Import & Parse

Papers, photos, notes, even content drafted by another AI, all turn into experiments in one shot.

Lesson 2Getting a protocol in, and checking the AI's numbers2:36

Ways to import

Five sources, all going through Smart Review before becoming a full experiment:

Source	How to import	Notes
Paper PDF	“Upload a file” card on the starter or new-experiment page / “Upload a protocol or paper” in the New menu / drop onto the home page / drop into Ask Vara / `+ Attach`	Auto-detects discipline, extracts accordingly
Experiment photos · handwritten notes	Same as above, vision model reads the image	Upload iPhone HEIC directly, auto-converted to JPG
Multiple images	Up to 3 at a time	Combined extraction across pages, no duplicates
Web / note text	Paste into the chat, say "turn this into an experiment"	Plain text parses too
Markdown drafted by an external AI	Paste in using the import template	No cold start: bring over work organized elsewhere

Limits: documents up to 20MB, and only the first 200K characters of text are read; 10MB per image (large phone photos auto-compressed); up to 10 uploads per 60 seconds.

Two parse modes

New experiment: parse a whole paper / multiple experiments → create an experiment + N steps.
Append sub-experiment: within an existing experiment, parse a single protocol step and add it.

Long PDFs (within 200K chars) parse in full; oversized text is truncated for the agent to grasp an overview.

Starting an empty experiment: paste stepsno AI

When an experiment you created by hand has no sub-experiments yet, the page shows three ways to start: Paste steps, Upload a file (the same quick import as in Import sub-experiment at the top) and Pick from templates.

Paste the protocol text you already have. It understands 1. 1) (1) ①, Step 1, bullets and indented sub-items. Without numbering it splits by blank lines, or one item per line if there is a single paragraph. Values such as 1.5 mL or -20 °C are not treated as numbering.
The split is previewed live below. You can edit titles and remove items. With headings or second-level items, each top-level item becomes a sub-experiment and its sub-items become procedure steps by default; you can switch to One sub-experiment.
Click Create to save everything at once. Only the numbering is removed and the text is not rewritten: the first sentence of each item is its title and the rest is its description.

A sub-experiment page with no procedure steps also offers Paste steps; there everything becomes procedure steps of that sub-experiment. One paste creates at most 20 sub-experiments and 100 steps, and anything beyond that is reported as not imported.

Import from an external AIno cold start

Already drafted a protocol in ChatGPT / Claude / DeepSeek? No need to redo it in Vara: bring the whole set in and fill a new account instantly. This is an experiment-level import (sub-experiments, parameters, materials), different from parsing a single document.

Entry: on the starter page or the new-experiment page, open "Want a more accurate parse? Copy a prompt for your AI" under the "Paste text" card. If you already have experiments, [Import from external AI] in the New menu (sidebar, or next to Recent experiments on the home page) works too. The steps:

Copy the Prompt: the page gives you a ready-made structured prompt; click "Copy Prompt".
Feed it to the external AI: paste the Prompt together with your paper / notes / even a one-line research idea into ChatGPT / Claude / DeepSeek and let it generate the result.
Paste back into Vara: paste the AI's Markdown output back in, or upload a .md / .txt file.

After submitting, it enters Smart Review (including material confirmation); confirm to generate the full experiment. The Prompt is provided in Chinese or English to match your interface language.

Why this avoids a cold start: new users have often already drafted a protocol in another AI, and this path moves that work over in one click, instead of starting from a blank account.

How it differs from Quick Import: Quick Import parses a single PDF / image into one sub-experiment in a single local step; external-AI import brings over the whole experiment and needs one round trip through the external AI.

Import protocols from protocols.ioprotocol library

protocols.io hosts a huge library of public protocols. In the Template Library, click [Import from protocols.io], find a protocol, save it as your own experiment template, then apply it to create an experiment. Public protocols only, no protocols.io login needed.

Search: type a keyword (Chinese works too: auto-translated to English for the search); results show step count / reagent count / DOI.
Save as template: pick one to store its steps + reagents as an experiment template, filed under the "protocols.io" source, with the original DOI link kept.
Apply: on a template, click "Use / create experiment" to turn the section-grouped steps into a new experiment.

Auto-translation for Chinese users: when your interface is in Chinese, protocol titles / steps are auto-translated (reagent names, catalog numbers, and technical abbreviations stay in English for searchability), labeled "AI-translated, please verify against the source".