II.3 Tasks, options and results
A task is a promise with three parts
When you press a run button, Documentize creates a task. Think of a task as a work order handed to a specialist:
INPUT OPTIONS RESULT
your file(s) + a few choices -> new file(s)
(never changed) (defaults ready) (kept for a while)
The specialist (a worker dedicated to that app) takes the order from a queue, does one job, and puts the finished files on a shelf for you to collect. That is also why heavy jobs show progress instead of freezing the page, and why the same work can be driven from a script or an AI assistant: it is just a task with a number.
Options: required, optional, defaulted
Options come in three kinds.
| Kind | Example | What to do |
|---|---|---|
| Required to start | Unlock needs the password. Sign needs a .pfx certificate and its password. Delete Page needs the pages to delete. Add TOC needs the entries you type | The app cannot guess. It will say what is missing (for example "Type the password, please.") |
| Choices with a default | Compressor level (Medium pre-selected); OCR document language (English); Abstracts style | Leave alone until the result is not what you wanted |
| Switches that add work | OCR while merging; AI correction in Searchable; Recovery's sanitize options | Off by default. Turn on when you know why |
The options panel is folded away behind Show all settings on purpose. A good default means a plain press of the button is a sensible first try.
"What changes if I change this?"
Ask this question about every option. A few real examples from the chapters show how different the answers can be:
| App | Option | What actually changes |
|---|---|---|
| Compressor | Low / Medium / High | Only the resolution ceiling for pictures (150, 120 or 96 DPI). Text stays sharp. Higher means smaller and softer |
| Merger | Mode Sequential / Grid | How images are laid out. PDFs still follow one after another |
| Searchable | Document language | Which alphabet and words the recogniser expects. The wrong one is the commonest cause of gibberish |
| Abstracts | Style, length, language | What the model is told to write. Sentence counts are instructions, not guarantees |
| Protector | Password | Who can open the file. There is no way to read it back from the result |
When an option is risky or irreversible, the app chapter says so under "Options explained" and "Common mistakes & fixes".
What you get: results
A result is new files. Your original is never modified by any app; this holds across every family chapter.
Typical shapes of result:
| Shape | Examples |
|---|---|
| One PDF | Merger, Rotation, Flatten, Add TOC |
| Many files in a ZIP | Splitter, Compressor with several inputs, Converter with several inputs, Parser |
| A different format | Extract Text (.txt), Extract Form Data (CSV), PDF Knowledge Base (Markdown ZIP) |
| A report | PDF/A Validator, Recovery (repaired PDF plus report) |
| A verdict on screen | Verify Signature |
| A conversation | Ask PDF (cited answers), Assembly Chat |
| A pull request | Repo Docs, Refactor Assistant, NuGet Updater (in apply mode) |
Naming
Output names usually follow the original with a suffix, or a ZIP is named after the operation. Observed examples: …_optimized.pdf, …_cheat_sheet, Compressed_documents, Converted_<from>_to_<to>. The exact name can vary by app; check the "What you get" section of the chapter.
What the result tells you
Some apps give you numbers: Compressor shows original size, new size, bytes saved and percentage, and honestly reports small or even negative savings. Others show what they did in steps. Read the summary before you build on the file.
Results are partial sometimes
On the free plan, a long job may stop at a time limit. You still get a real result for the files that were done, plus a way to finish the rest. See II.4.
Redo, don't patch
Because every run starts from your original and creates a new file, the cheapest fix for a bad result is to run again with different options, from the original. Do not push a lossy result through the same lossy app a second time. Compressing twice is not better, only blurrier (Compressor).
Task numbers, other channels
The same task has a number behind it. In the API, a task id is how you ask "is it ready?" and then download; the MCP and CLI channels expose the same idea. Details are in Beyond the browser. In the web app you rarely see the number, but it is why results can be fetched and deleted individually.
Try it
Run Compressor twice on the same PDF, once at Low and once at High, each time from the original. Compare file sizes and zoom to 200% on a photograph. You have just learned what an option does, from the outside.