📊 Structured Data Inspection Workbench

Profile large structured files at the existing Log File Analyzer URL: detect formats, inspect columns and malformed records, search rows, and export bounded evidence.

Browser-local: files are read in chunks and never uploaded. Maximum input: 10 MB. Profiles, evidence, previews, and exports are deliberately bounded.

Package metadata-only findings in the Context Workbench with Log Triage or Structured Data Quality Packs. Raw log lines, rows, and cell values are excluded.

Load a local .csv, .tsv, .json, .jsonl, .log, or .txt file

Data to inspect

id,name,active,score,note
1,Ada,true,9.5,"quoted, value"
2,Grace,false,,missing score
3,Linus,true,7,ok
broken,row

FormatDelimiter override First row is a headerSearch regexContext lines

Declarative validation rules (JSON, maximum 50)

[
  {"field":"id","required":true,"type":"number","unique":true},
  {"field":"active","type":"boolean"},
  {"field":"score","min":0,"max":10}
]

Supported rules: required, type, enum, regex, min, max, and unique. Violation evidence reports only field, row, and typed code—never raw cell values.

Bounded previewColumn/type/null profileMalformed-record evidenceRule-violation evidenceInspection report

Metadata-only Log Triage / Structured Data Quality report

Runs in your browser

Files and pasted data stay in this browser and are never uploaded. Redact sensitive values before copying reports or filtered evidence.

Include current input in shared/debug payload

What this tool does

Inspects large CSV, TSV, JSON, JSON Lines, and plain logs incrementally, then reports format confidence, row counts, column types, null frequency, malformed-record evidence, bounded previews, and search results.

Why it is useful

It gives humans and agents a fast local data-quality pass without sending datasets or production logs to a server.

Bounded by design

Input is capped at 10 MB, malformed evidence at 100 records, previews at 200 lines, and filtered exports at 1,000 matching rows. Every truncation is explicit.

Detection and search semantics

Auto detection reports confidence and can be overridden. Search uses a case-insensitive regular expression; invalid expressions and delimiters return typed errors.

Privacy note

File loading and inspection run locally in the browser.

Important limitation

Detection and inferred types are inspection hints, not a schema guarantee. Keep the original file and review malformed evidence before transforming data.

Quick answers

Are files uploaded?

No. Files are read in bounded chunks and inspected in your browser.

Why is the preview truncated?

Preview, malformed evidence, and exports have explicit limits so large inputs remain responsive.

Common Use Cases

Profile a vendor CSV

Detect the delimiter, inspect null and mixed-type frequency, then review inconsistent row widths.

Check a JSONL dataset

Find malformed lines, search matching objects, and export a bounded JSONL subset.

Summarize a deploy incident

Choose Plain log to retain severity, key-frequency, context-window, and long-line inspection.

Export timeout evidence

Search for timeout, inspect bounded context windows, then export matching lines.

View all in Developer →