Browsevio / User guide

A clearer path to getting things done.

Setup, everyday workflows and the details behind files, privacy and website testing.

Browse the guide or search for a task.

Overview

Browsevio is a desktop browser with an AI assistant, local profile storage, document tools and website automation. Start with a normal chat request and add a page, document or local folder when the task needs it.

This guide describes the current development preview. The download page states which packaged version is available; behavior can differ in older installed builds.

Start here: install Browsevio, open a website, then ask “Summarize this page.” Expand the chat controls to find attachments, history, memory, settings and browser tools.
You want to…Start with…
Understand a pageOpen it and ask a question in chat.
Edit a Word documentAttach the DOCX and describe the change.
Use project filesEnable Local access to folder, then Set folder path.
Test a websiteProvide its URL, requirements and the action you want.
Repeat form entrySupply a data file, target URL and intended operation.

Install & first launch

  1. Visit the download page and check the listed version and platform.
  2. Download the Windows x64 installer and run it. Follow the setup prompts and choose an installation folder.
  3. Open Browsevio. Select Continue with Google to authenticate the current profile. Return to the app after sign-in completes.
  4. Open a website using the address bar. The floating assistant can work alongside it.
  5. Ask your first question. Expand the chat control bar for settings, files, saved chats and other tools.

Before installing a preview

Check the signing information on the download page and only use an installer you trust. Do not disable operating-system protections to install a blocked build. The .sha256 file lets you compare the installer’s SHA-256 hash after download.

AI features require a working internet connection and authentication or a configured supported provider. This guide does not promise free or unlimited AI usage; availability depends on the account and provider configuration.

Platforms

The downloadable target is Windows x64. No macOS, Android or iOS installer is published on this site.

Chat & browser controls

Use chat to ask a question, request a page action, or continue an existing task. Refer to the current page, name a website, or identify tabs by title or order when comparing them. Give an explicit URL when it matters.

A compact workspace

The floating chat window prioritizes messages and typing. Its compact control bar contains response status, the latest-response arrow and the expand control. Expand it to access saved chats, attachments, the project folder, settings and browser tools. Controls can be collapsed again to reclaim space.

  • Latest-response arrow: bring the newest assistant response into view.
  • Move: use a move grip near any corner of the chat window.
  • Resize: drag a resize corner. The window also adapts to smaller available screen areas.
  • Pause or stop: use the task controls while a task is running.
  • New chat: start a separate conversation when the topic changes.

Example requests

  • “Summarize this article in five points.”
  • “Compare the first two tabs.”
  • “Write a short thank-you reply here.”
  • “Find this item on Walmart and add one to the cart.”

For travel and booking searches, provide required details such as route, dates and trip type. If an essential choice is missing, the assistant should ask before dependent work.

Drafting is separate from sending. Email sending has a dedicated review and approval step. Final purchases, bookings and other protected commitments require your completion.

Attachments & documents

Expand the chat controls and choose Attachments to add a file. Supported workflows include PDF, Word, text, CSV, Excel, JSON and code. Tell the assistant what to do with the file; attaching it alone does not define the task.

Read and summarize

Text extraction happens locally. Relevant extracted content can be sent to the selected AI provider to answer your request. A PDF is suitable for reading; use DOCX when you need changes to a Word document’s original structure.

Edit an existing document

Attach the original DOCX and describe the changes, for example: “Add this work experience as a separate entry.” The document tools inspect paragraphs, headings, tables and supported formatting, then save an edited copy. The original remains available.

Existing styles and unrelated structure are preserved where supported. Longer text can change line wrapping and page count. Arbitrary chart, image or table-row restructuring, protected complex content, legacy DOC editing and original-layout PDF editing are limited or unsupported.

Create a new document

Describe the output and content you need. Built-in Word layouts cover letters, reports, meeting notes, proposals and resumes. A supplied DOCX or DOTX can also be used as a template.

Find and upload files

Generated documents appear as saved files with download/open controls. After a successful response, a consumed attachment is deselected for the next message but retained for reuse. Asking to upload a file to a website is a separate action that shares the actual file with that website. Upload acceptance must be verified; selecting a file alone is not proof of submission.

Local folder access

  1. Open Settings → Local files.
  2. Enable Local access to folder.
  3. Expand the chat controls and open Project folder.
  4. Select Set folder path and choose the project folder.
  5. Place requirements, data and source files there, then request the action in chat.

No project name or separate testing setup form is required. A different selected folder uses its own workspace context. Disabling local access revokes folder tools and hides the picker.

The assistant can list, read, create and edit supported files within the selected folder and its subfolders. Existing file edits require matching the current content and keep backups. Secret files, credential files, repository metadata, dependencies, linked paths and files outside the allowed folder are excluded.

Local file access is not terminal access. These tools do not execute arbitrary programs, shell commands or generated framework code.

Website testing workflow

Provide the website URL, functional requirements, relevant environments, account roles and the testing action you want. Attach requirements or keep them in the selected project folder. Use a URL accessible from the computer running Browsevio.

  1. Create the manual cases. Ask “Create manual test cases from these requirements for this URL.” The assistant writes cases with steps, prerequisites and expected results.
  2. Review and approve. Check the manual cases and explicitly approve them in chat. A changed cases file needs approval again.
  3. Request execution. Ask to run the approved cases against the target URL. Login cases use the local account file described below.
  4. Record what happened. Cases are marked passed, failed, blocked or not run based on fresh observations. A previous saved report is historical evidence, not a fresh result.
  5. Generate reusable files. Ask for Playwright files from the recorded run. Observed steps and verified locators become config, fixtures, page objects and specs.
  6. Run the exported suite separately. Follow the generated README in your development environment.

Creating cases alone does not authorize running them. Browsevio does not execute arbitrary generated Playwright, Selenium or other framework programs through its chat file tools.

When requirements change

Supply the separate change document and point to the original requirements and test files in the selected folder. The local requirement index links requirement IDs to coverage, identifies affected areas and can adopt existing tests. Ambiguous changes should be clarified before editing related tests.

Execution limits

Fresh isolated login sessions do not reset server-side data. Use appropriate test environments and accounts. CAPTCHA, rate limits, inaccessible pages, unsupported controls or missing credentials can block a case. Multi-origin SSO journeys have additional limits.

Local test credentials

If a test requires login, create a file named exactly test-accounts.json in the selected project folder’s root. Do not attach it to chat or paste its passwords into prompts or requirements. Create and edit it locally.

The following is a format example only. Replace the example values on your computer:

{
  "sites": [
    {
      "origin": "http://127.0.0.1:8085",
      "accounts": [
        {
          "name": "Employee",
          "username": "REPLACE_LOCALLY",
          "password": "REPLACE_LOCALLY"
        }
      ]
    }
  ]
}
  • The origin is scheme, host and port, with no page path or trailing slash.
  • localhost and 127.0.0.1 are different origins. Use the exact one opened by the test.
  • Use unique role names and nonempty usernames and passwords. Add a separate site entry for each environment.
  • Add test-accounts.json to your project’s .gitignore.
  • If the file, site or role is missing, fix it locally and retry. The assistant should explain what is missing.

The local login mechanism gives the model account names and opaque references, not the usernames or passwords. The browser fills the requested test login locally. The source JSON remains a plaintext local file; the imported copy is encrypted in the browser’s local storage.

Repeated data entry

Attach a supported data file or place it in the selected folder. Supply the target website, identify the relevant rows and columns, and describe the operation you want repeated.

Example: “Use this CSV to fill the customer form on this URL for each row. Show me the field mapping before starting.”

Have the assistant inspect the file, resolve field mappings and clarify missing values before dependent work. Repeated entry and supplied-data testing use bounded website actions; they do not imply permission to complete purchases, send messages or bypass protected submissions.

Check row-level results and failures. A partially completed batch should not be reported as fully successful. For account-based tests, use the dedicated local credential file rather than passwords in the data attachment.

Privacy & data handling

What stays on your device

Profile records, chat history, saved form values and supported attachment metadata use local storage. Sensitive database records use operating-system encryption. Profiles are kept separate; private browsing uses temporary storage for its supported records.

What can be sent to AI

Relevant chat messages, page excerpts and document text can be sent to the selected AI provider. Text extraction is local, but that does not make subsequent AI processing offline. Known saved sensitive values and supported credential patterns are masked before applicable model requests. Arbitrary unknown personal text is not guaranteed to be detected.

Credentials and websites

Dedicated local mechanisms resolve authorized saved values and test-account credentials without putting their original values in GPT prompts. Filling a form intentionally shares the requested values with that website. Uploading a document shares the actual file with the target website, including any information it contains.

Files outside the encrypted database

Source documents, exported reports, downloaded files and test-accounts.json are ordinary local files. They are not encrypted merely because the browser’s metadata is encrypted. Treat those files according to the sensitivity of their contents.

Review what you attach and upload. Local storage and masking reduce some exposure; they do not make all browsing or all AI requests private by default.

Logs & troubleshooting

The browser does not open

Use the current completed installer rather than an .incomplete file. Record the exact startup error. Preserve your local data and logs; do not delete the profile folder as a first troubleshooting step.

Where logs live

Encrypted logs are stored in the logs folder inside Browsevio’s local data directory. On the current Windows build the inherited default data directory is %APPDATA%\Smart Browser. Keeping this name preserves existing profile continuity.

Logs rotate at approximately 10 MB, rotated files are compressed, and retention is capped at approximately 200 MB by removing older files. Use the supported local decryption tooling when diagnosing an exported log; log files are not readable plaintext.

Google sign-in or server connection fails

Check internet access, retry sign-in and note the displayed error. The managed backend address is configured by the application. There is no server-address setting in the normal UI.

Tests need credentials

Check the selected folder, exact test-accounts.json filename, valid JSON, website origin and requested role. See Local test credentials. Do not paste the password into chat as a workaround.

A website stops automation

Respect CAPTCHA, access checks and rate limits. Ask for the actual observed blocker and a partial-results report rather than retrying indefinitely.

Updates & preview limits

Installed builds include update support, subject to a compatible release being published to the configured update feed. When an update is available, the app can offer installation and restart. The first locally staged preview is not proof of an end-to-end public update rollout.

Downloads on this website are independent of publishing an update release. A new installer must be built, verified and staged before the download page points to it.

  • Windows is the currently staged platform; macOS has not been verified here, and mobile apps are not implemented.
  • AI outputs and important website actions should be checked.
  • Document structure retention is not a guarantee of identical pagination.
  • Exported test code is run in your own development environment.
  • Live sites can change controls, require verification or block automation.

For help, contact Shrismit with the app version, steps and observed error. Keep credentials and other sensitive data out of your report.