Back home

Changelog

Every user-facing change in Vara. The source of truth is GitHub commits; this is the friendly summary.

Platform

Vara has a new address: varaos.com

  • • Vara's main site is now varaos.com. The old address varaapp.site still works and takes you to the same page on the new domain, so bookmarks and open tabs keep working.
  • • Your account and data are unchanged. You sign in once on the new address. Existing connections in Claude, ChatGPT and other AI assistants, API keys and calendar subscriptions keep working with no changes.
  • • Support email is now support@varaos.com; the old address still receives mail.
1.2.0
PlatformFeature

Mac app 1.2.0

  • • The Mac app is now 1.2.0. Download it from the download page; installed copies update automatically.
  • • Getting started: with no experiments yet, the home page offers three ways in: import from a paper, import from an external AI, or create one by hand. You can also open a sample experiment.
  • • Smoother runs: one Start run button that becomes Continue: step n while running; press Return to save a note; a one-line conclusion is enough to finish; each experiment shows its run history.
  • • Visible sync status: a status dot in the lower-left shows synced, waiting to upload, offline or failed uploads. Click it to sync now or resolve conflicts.
  • • More like a Mac app: the sidebar matches the web, with a new Run menu and common shortcuts. Press ⌃⌥N in any app for a quick note, drop files onto the home page to upload and parse, and ⌘K search works like the web.
  • • Vara AI chats are saved to the cloud: what you ask on the Mac shows up on the web, and you can continue web chats on the Mac.
  • • Clearer interface: higher text contrast, an 11pt minimum font size, plain-language error messages, and a nearly complete English interface.
Feature

Export all your data yourself from Settings

  • • Go to Settings → Data → Export all my data and click Export ZIP. Your browser downloads one ZIP file you can keep as a full backup, with no need to email us.
  • • It contains all your records (experiments, sub-experiments, procedure steps, parameters and their change history, runs and per-step results, materials and material logs, references, templates, plan items, daily notes, AI conversations, account settings), 5 tables that open in Excel, the original attachment files, and the original literature, kit manual and SOP files you uploaded to your experiments, plus notes in Chinese and English.
  • • Attachments and paper files together are bundled up to 500 MB, in upload order. Files that don't fit, outside links, and files that exist only on another device are listed with the reason, and you can download them from the experiment page.
  • • It covers experiments you created (including ones moved into a team). Deleted items and credentials such as passwords and API keys are not exported. You can export once every 10 minutes.
Feature

Use your own model key in Vara AI on Pro and Pro+, now with more providers

  • • Add your own model key under Settings → AI → My model, and Vara AI chat, Q&A, and parsing can run on your model. Your provider bills you directly, and it doesn't count against your Vara quota.
  • • Supported providers are OpenAI, Claude, Gemini, Grok, DeepSeek, Kimi, Qwen, GLM, MiniMax, Doubao, OpenRouter and SiliconFlow, or any OpenAI-compatible address. For providers with separate China and international endpoints, you can switch between them in one click.
  • • Pick “My model” in the chat model menu, or make it the default. Image understanding still uses Vara's model.
  • • “Test connection and save” calls your provider once before saving. When it fails, it tells you why, such as an invalid key, an account out of credit, or a wrong model name.
Feature

Free lab calculators, no login needed

  • • New public Lab calculators pages: antibody dilution (1:N), solution prep (% w/v) and PCR / qPCR reaction mix. They work the moment you open them, with no account or login.
  • • Each calculator shows the formula, converts units automatically (μL and mL), and includes bench tips and FAQs, in both Chinese and English.
  • • The math is identical to the calculator assistant inside experiment steps, so you can carry the numbers straight into a linked experiment in Vara.
Feature

Upgrade from Pro to Pro+ in one click; 10% off the first month for survey participants

  • • If you are on Pro, click Upgrade to Pro+ on the Pro+ card of the pricing page or in Settings > Billing. The price difference is prorated for the rest of the current period, so there is no need to cancel and resubscribe.
  • • Beta users who completed the pricing survey before October 7, 2026 (Beijing time) get 10% off their first subscription, applied automatically at checkout.
Feature

New protocol reproduction guides

  • • Ten new guides (qPCR, Western blot, ELISA, RNA extraction, transfection, cell passaging, CRISPR knockout, organoids, IHC/IF, flow cytometry) cover what methods sections usually leave out, which parameters change together, how to troubleshoot, and a worked example tagged by source confidence.
  • • Find them at /how-to-reproduce. Every reference links to its source.
Feature

Pro and Pro+ subscriptions now available (USD)

  • • The pricing page shows USD prices: Pro at $6/month and Pro+ at $19/month, and you can subscribe directly.
  • • Payments are processed by Creem as Merchant of Record. Plans renew monthly, you can cancel anytime, and you keep access until the end of the paid period.
  • • Settings > Billing shows your current plan and renewal date. Use Manage subscription to cancel, update your card, or download receipts.
  • • New Refund & Subscription Policy page explains renewal, cancellation, and how to request a refund.
Fix

Paper parsing is faster and more reliable

  • • Parsing papers, quick import and AI optimization no longer let the model think at length first; it extracts in the required format directly. Long papers now parse in under a minute instead of more than ten, so they no longer fail on timeout.
  • • If the model occasionally returns an empty result, it is retried once automatically.
Fix

Clear notice when a long paper is parsed incompletely

  • • For very long papers, the AI output can occasionally hit its length limit, so reproduction risks, AI inferences or the last few steps may be missing. This used to look the same as a normal result. The review page now tells you the result may be incomplete and offers a way to parse again.
  • • The default parsing model can now write twice as much, so long papers get cut off much less often.
Improvement

Paper parsing reads more of the paper and flags missing details

  • • When you upload a paper to reproduce, parsing now reads most of the text (up to about 120,000 characters) instead of only the first 15,000. Papers whose methods appear later get more complete steps and parameters.
  • • Parsing checks the details most often left out, such as how replicates were done, whether error bars are SD or SEM, antibody and cell line sources, how sample size was chosen, and where data and code are available. Anything the paper does not state is listed as a reproduction risk for you to check.
  • • Items that do not apply are left out. A purely computational paper will not be asked for cell line authentication, for example.
Feature

The flow view now shows lanes

  • • The main line is the longest chain from start to finish. If any step on it is delayed, the whole experiment is delayed.
  • • Branches that can run at the same time each get their own row, and parallel steps with the same structure (such as several samples prepared separately) are merged into one row.
  • • Click a sub-experiment to see what it waits for and what it affects. You can also highlight the critical path, search by name or number, and jump around with the overview at the bottom.
  • • The list is grouped into the same lanes, with the same numbers as the flow view.
Feature

Sub-experiment order is now marked automatically on import

  • • When you import a paper or paste a method, the AI also marks which sub-experiments must wait for others and which can run at the same time. You can see and edit this on the review page.
  • • On the experiment page, click "After #N" in the list or "Set order" in the menu to change the order directly. Options that would create a loop cannot be selected.
  • • The order is kept when you create a new version of a sub-experiment or duplicate an experiment.
Improvement

Sub-experiments now show as a list by default

  • • Sub-experiments on the experiment page now show as an ordered list, grouped into stages when some can run in parallel, the same on phone and desktop.
  • • To see the connections, switch to "Flow" above the list. Your choice is remembered.
  • • The Flow view now automatically zooms to fit the whole graph when it opens, and no longer draws gray dashed lines when no dependencies are set.
FeatureFix

Focus on the current step while running an experiment

  • • After you start a run, the screen shows only the current step: what to do, what to watch out for, which materials it uses, and what changed since last time, with the next step previewed at the bottom. Open "All steps" to see overall progress.
  • • Sub-steps no longer have stopwatches. Durations such as a 60 min incubation are listed with the key points.
  • • Materials linked to the step show batch, passage and expiry date, and expired or used-up items are highlighted. Reagents written in the step but not linked yet come with a suggestion to link them or add them to your materials; nothing changes until you click.
  • • Changes from the previous version of the design, and the note you wrote for this step in the last run, are shown right on the card.
  • • While a run is in progress, opening a step's detail page shows a prompt to go back to the step card, which opens right at that step.
  • • Fixed: marking a step as done no longer clears the note you already wrote for it.
ImprovementDocs

Help center FAQ expanded

  • • The FAQ now answers the basics: what Vara is, how to manage experiment records, reproducing a paper, whether you can trust AI extraction, supported disciplines, and pricing.
  • • The "What is Vara" page now lists what Vara does at a glance.
FeaturePlatform

macOS v1.1.3 released

  • • In-app auto-update is here — upgrade with one click from now on.
  • • Vara can now optimize a sub-experiment: describe what you want in chat, review, and confirm to create a new version (the original is kept).
  • • Review pages show provenance badges per step, plus a change-review dialog before confirming.
  • • Referrals are live — share your code from Settings and you both earn Pro days.
  • • BYOK in-app — pick "My model" in chat, test the connection in Settings.
  • • Rate each parse after confirming — your scores feed our public accuracy stats.
FeaturePlatform

macOS app updates

  • • Updated AI model list — added MiniMax M3, Xiaomi MiMo 2.5 Pro, and Kimi K2.7 Code; GLM upgraded to 5.2.
  • • Bug reports now attach a snapshot of your recently viewed experiment so we can reproduce issues faster.
Fix

Fixed dialogs closing mid-typing on mobile

  • • On phones, a dialog you were typing into (a survey, or a form to add an experiment, step, or record) could close on its own mid-typing and lose what you had entered. It now stays open until you explicitly close, cancel, or submit it.
Improvement

Clearer research disciplines: Life Science vs Biology

  • • Biology-related fields are now split more precisely: 'Life Science' covers human / disease / clinical / drug work (including dentistry and other medicine), while 'Biology' covers basic research on plants, animals, microbes, model organisms, and ecology.
  • • Picking the right discipline lets the AI understand your experiments better and give sharper defaults and troubleshooting. Disciplines you picked before were moved to the matching new category automatically — nothing to change on your side.
Feature

Xiaomi MiMo split in two — MiMo 2.5 (vision) and MiMo 2.5 Pro (reasoning)

  • • Xiaomi MiMo now comes in two tiers in the chat box — MiMo 2.5 for image recognition / vision (1×), and MiMo 2.5 Pro for complex text reasoning (3×). Pick the right one to spend your quota wisely.
  • • Choose MiMo 2.5 (vision) when uploading images; choose MiMo 2.5 Pro for strong reasoning, long-form writing, or complex analysis.
  • • Your default model is unchanged — just tap the switcher at the bottom of the chat box.
FeatureImprovement

New chat models — MiniMax M3, GLM 5.2, and a Kimi code model

  • • Three new options in the Vara chat model menu — MiniMax M3 and GLM 5.2, two stronger next-gen general models. The earlier M2.7 / GLM 5.1 stay available so you can switch and compare anytime.
  • • New Kimi K2.7 Code — a code-specialized model from Moonshot, handy for writing code and data-processing scripts.
  • • Your default model is unchanged — just tap the switcher at the bottom of the chat box to try the new ones.
Feature

Invite friends, you both get Pro time

  • • Grab your personal referral code and link on the "Invite friends" page (also in the sidebar)
  • • When a friend signs up with your code, they get +14 days of Pro right away (about 21 days stacked with the signup trial)
  • • After your friend creates their first experiment, you get +14 days of Pro too. The more you invite, the longer your Pro
  • • Works through your shared link, a code on the signup page, or one-tap Google sign-up
Feature

Import protocols from protocols.io, turned into Chinese-ready templates

  • • In the Template Library, tap "Import from protocols.io" to search public protocols and save any one as your own experiment template, no protocols.io login needed
  • • Search in Chinese too: Chinese keywords are auto-translated to English for the search, and result titles show in Chinese
  • • Imported protocol steps are auto-translated to Chinese, with reagent names and catalog numbers kept in English for searchability, and the original DOI preserved
  • • Tap "Apply" on a template to generate a full experiment with section-grouped steps, ready to use
macOS v1.1.1
FeatureImprovement

Mac sidebar upgrade and sync fixes

  • • Pin items in the sidebar — right-click any experiment, material, or reference to pin it to a dedicated group at the top
  • • Sidebar auto-groups by time — Today, Yesterday, Last 7 days, Last 30 days, Earlier — so it stays easy to scan
  • • Step dependencies, reproduction risks, and evaluation baselines now sync and display correctly
Improvement

A new Help Center, organized by task, open to everyone

  • • The single-page docs are now a multi-page Help Center, split into six areas: Getting Started, Guides, Vara AI Assistant, Reference, Account & Settings, and Help
  • • 18 fully bilingual pages, so you find help by what you want to do instead of scrolling through everything
  • • Open without signing in, so people who haven't registered yet can see how Vara works first
Improvement

The Vara side-panel chat now remembers each experiment

  • • The Vara chat in the side panel now keeps a separate conversation for each experiment, so switching experiments no longer carries over the previous one
  • • The empty chat state is now clickable suggestion bubbles, so it's easier to get started
Feature

Daily Plan is here — schedule your day on the dashboard, or let Vara plan it

  • • New "Today's plan" tab on the dashboard — one line per item: optional time + what to do, press Enter to add. Check done, edit the time, or delete, all inline.
  • • New dedicated "Plan" page — a full-month calendar on desktop, day-by-day on mobile. Tap any day (past / today / future) to plan it, and see at a glance what you did this week and what's coming.
  • • Let Vara plan for you — just say "plan my day: PCR at 9am, Western blot at 2pm" and confirm the card to schedule the whole day.
  • • Or snap a photo of a schedule table — show Vara a handwritten or printed agenda and it reads each row into plan items, even across several days.
  • • Planning stays separate from your experiment data — checking an item done won't touch your run logs. Web for now.
FeatureImprovement

The reproduction review page is now editable — and shows what changed before you commit

  • • After the AI parses a paper or protocol into an experiment, every field on the review page is now editable — experiment name, sub-experiments, each step, each parameter — no more creating it first and fixing things afterward.
  • • Before you confirm, a "Review your changes" dialog lists exactly what you changed versus the AI's parse (old → new, added, removed), so you verify before it lands in the database.
  • • Required fields and duplicate parameter names are caught and highlighted on the spot.
Improvement

A refreshed emerald look across the app

  • • Unified the brand accent to emerald and polished the landing page and AI/feature highlights for a cleaner, more consistent look.
Feature

Reference Library — paste a DOI to file it, then reproduce it in one click

  • • New "Reference Library" in the sidebar — paste DOIs (one per line, up to 50) to auto-fetch title / authors / journal / year; no DOI? Upload the PDF and Vara extracts the bibliographic details.
  • • Each reference is tagged by full-text availability — "Open Access" / "Institution access" / "Local PDF" at a glance.
  • • Link references to any experiment — attach them under "References" on the experiment page to track which paper an experiment reproduces.
  • • ⭐ Reproduce in one click — on a reference card, "⋯" menu → "Reproduce": open-access papers are fetched and parsed automatically; otherwise you're prompted to upload the PDF. After parsing you land in the review page, and the new experiment is auto-linked back to the reference (Pro).
Feature

Compare result images — line up your results side by side

  • • A new "Compare results" entry in the "Experiment attachments" area — line up result images (qPCR gels, cell photos, etc.) in columns side by side.
  • • Three comparison axes: across repeat runs of the same experiment, across versions, or even across different experiments via search.
  • • Give each result image a caption (e.g. "qPCR run 3") so you can tell them apart at a glance; click any image to view it full size.
Improvement

Materials can now be deleted

  • • Material list cards now show a ⋯ menu in the top-right — visible on hover on desktop, always visible on mobile. Confirm in the dialog and the material is hidden from the list.
  • • Material detail page header gets the same delete entry; after deletion you're redirected back to /materials.
  • • Soft delete (record kept in DB, recoverable on request — consistent with how experiments and steps are deleted).
FeatureImprovement

Mobile rebuilt to app-grade — no longer a shrunken website

  • • Brand-new bottom navigation bar — Home / Experiments / Materials / Me as four core destinations, plus a center Vara chat button, all within thumb reach. As you drill into experiments and steps, the top bar always shows the current title and a back button, so you never get lost.
  • • Experiment detail, steps, and the workflow canvas are all re-laid-out for phones — secondary content on long pages is collapsed by default and tool buttons move into a top "⋯" menu, so you scroll far less.
  • • Mobile Vara chat now has the same full capability as on desktop — it can directly create / look up / edit your experiment data, runs full-screen, and keeps a browsable history of past conversations.
  • • Dialogs now slide up from the bottom of the screen for a feel closer to a native mobile app; the Settings, API, Weekly Digest, Reproduction Review, Version Compare, and Material Detail pages are all adapted for mobile too.
  • • More mobile polish — larger touch targets, tapping an input no longer zooms the page, and bottom-anchored dialogs clear the iPhone home indicator.
Feature

Parsing models are now selectable

  • • /settings adds a [Parsing model preferences] section. Pro/Pro+ users can pick which AI model handles literature reproduction, smart review, and quick import.
  • • 9 models available: Doubao Lite/Pro/Code, DeepSeek V3.2, Kimi K2.6, GLM 5.1, MiniMax M2.7, MiMo V2.5.
  • • Dropdown shows live per-call cost estimates (¥0.008–0.20) so you see the tradeoff at a glance.
  • • Defaults pick the best price/performance balance (Doubao Lite for reproduction, Doubao Pro for review).
  • • Free tier is locked to defaults — upgrade to Pro to customize.
FeatureImprovement

Big UI update, plus materials and samples, and a media recipe calculator

  • • New Materials / Samples page: keep cell lines, media, reagents and antibodies in one place and reuse them across experiments instead of re-entering them for each one.
  • • Media details now have a recipe editor: add as many components as you like (mixing mL, % and L units), enter a target volume, and Vara works out how much of each to add.
  • • Two-way links between materials and sub-experiments: see which materials a step uses, see which steps use a material, and jump between them in one click.
  • • Exported run sheets let you pick what to include (summary, duration, steps, notes), in Word or PDF.
  • • The Vara chat context picker now has a Materials section next to Experiments. Both sections collapse, so long lists stay readable.
  • • Long-term memory, like ChatGPT: preferences and conventions you mention in chat are saved and applied next time.
  • • Fixed: parsing an English PDF now outputs in the language you chose in settings, instead of always staying in English.
  • • Paid plans are not open during the beta. The four pricing tiers are shown as a preliminary reference, with launch planned for mid to late June.
Feature

Pick from 10+ models in Vara chat

  • • Switch models at the bottom of the chat box: Doubao Seed 2.0 (4 versions), Kimi K2.6, GLM 5.1, DeepSeek V3.2, MiniMax M2.7 and Xiaomi MiMo.
  • • Each model shows a cost multiplier (0.25× to 4.5×), which is easier to compare than raw prices.
  • • When you upload an image, Vara suggests switching to a vision-capable model (MiMo, Doubao or Kimi).
Feature

Use @ to pick context in Vara chat

  • • A new [@ Add context] button above the chat box, the same idea as in Claude Code.
  • • Once you pick an experiment or material, Vara answers with its full data, without looking it up again each time.
  • • On experiment, sub-experiment and material pages, Vara picks up the current context automatically.
Feature

Export a journal club deck in one click

  • • After a paper is parsed, export a PPT draft in one click, ready for journal club or a literature talk.
  • • Two lengths: Full (10 to 12 slides, for a 30-minute talk) or Concise (6 to 8 slides, for a 15-minute lightning talk).
  • • Vara condenses the key points of the paper into an academic template with figure placeholders. Drop in your figures and present.
  • • Not happy with it? Ask Vara to regenerate. The content follows your language setting (Chinese or English).
Feature

Experiment photos and attachments, kept with the experiment

  • • Experiment pages have a new Photos & Attachments section. Upload bench photos, instrument screenshots and result images alongside the source paper PDF or SOP.
  • • Large phone photos are compressed automatically, and iPhone photos (HEIC) upload as they are.
  • • Three places to upload in each experiment: experiment photos, experiment references, and in-process photos on a step's run log.
Feature

Quick import now takes several images at once

  • • Notes that don't fit on one page, multi-page protocols, shots from several angles: you can now upload up to 3 images at once.
  • • Vara reads them together and pulls out the complete set of steps without duplicates.
Feature

Danger zone in settings: delete your account

  • • The Danger zone at the bottom of Settings lets you delete your account and permanently erase all your data (as required by Article 47 of China's Personal Information Protection Law).
  • • Three confirmations prevent accidents: enter your password, type the confirmation phrase, then click the red confirm button.
  • • Deletion is permanent, not hidden. Use this if you need your data formally erased.
macOS v1.0.16
Platform

macOS v1.0.16: windows fit 13-inch and 4K screens

  • • Several dialogs and settings panels now resize instead of using fixed sizes, so they don't overflow on small screens or look empty on large ones.
  • • Notarized and stapled by Apple, so Gatekeeper lets it through right away.
Feature

AI inferences now cite the source, so you can decide before accepting

  • • When Vara infers a parameter the paper doesn't state outright, it shows which passage of the paper the inference is based on.
  • • Open that passage in one click and decide whether to accept, adjust or skip.
  • • For the same field, you can also look up the related studies Vara used, to see what values other papers chose.
Feature

Run sheets can now be exported as PDF as well as Word

  • • Once an experiment is ready to run, export each sub-experiment's run sheet as Word or PDF.
  • • PDF is ready to print and pin next to your bench, with no Word layout differences.

Want to know when new versions ship?

Download the app →