The tools' API

Availability: in production since 20 October 2026. Uploaded-file jobs and download links follow separately.

The Matrix Tools pages stay as they are: they run in your browser, and the files you open there are never uploaded.

What it does

The API provides 50 Matrix Tools as 63 operations on our server, using the same engines as the tool pages. Send a tool's arguments and its files in one request. The answer holds the tool's result and every file the tool wrote. Each run gets a process of its own with no network access and its own memory limit, and the process ends with the run.

Encrypt a file with a passphrase, Decrypt a file with its passphrase and the CSR Generator are not offered. They exist so that a passphrase or a private key never leaves your device, so they run only in your browser.

Your files

• Files are sent inside the request, base64-encoded, up to 700 KB together.
• They are held in memory for the run only: the API stores them nowhere, and they are gone when the run ends. The files the tool writes come back in the answer and are not kept either.
• An error's message can name a file you sent. It goes back to you in the answer and is not logged.
• A run keeps nothing, so it takes no Idempotency-Key: sending a run again runs the tool again.
• Each call made with your API Key is counted for its usage, as /privacy/ describes: the action, the result (the HTTP status), the tool's name and the bytes sent and returned. The record never holds your files or their names.

Authentication

Send your API Key in the x-matrix-api-key header. A Key needs the Scope tools.run to run a tool, and tools.read to list the tools. A Workspace's Key runs in its Workspace. A person's own Key runs in their personal space, or in a Workspace they belong to when the body names it in workspace_id.

List the tools

GET /v1/tools answers every tool with its name, title and description, the JSON Schema of its arguments (inputSchema) and of its result (outputSchema), and which arguments name files (files).

Run a tool

POST /v1/tools/{name}/runs, with a JSON body:

curl -s https://api.matrixai.group/v1/tools/merge_pdf/runs \
  -H "x-matrix-api-key: $MATRIX_API_KEY" -H "content-type: application/json" \
  -d '{"arguments":{"files":["first.pdf","second.pdf"]},
       "files":[{"name":"first.pdf","content_base64":"JVBERi0..."},{"name":"second.pdf","content_base64":"JVBERi0..."}]}'

• arguments: the tool's arguments, as its inputSchema says. An argument marked (file) takes the name of one of the run's files.
• files: the run's files, each with a unique name and its bytes in content_base64. Names are case-sensitive, at most 255 characters, with no slashes, backslashes, surrounding whitespace or control characters; . and .. are refused.
• files[].last_modified is optional: the original modification time as integer milliseconds since the Unix epoch (0 through 253402300799999). Date-prefixed renaming and oldest/newest duplicate selection require it for every selected file. Without it, duplicate reports leave the date blank; ZIP entries use the run's time. Server dates use UTC.
• output, where a tool has it, names the file the tool writes; a folder in the name is dropped. A tool that answers inline writes its result as a file too when output is given, as its description says.
• outputDir and overwrite belong to a person's own disk and are refused. Every file a tool writes comes back, and a name that repeats gets a number, as on the pages.
• color_palette, bill_split and font_pairs write their files only into a folder. The API sets that folder itself, so their files come back too.

The answer, 200 (a real one, its bytes shortened):

{
  "data": {
    "tool": "merge_pdf",
    "summary": "Merged 2 PDFs into merged.pdf (2 KB), 8 pages.",
    "result": {
      "path": "merged.pdf",
      "pageCount": 8,
      "bytes": 2228
    },
    "outputs": [
      {
        "name": "merged.pdf",
        "size": 2228,
        "content_base64": "JVBERi0xLjcK..."
      }
    ]
  },
  "meta": {
    "request_id": "5a0f6a1e-6f0b-4d1c-9f6e-2b8d1c3e4f50"
  }
}

• summary: one line about the run, as the tool says it.
• text: the tool's text result, when it has one (such as the formatted JSON).
• result: the tool's result, as its outputSchema says. A file is named by its name in outputs.
• outputs: every file the tool wrote, with its size and its bytes in content_base64.

Limits

• A run has 30 seconds of wall-clock time, the same CPU-time limit, and a 256 MiB JavaScript heap. It is stopped past a limit.
• 700 KB of files in a request, under the API's 1 MB body.
• Your API Key's requests per minute.
• One active run per API Key family. A rotated replacement shares this limit with the previous Key. If a call disconnects or the runner's response is lost, this lock can remain until 40 seconds after the call started.
• Inline results: 20 MiB of output files, within a 32 MiB JSON response.

Errors

Every error answers { code, message, details, request_id }.

A 413 says in its message what was too large and the limit, for example:

• one file: The file "scan.pdf" is 701 KB. Files sent inline can be at most 700 KB together; larger files go through an upload (not available yet).
• files together: The 3 files sent come to 720 KB. Files sent inline can be at most 700 KB together; larger files go through an upload (not available yet).
• the request: This request is 2.7 MB, more than the 1 MB the API reads in one request. Files sent inline can be at most 700 KB together; larger files go through an upload (not available yet).

Tool reference

merge_pdf: Merge PDF

Merge two or more PDF files into one, in the order given. Writes merged.pdf unless output is given.

Result: path, pageCount, bytes.

split_pdf: Split PDF

Split one PDF into several files: every N pages (mode every, the default, with every: N), one file per comma-separated range such as "1-3, 4, 5-8" (mode ranges), or one file per page (mode each). Writes <name>_part1.pdf, <name>_part2.pdf, ..., and a ZIP of all parts when zip is true.

Result: files, zip, pageCount.

rotate_pdf: Rotate PDF

Rotate all or some pages of a PDF by 90 degrees clockwise (the default), 180, or 270 (counter-clockwise), on top of any rotation the pages already have. pages is "all" (default), "odd", "even", a list such as "1-3, 5", or an array of page numbers. Writes <name>_rotated.pdf.

Result: path, pages, pageCount.

organize_pdf: Organize PDF

Rebuild a PDF from a page plan: an array of page numbers, or of {page, rotate} objects (rotate 0, 90, 180 or 270), in the order the pages are to appear. Pages left out are dropped, each page can appear once, and at least one page must be kept. Writes <name>_organized.pdf.

Result: path, pageCount.

image_to_pdf: Image to PDF

Make a PDF from PNG or JPEG images, one page per image: at the image's own size (pageSize image, the page being the picture's size at 72, 150 or 300 dpi) or fitted on A4 or Letter with a margin. The pixels are embedded as they are; a phone photo stored sideways is turned upright. Writes <name>.pdf, images.pdf. WebP, GIF and BMP input, mirrored photos (EXIF orientation 2, 4, 5 or 7) and the web page's "Shrink large photos" option need the browser's canvas. Scope decision: no Node canvas in this version; these inputs are refused with needs-canvas and the error names the reason (the shrink option is not offered, so passing it is invalid-input). Convert such images to PNG or JPEG first.

Result: path, pageCount, turned.

inspect_pdf: Inspect PDF

Page count, each page's size in points and rotation, and whether the PDF has form fields. Reads the file only.

Result: pageCount, pages, hasForm.

describe_pdf_form: Describe PDF form

List the fillable fields of a PDF form: name, kind (text, checkbox, radio, dropdown, list), current value, options and whether the field is read-only. Call this before fill_pdf_form to learn the field names.

Result: hasXFA, fields.

fill_pdf_form: Fill PDF form

Fill the fields of a PDF form by name: a string for text, radio and dropdown fields, true or false for a checkbox, a string or an array of strings for a list. flatten burns the values into the page so the form can no longer be edited. The text a field shows is drawn in one font: Helvetica when it has every character, else Noto Sans CJK, embedded as a subset, all of it, digits and Latin letters too. Skipped and listed: text with a character neither font has (Thai, Arabic, an emoji) or that would need both fonts (a character only Noto Sans CJK has beside Š, š, Ž, ž, Ÿ, ˆ or ˜, which only Helvetica has), and read-only fields. Writes <name>_filled.pdf or <name>_flattened.pdf.

Result: path, written, skipped.

sign_pdf: Sign PDF

Place a PNG or JPEG signature image on one page of a PDF. Every fraction is of the page as viewers show it (its crop box), turned by the page's rotation: x and y are the top-left corner as fractions of its width and height (defaults 0.1 and 0.8), width the fraction of its width from 0.05 to 0.6 (default 0.25); the height follows the image's proportions on that page unless given. A box that would cross the right or bottom edge is moved in. This is a picture of a signature, not a certificate-based digital signature. Writes <name>_signed.pdf.

Result: path, page, placement.

describe_image_metadata: Describe image metadata

What a JPEG, PNG or WebP file carries besides its pixels: EXIF (camera, date, GPS location), XMP, IPTC, comments, text chunks, the colour profile, and the EXIF orientation. Reads the file only.

Result: format, items, orientation, removableBytes, summary.

remove_exif: Remove EXIF and metadata

Copy a JPEG, PNG or WebP image without its metadata; the compressed pixels are copied byte for byte, nothing is recompressed. The colour profile is kept unless keepIcc is false, and a sideways photo keeps a minimal orientation tag so it still displays upright. Writes <name>_clean.<ext>; a file with nothing to remove is not copied (path null) unless output is given.

Result: path, changed, format, removed, kept, removedBytes, keptOrientation.

json_format: JSON Formatter

Reprint JSON with an indent of 2 (default) or 4 spaces, a tab, or minified on one line (indent "min"), optionally with keys sorted. Every string and number is copied exactly as written. A syntax error is reported with its line and column. The text itself (text) or a file to read it from (file): exactly one of the two. The result is returned inline; results over 65536 characters, or when output is given, are written to a file.

Result: text, textTruncated, path, stats, summary.

text_diff: Text Diff

Line diff of two texts: a unified diff with three lines of context (empty when the texts are identical) and counts of added, removed and unchanged lines. Each side is given inline (original, changed) or as a file (originalFile, changedFile). Options: ignoreCase, whitespace exact (default), trailing or all. Diffs over 65536 characters, or when output is given, are written to a file; with output, identical texts give an empty file.

Result: identical, stats, unified, unifiedTruncated, path, timedOut.

regex_tester: Regex Tester

Test a JavaScript regular expression on a text: every match with its index, end, groups and named groups (the first limit matches are returned, default 100), the capture group names, and, when replacement is given, the text with the replacement applied ($1, $<name>, $& and $$ as usual). The text itself (text) or a file to read it from (file): exactly one of the two. A replaced text over 65536 characters, or when output is given, is written to a file.

Result: pattern, flags, matchCount, truncated, matches, groupNames, replaced, replacedTruncated, path.

markdown_preview: Markdown Preview

Render Markdown to HTML with the web page's renderer: headings with ids, lists, task lists, tables, fenced code (in colour in the languages the page names), footnotes, blockquotes, links and images; raw HTML is shown as text. Returns the body HTML, the title, the headings and counts. The text itself (text) or a file to read it from (file): exactly one of the two. With output, a standalone HTML document is written; HTML over 65536 characters is written too and cut inline.

Result: title, toc, stats, html, htmlTruncated, path.

unit_convert: Unit Converter

Convert a value between two units of one quantity (length, area, volume, mass, temperature, speed, pressure, energy, power, data, time, fuel economy, ...), with the exact factor in words and the value in every unit of the quantity. Call unit_categories for the quantity and unit ids.

Result: value, result, line, explanation, table.

unit_categories: Unit categories

The quantities and unit ids unit_convert accepts.

No arguments.

Result: categories.

date_calc: Date Calculator

Calendar arithmetic without time zones. mode diff: the days from start to end (the start date counts, the end date only with includeEnd) with weeks, months, years and working days. mode add: start moved by years, months, weeks, days and workdays (whole numbers up to 1,000,000; direction subtract reverses them). mode facts: weekday, day of the year, ISO week, quarter, leap year, days in the month and Unix time of a date. Dates are YYYY-MM-DD. Every result has the page's headline line and its rows.

Result: mode, line, rows.

timezone_compare: Time Zone Compare

The same moment in several time zones: a date (YYYY-MM-DD) and time (HH:MM) read in the reference zone, with each zone's local time, date, offset, difference from the reference and abbreviation, and an hour-by-hour planner of the reference day. A time inside a daylight-saving gap is moved forward, a time that happens twice takes the first occurrence; note says so. Zone names are IANA names; call time_zones for the list.

Result: line, note, epochMs, shifted, ambiguous, rows, planner.

time_zones: Time zones

The IANA time-zone names this server knows, sorted.

No arguments.

Result: zones.

extract_pages: Extract pages

Pull the chosen pages out of a PDF into a new file, in document order. With includeRest, the pages left over are written as a second file. Writes <name>_pages.pdf, and <name>_remaining.pdf.

Result: path, pages, pageCount, restPath.

page_numbers_pdf: Add page numbers

Stamp page numbers on a PDF. Page p is numbered start + p - 1, so numbering can begin at any figure. The text sits inside the crop box and follows the page's rotation so it reads upright. Format custom takes your own template ({n} the number, {p} the total). The labels are written in Helvetica when it has every character in them; otherwise, as for Ā {n} or 第 {n} 页, in Noto Sans CJK, embedded as a subset, digits and Latin letters too, except a few characters, such as Š, that stay in Helvetica. Writes <name>_numbered.pdf.

Result: path, count, total, firstLabel, unnumbered, pageCount.

watermark_pdf: Add a text watermark

Stamp text across the chosen pages of a PDF, diagonally, horizontally through the centre, or tiled over the whole page. Writes <name>_watermarked.pdf.

Result: path, count, pageCount.

crop_pdf: Crop PDF

Trim margins off the chosen pages of a PDF. The margins are given in millimetres from each edge of the page as viewers show it (its crop box), turned by the page's rotation; nothing is re-rendered, the visible area is moved in. The web page can also measure the white margins and trim automatically, which reads the pages as pictures and is not available here. Writes <name>_cropped.pdf.

Result: path, count, pageCount, margins.

compress_pdf: Compress PDF

Make a PDF smaller by tidying its structure: streams are recompressed and objects nothing points to are dropped. The pictures are left byte for byte as they are. The web page's balanced and small presets, which re-encode pictures, need the browser's canvas and are refused here. When nothing is gained the original bytes are written back, as on the page. Writes <name>_compressed.pdf.

Result: path, before, after, saved, changed, preset.

describe_pdf_metadata: Describe PDF metadata

What a PDF says about itself besides its pages: the document information entries such as author and producer, the XMP stream, the file identifiers, private application data and the authors and dates on annotations. Reads the file only.

Result: pageCount, info.

clean_pdf_metadata: Clean PDF metadata

Remove what a PDF says about its author, tools and history, keeping the pages untouched. Each part can be kept: the information entries, the title within them, the XMP stream, the file identifiers, private application data and annotation authors. Writes <name>_clean.pdf.

Result: path, removed, remaining.

json_schema_validate: JSON Schema Validator

Check a JSON document against a JSON Schema (draft 2020-12 and draft-07 keywords) and report every error with the path to the value that broke the rule. Each side is given inline (schema, data) or as a file (schemaFile, dataFile). Numbers are compared exactly as written, and $ref follows JSON Pointers, $anchor and $id within the schema. With formats, the format keyword becomes an assertion for date, time, date-time, email, uuid, ipv4, ipv6, hostname, uri, uri-reference and regex.

Result: valid, complete, unchecked, errors, notes, truncated.

mock_data: Mock Data Generator

Generate synthetic rows that look real and repeat exactly for the same seed, for testing without real people's data. Give either a ready-made preset (people, orders, products, events) or a list of fields, each with a name, a type and options. The types are sequence, uuid, firstName, lastName, fullName, email, username, phone, company, jobTitle, street, city, country, countryCode, postcode, date, datetime, integer, decimal, boolean, enum, words, sentence, paragraph, url, ipv4, hexColor. Returned inline, and written to a file when output is given or the result is long.

Result: text, textTruncated, path, seed, rowCount, fields.

cron_explain: Cron Expression Explainer

Say in words what a cron expression means and when it runs next, in a named time zone. Five fields (minute hour day-of-month month day-of-week) or six with a leading seconds field; the usual macros (@daily, @hourly, @weekly, @monthly, @yearly, @reboot) are understood.

Result: expression, sentence, fields, timeZone, runs, exhausted.

csv_sql: SQL on CSV

Run a SQL query over one or more CSV files held in memory: SELECT with WHERE, GROUP BY, HAVING, ORDER BY, LIMIT, JOIN (also LEFT, RIGHT and FULL), sub-queries, WITH (also recursive), UNION, INTERSECT and EXCEPT, window functions and the usual aggregate and string functions, with SQLite's rules. Column types are read from the data, and numbers keep their exact decimal value. Each table is a file or inline CSV with a name to use in the query.

Result: columns, rows, rowCount, csv, csvTruncated, path, stats, tables.

cert_decode: Certificate Decoder

Read X.509 certificates and certificate requests and say what is in them: subject and issuer, validity, key, signature algorithm, extensions, the alternative names and the fingerprints. Several PEM blocks are read at once and ordered into chains, with each link checked against the one above it. Give PEM text or a DER file.

Result: items, privateKeys, errors, chains, unrelated.

csv_to_json: CSV to JSON

Convert CSV, TSV or other delimited text to JSON: an array of objects (one per row, the default), JSON Lines, an array of arrays, or one array per column. The delimiter is detected unless given. Numbers and true/false are converted and empty cells become null unless types is false; numbers with leading zeros, integers over 15 digits and any number a JSON number cannot hold exactly (123456789012345.68) stay text, so codes and amounts are never altered. Each column keeps one type: where a value of a column stays text, its numbers and true/false stay text too, as typed, and a warning names the column. The text itself (text) or a file to read it from (file): exactly one of the two.

Result: text, textTruncated, path, stats, warnings, summary.

csv_clean: CSV Cleaner

Clean a delimited table: trim cells, remove invisible and control characters, fix header names, even out row lengths, remove blank and duplicate rows, and save with a chosen delimiter and line ending. Every step can be switched off. Nothing is guessed about the values themselves: no number or date is rewritten. The text itself (text) or a file to read it from (file): exactly one of the two.

Result: text, textTruncated, path, stats, changes, unchanged, summary.

compare_sheets: Compare Sheets

Compare two versions of a delimited table. Rows are matched by one or two key columns (header names that exist in both tables) or, without a key, by their position; columns are matched by header name. Returns the changed cells, the added and removed rows, and a CSV report with one line per changed cell and per cell of every added or removed row. Each table is given inline (a, b) or as a file (aFile, bFile).

Result: identical, headline, summary, stats, changed, added, removed, warnings, report, reportTruncated, path.

reconcile_sheets: Reconcile Sheets

Reconcile two tables that should describe the same records, such as a ledger and a bank statement. Rows are matched by key (one or two columns, named per table, so the names may differ); with an amount column on both sides the amounts of each key are added up and compared exactly, as decimals, never as floating point. Returns what matches, what differs, what is missing on either side, rows without a key, and possible matches among the leftovers, which count as matched only when their ids are passed in accepted. Writes the full report and the exceptions as CSV on request. Each table is given inline (a, b) or as a file (aFile, bFile).

Result: headline, summary, counts, compared, groups, noKey, suggestions, warnings, exceptions, exceptionsTruncated, path, exceptionsPath.

describe_csv_columns: CSV Anonymizer: what the columns hold

Read one or more CSV, TSV or JSON tables and say what each column looks like (e-mail address, phone number, person's name, identifier, date, IP address, number or text), with the evidence and the action csv_anonymize would take by default. Call it first and check the guesses: only the data's owner knows what a column means.

Result: files, columns.

csv_anonymize: CSV Anonymizer

Replace what looks personal in one or more CSV, TSV or JSON tables with made-up values, using one mapping for the whole call: the same original gets the same replacement in every file, so keys still join. Columns are handled by what they look like unless actions names an action for a column: keep, email, phone, name, id, ip, hash, generalise-email, generalise-phone, generalise-name, generalise-id, month, shift, remove. Dates can be shifted by one offset for the whole run. Writes <name>_anon.<ext> for every file and, unless report is false, anonymize-report.txt with counts and rules, never a value. This is pseudonymisation, not a guarantee of anonymity: read the residual findings. The key of the mapping is drawn at random for every call; a repeatable run from a passphrase is offered by the library function and the web page only, because a passphrase should not travel through a model's context.

Result: files, report, zip, columns, mapped, residual, findings, offset, summary, note.

zip_create: Create ZIP

Pack files into one .zip. Every file is stored under its own name (no folders); a name already used gets a number, as "report (2).pdf". With deflate (the default) a file is compressed unless its first 64 KB show it would not shrink, as photos, videos and office documents do not; store packs everything as it is. Each entry keeps the time its file was last changed. Writes archive.zip unless output is given.

Result: path, bytes, entries, renamed, summary.

zip_list: Extract ZIP: what the archive holds

List a .zip without taking anything out: every entry with its path, size, compression method and the time it was last changed, and whether it can be extracted here (password-protected entries and methods other than deflate and stored cannot, with the reason). ZIP64 archives are read.

Result: entries, files, folders, comment, zip64, totalSize, totalCompressed.

zip_extract: Extract ZIP

Take files out of a .zip and verify the CRC-32 of each: all files that can be extracted, or the ones named in entries by name, path or index (see zip_list). Every file is written with the folders it had in the archive in its name (docs/report.txt as docs - report.txt), as the page's downloads are named, and a name still alike gets a number; an entry that fails is reported and the others are still written.

Result: files, failures, summary.

plan_rename: Batch rename: the preview

Say what new name each file would get under renaming rules, and why a name could not be used (a character Windows does not allow, two files ending with the same name). Nothing is read or written. Rules are applied in this order: find and replace in the name without its extension, change of case, text before and after, numbering in list order, the file's modified date in front, and the extension.

Result: plan, changed, problems, ruleProblem, ready, summary.

batch_rename: Batch rename

Make renamed copies of files under renaming rules (see plan_rename for the rules and the preview). The files themselves are never renamed, moved or deleted: each file whose name changes is written under its new name, and with zip the page's renamed-files.zip is written as well, holding every file, renamed or not. Refused when the preview shows a problem.

Result: files, zip, plan, changed, summary.

find_duplicates: Find duplicate files

Find byte-identical files among the files given: files of the same size are compared by SHA-256, so files that are alone in their size are never read twice. Returns the groups, which copy of each to keep under the rule (the first in the list, the oldest or newest by modified time, or the one with the shortest path), and the page's CSV report. Nothing is deleted: the report says what could be removed.

Result: groups, hashed, redundantFiles, redundantBytes, report, reportTruncated, path, summary.

split_file: Split a file

Split one file into numbered parts of partSize bytes (<name>.001, <name>.002, ..., the naming of HJSplit and 7-Zip volumes) and write <name>.sha256 with the SHA-256 of the whole file, so a join can be verified. The parts are written.

Result: parts, checksum, hash, summary.

join_files: Join the parts of a file

Join parts named <name>.001, <name>.002, ... back into the file, in the order of their numbers whatever the order given. With the .sha256 file among the files the result is verified against it, and nothing is written when it does not match. Writes <name> unless output is given.

Result: path, bytes, hash, verified, summary.

pdf_to_text: PDF to text

Read the text of a PDF's pages, in reading order. layout paragraphs (the default) joins the lines of a paragraph and leaves a blank line between paragraphs; lines gives one line of the page per line of text. Each page starts with a line such as "---- Page 3 ----" unless pageBreaks is false. pages is "all" (default), "odd", "even", a list such as "1-3, 5", or an array of page numbers. A scanned PDF holds pictures of pages and no text: it is refused (no-text), there is no OCR here. Writes <name>.txt when output is given or the text is long.

Result: text, textTruncated, path, pages, pagesWithText, characters, summary.

subtitle_shift: Subtitle Shift

Retime SRT or WebVTT subtitles: move every cue by offsetMs milliseconds (negative is earlier; a cue that would start before 0 is clamped to 0), stretch the times for a change of frame rate (from and to), and write SRT or WebVTT. Blocks that cannot be read are left out and counted. The text itself (text) or a file to read it from (file): exactly one of the two.

Result: text, textTruncated, path, format, inputFormat, cues, clamped, skipped, warnings, problems, headline, summary.

color_contrast: Colour contrast checker

The contrast ratio of a text colour on a background colour, which of the WCAG 2.2 contrast criteria it meets (AA and AAA for normal and large text, and 3:1 for user interface parts), and for each criterion it fails the nearest text colour and the nearest background colour that would pass: the same hue and colourfulness in OKLCH, only the lightness moved. Colours are given as hex, rgb(), hsl() or a CSS colour name. The verdict is taken from the exact ratio, before it is rounded for display.

Result: text, background, notes, ratio, ratioText, checks, passed, summary, suggestions, css.

color_palette: Colour palette generator

Build a palette from one colour: a tonal scale from 50 to 950 with the colour at its nearest step unchanged, and the companions of a harmony (complementary, analogous, triadic, split, tetradic), all computed in OKLCH. Every swatch names the text colour, black or white, that reads better on it and its WCAG ratio. Returns the palette as CSS custom properties, JSON and an SVG sheet, and writes the three files.

Result: color, note, harmony, scale, companions, css, json, svg, files.

qr_code: QR Code Generator

Make a static QR code and write it as SVG and PNG. content says what the code does: {type: "url", url, utm} opens a web address (https:// is added when no scheme is given, campaign parameters are appended from utm), {type: "text", text}, {type: "wifi", ssid, password, security, hidden} joins a network, {type: "email", to, subject, body}, {type: "phone", number}, {type: "sms", number, message}. The code is static: what it holds is printed into it and cannot be changed later. The PNG gets whole pixels per module and prints at exactly widthMm. Scan the result once before printing many.

Result: svg, png, payload, summary, version, modules, level, levelRaised, print, notes.

parse_vcard: vCard Maker: open a .vcf file

Read the first contact of a .vcf file (vCard 2.1, 3.0 or 4.0) as a card that vcard_maker takes, and name what the file holds that a card does not, so nothing is lost without a word. The text itself (text) or a file to read it from (file): exactly one of the two.

Result: card, version, count, ignored.

vcard_maker: vCard Maker

Make a contact's .vcf file (vCard 3.0, the most compatible, or 4.0), a QR code that adds the contact to a phone, and an email signature as HTML and plain text, and write them. card holds name {given, family, prefix, additional, suffix}, formatted, org, department, title, phones, emails and urls (each a list of {type, value} or of plain values), address {street, locality, region, postcode, country, type}, note and photo {mime, base64}. The photo goes into the file only, never into the QR code. Each output stands on its own: a contact too long for a QR code still gets its file and its signature, and the result says why the code is missing.

Result: name, vcf, vcard, qrSvg, qrPng, signatureHtml, signatureText, signature, problems, notes.

bill_split: Bill Splitter

Work out who owes whom after shared expenses, in the fewest payments. All sums are done in whole minor units (cents), so every split adds up to its expense exactly. Pass the split itself (split) or a file the page or this tool saved (file): exactly one of the two. split holds title, currency (such as EUR), decimal ("." or ",": how its amounts are written), people (their names) and expenses: each with description, amount, payer (the position of the person who paid, from 0), method (equal, exact, percent or shares), take (for equal: who shares it, true or false per person; everyone when left out) and values (for the other methods: the amount, percentage or share count of each person, in the order of people). An expense without an amount is not counted. The CSV and the split as a file are written too.

Result: headline, proof, summary, currency, decimals, total, people, transfers, balances, expenses, provenFewest, notCounted, notes, csv, csvPath, savedPath.

print_templates: Print templates

Make printable paper, a calendar or a label sheet as a vector PDF: lined, handwriting rows, square grid, dot grid, graph paper, tian zi ge writing boxes, music staff, Cornell notes, a month or a whole year, or the outlines or the text of a sheet of labels. Give only the options that belong to the template: one that the web page does not show for it is refused. The PDF carries no date, so the same options give the same bytes. Print at 100 % (actual size), not fit to page.

Result: path, pages, title, summary, warnings, counts.

resume_to_pdf: Resume Builder

Make a resume as a one-column PDF with real text that copies in reading order, clickable email and web links and no watermark, from a JSON Resume (jsonresume.org). Pass the resume itself (resume) or a file (file): exactly one of the two. basics (name, label, email, phone, url, summary, location, profiles), work, education and skills fill the resume; projects, certificates, awards, languages, volunteer, publications and interests become sections of their own; references are not taken. A file saved by the web page or by this tool opens exactly as it was, with its layout, headings and order of sections. The PDF carries no date, so the same resume gives the same bytes.

Result: path, pages, paper, summary, warnings, notes, layout, json.

tag_desk: Label Maker

Make labels for a label sheet from a list, as a PDF: text with {{Column}} fields and an optional Code 128, EAN-13 or QR code per label, all as vector shapes. The list is CSV or cells copied from a spreadsheet, given as text or as a file (at most one of the two); without a list the labels are all the same apart from the running number {{#}}, and print.count says how many. designFile takes a design saved by the web page or by this tool; header, sheet, design and print of the call replace those of the file. A row that cannot be printed as asked (a wrong EAN check digit, a character the font cannot draw) stops the PDF unless print.leaveOut is true. Print at 100 % (actual size), never fit to page; tag_desk_calibration makes a page to check the sheet against.

Result: path, pages, title, summary, sheet, columns, delimiter, counts, leftOut, problems, notes, design.

tag_desk_calibration: Label Maker: calibration page

Make the calibration page of a label sheet: the outline of every label, moved by the offsets, and a horizontal and a vertical reference line, 100 mm long where the paper has room (on smaller paper the longest of 50, 20 or 10 mm that fits, its length printed beside it). Printed on plain paper at 100 % and held against a label sheet, it shows whether the printer scales and by how much the print must be moved.

Result: path, sheet, summary, marks.

describe_docx_template: Mail Merge: read a template

Read a Word template (.docx or .dotx) before a mail merge: its fields ({{Name}} placeholders and Word merge fields) and where they stand, its conditions, and the problems that must be fixed in Word first. Call it to know which columns a list needs.

Result: name, size, summary, fields, conditions, mergeFields, usesRowNumber, isTemplate, problems.

merge_docx: Mail Merge

Fill a Word template from a list: one Word file per row (output files), or one document with every row, each on a new page (output combined). The list is CSV or cells copied from a spreadsheet, given as text or as a file (exactly one of the two); its first row names the columns. A field takes the column of its own name (case, spaces and underscores do not matter); mapping changes that per field: {"Company": {"column": "Firm"}} takes another column, {"Company": {"column": null, "fallback": "ACME Ltd"}} takes no column and writes the text, {"Grade": {"fallback": "n/a"}} writes the text where the cell is empty. A field without a column and without a text stops the merge. checks in the result names the rows with empty fields, values a format could not read and unusually long values.

Result: output, files, list, zip, path, summary, notes, rows, columns, mapping, checks.

csv_to_chart: CSV to Chart

Draw a table as a column, bar, line or area chart and write it as SVG. The first row of the table names the columns; one column gives the category names and up to eight columns of numbers give the series. Left to itself the tool reads the table as the web page does: labels from the first column of dates or of text, the first four columns of numbers as series, and the chart type that suits the table (a line for dates, bars for many or long names, columns otherwise). The result says what was read, and lists cells that are not numbers and were left out. The text itself (text) or a file to read it from (file): exactly one of the two.

Result: path, heading, title, description, type, suggested, stacked, width, height, summary, notes, label, labels, series, columns, delimiter, decimal.

font_pairs: Font pairs

Make the CSS and the JSON design tokens of a type system: a heading font, a body font and a modular scale. Each font is one of: {stack: id} for a system font stack that needs no download (system-sans, helvetica, verdana, trebuchet, gill, georgia, times, palatino, garamond, baskerville, mono, cjk-sans, cjk-serif), {family: name, category} for a font by name with its generic fallback, or {file: path} for a font file, which is read for its family name and axes only and is never copied into an export. The two files are written too.

Result: css, tokens, heading, body, scale, notes, files.

svg_pattern: SVG pattern maker

Make a seamless repeat tile and a 1024 px square filled with it, as SVG, from a built-in shape (dot, ring, square, diamond, triangle, hexagon, cross, star, stripe) or from a motif of one's own (an SVG or a PNG file). Whatever crosses one edge of the tile comes back on the opposite edge, so the tile repeats without a seam.

Result: tile, fill, width, height, cell, placements, shape, layout, color, background, summary.