundaries.
-
AI-ready structure
Cleaner data shape and clearer review context make future AI actions less magical and less hazardous.