Back to site·Changelog

What's new in Genzbots

All improvements, new features, and bug fixes — every release, documented.

Major release Minor release Patch
v2.1.6patchSeptember 2, 2026

Distinct "Launch Application" and "Launch Browser" activities, plus reliability and security refinements

Improvements

  • Clearer automation building blocks: "Launch Application" (for desktop apps) and "Launch Browser" (for websites) are now two distinct, clearly-labeled activities - including in AI-generated steps - so it's obvious which to use. Your existing saved workflows continue to run unchanged.
  • Reliability, quality and security refinements across Studio and the AI service.
v2.1.5patchAugust 31, 2026

Consistent data-field references across AI-built steps, AI Fill and drag-and-drop

Improvements

  • Consistent field references: steps generated by AI and by AI Fill now use the same clear row-field references as drag-and-drop and the input/output drawer, so data-driven agents read consistently from end to end. Your existing saved workflows continue to run unchanged.
  • Stability and security refinements across Studio and the AI service.
v2.1.4patchAugust 31, 2026

Agents that learn from past runs, plus more dependable data-driven automation

Improvements

  • Self-improving reliability: agents can now learn from your previous runs to make more reliable execution decisions and to recover from transient failures automatically. This learning is stored privately and encrypted on your own device, and can be turned off at any time.
  • More dependable data-driven agents: reading spreadsheets and data sources, processing each row with conditional logic, and writing results are now more reliable, with accurate per-row progress and honest run status.
  • Smoother agent building: add breakpoints while a run is in progress, a cleaner run drawer with per-iteration output, and clearer, uniquely-named step outputs.
v2.1.3patchAugust 28, 2026

More resilient automation across web and desktop

Improvements

  • More reliable element targeting - when a saved target drifts, Studio automatically falls back through layered strategies (page structure, accessibility information, and on-screen text and imagery) to find and act on the right element instead of failing.
  • Desktop application automation - configured actions can now target native Windows desktop controls through the same resilient resolver Studio uses for the browser, so a single agent can work across web and desktop surfaces.
  • Honest, verified execution - every action confirms the intended outcome actually happened and reports truthful successes and failures; when a target is ambiguous the action stops safely rather than acting on the wrong element.
  • Includes all reliability improvements from 2.1.2, plus performance and stability refinements across Studio.
v2.1.2patchAugust 27, 2026

Smoother Authoring

Improvements

  • Change a block's type in place - click the type chip on any block to convert it to a compatible type, keeping its position, connections and compatible settings.
  • More reliable Copilot editing - activities are placed into the correct Try/Catch/Finally branch, and automatically generated flows no longer create invalid loops.
  • Reliability and performance improvements across Studio.
v2.1.1patchAugust 26, 2026

Real-Time Voice

Improvements

  • Real-time voice transcription - see your words appear live as you speak when dictating to the Copilot or describing an automation by voice.
  • Smoother voice dictation - clear what you have written and keep speaking without earlier words returning.
  • Faster, more accurate speech-to-text across Studio.
v2.1.0minorAugust 18, 2026

Smarter Authoring, Clearer Execution

Improvements

  • Continue building existing agents naturally - add new automation through Chat, Voice, or Demonstrate without recreating completed workflow sections.
  • Improved demonstration-based automation - preserve multiple demonstration sessions and their context while creating incremental workflow sections.
  • More reliable browser automation - stronger element resolution, verification, recovery, interruption handling, and data-driven form interactions.
  • New Block Inputs and Outputs experience - inspect schemas, actual run results, execution status, and action-level outcomes through Schema, Table, and JSON views.
  • Visual data binding - drag available output fields into compatible parameters while Studio creates clean reusable bindings such as {{row.department}}.
  • Better execution transparency - inspect action status, verification, attempts, resolution strategy, recovery information, and truthful failures.
v2.0.10patchAugust 11, 2026

Improved Agentic Studio voice and demonstration experience

Improvements

  • Choose your microphone directly in Agentic Studio - select and reuse your preferred audio input for Goal dictation and Demonstration recording.
  • Improved Demonstration setup - microphone selection, microphone testing, live input-level monitoring, and transcript feedback are available before you start recording.
  • More reliable voice interactions - Goal dictation, microphone testing, and Demonstration recording each use an isolated voice session, so they never interfere with one another.
  • Better live speech feedback - improved multi-segment transcript handling and no-speech suppression give a cleaner, quieter dictation experience.
  • All Studio 2.0.9 Classic reliability improvements remain included - enhanced AI Copilot, automatic variable lifecycle management, Undo reliability, validation improvements, and workspace enhancements.
v2.0.9patchAugust 11, 2026

Classic Studio reliability and authoring improvements

Improvements

  • Smarter Classic AI Copilot - workflow-aware Questions, natural-language edits, and precise activity targeting.
  • Improved variable lifecycle management - automatic creation, binding, copy and delete, plus Clear Unused Variables and Arguments.
  • More reliable workflow editing - stronger Undo/Redo, better validation handling, and workflow tab integrity.
  • Improved Studio workspace - project-folder state is remembered and the empty designer supports quick workflow creation and opening.
  • Broad Classic authoring stability and reliability improvements.
v2.0.8patchAugust 5, 2026

Beta access extended through August 31, with updated privacy disclosures and safer update checks

Improved

  • Beta access extended through August 31, 2026 — Standard and Pro remain available during the extended beta.
  • Updated privacy, telemetry, and AI-processing disclosures.
  • Corrected minor product labels and descriptions.

Security

  • The updater now rejects unexpected redirects.
v2.0.7patchAugust 3, 2026

Agents now honour your approved dropdown fixes and stop repeating them

Fixed

  • When an agent picks a value from a dropdown, it now confirms the value actually changed - if it didn't, it says so clearly and lists the choices that were available, instead of quietly moving on and failing later.
  • A fix you approve for a step is now reliably applied on the next run, so the same item stops coming back to your review panel over and over.
  • Agents remember the correct element for a step and reuse it, so a step that needed a correction once runs cleanly afterwards.
v2.0.6patchAugust 2, 2026

A live AI budget and usage breakdown in Agentic Studio

New

  • Agentic Studio has a new AI Usage view - see how much of your monthly AI budget is left at a glance, and exactly where it went: building agents from your demonstrations, agent reasoning during runs, self-healing, voice and more.
  • Beta users now see their free credit clearly, so you always know where you stand.

Improved

  • Agent runs now pause automatically if you reach your AI budget for the month, so there are no surprises.
  • Your plan name now shows correctly everywhere across the app.
v2.0.5patchAugust 2, 2026

A clearer way to upgrade, and agents that learn from every run

Improved

  • Your agents now learn from every run: once you approve a fix in the review panel, it is applied automatically on future runs - so the same step stops failing and getting corrected over and over.
  • Agents remember the fields and dropdowns they have already worked with, so they run more reliably each time instead of guessing again.
  • A clearer upgrade experience: when something needs a higher plan, you now see exactly which plan unlocks it, with the key benefits laid out side by side and no clutter.
v2.0.4patchAugust 2, 2026

Recorded agents now self-check both their steps and their data

Improved

  • The review check now also verifies the data your agent fills in: every value it types is confirmed against what you actually typed during the recording, and anything it cannot recognise is flagged for you to review.
  • After building an agent from a recording you now get a clear self-check summary - either confirming every step and value traces to your recording, or listing exactly which items to review before running.
v2.0.3patchAugust 2, 2026

Agents built from recordings now flag unverified steps for review

Improved

  • When you build an agent by recording a demonstration, the assistant now checks that each step targets an on-screen element you actually interacted with. If a step points to something it cannot confirm from your recording, it is flagged for you to review before running - avoiding surprises where a step cannot find its target at run time.
  • Agents generated from recordings are now more consistent and reliable, with the assistant kept to the set of actions your recording actually demonstrated.
v2.0.2patchAugust 2, 2026

Agents are now built using the AI assistant by default

Fixed

  • Recording a step where you type a web address to open a page is now captured reliably as a navigation step, instead of occasionally being recorded as an unclear action that needed manual review.

Improved

  • When you build an agent by recording a demonstration, the AI assistant now handles it directly by default, producing more complete and accurate agents from what you show it.
v2.0.1patchAugust 2, 2026

Genzbots Studio 2.0 - smarter demonstration recording and agent-building improvements

Added

  • Demonstration recording is now more thorough. When you build an agent by showing a process, the assistant reviews the screen recording to understand moments that happen between your clicks - such as a page finishing loading or a menu appearing - so the generated agent reflects more of what you actually did.

Fixed

  • When recording in a web browser, typing a web address to open a page is now correctly captured as a navigation step in the agent. Previously this step could be missed.
  • Test results for a block now appear immediately after you run a test. Previously you had to close and reopen the block to see them.

Improved

  • New agents now receive clear, automatically numbered names, so multiple drafts are easy to tell apart.
  • Creating a new agent now opens directly into a full-screen canvas, giving you more room to design.
  • Strengthened the security of the connection between Studio and the AI service.
v1.4.61patchJune 9, 2026

fix: Ctrl+V now correctly pastes files and screenshots into Copilot chat

Fixed

  • Pasting files and screenshots into the Copilot chat input using Ctrl+V now works correctly. Files copied from Explorer (Ctrl+C) and images from all screenshot tools (Win+Shift+S, Snip & Sketch, Print Screen) are attached to the message. Previously, only plain text could be pasted; all other clipboard formats were silently ignored.
v1.4.60patchJune 9, 2026

fix: Copilot chat now accepts pasted screenshots and images from clipboard

Fixed

  • Screenshots and images copied to the clipboard (Win+Shift+S, Snip & Sketch, Print Screen) can now be pasted directly into the Copilot chat input using Ctrl+V. Previously, pasting an image silently did nothing. The image is attached as a file and included with the next message sent to the AI.
v1.4.59patchMay 29, 2026

Security: AI prompt privacy and training data encryption

Security

  • AI prompts are no longer retained in server logs. Only the character count is recorded, ensuring your automation instructions remain private.
  • Training data stored on the server is now encrypted at rest using AES-256-GCM. All previously stored training records have been migrated to encrypted storage.
v1.4.58patchMay 29, 2026

Improved reliability of AI-generated workflow quality checks

Fixed

  • AI-generated workflows now undergo an additional quality check before being returned. This step detects undefined variables, broken activity sequences, and incomplete actions, and automatically corrects them when issues are found.
v1.4.57patchMay 29, 2026

Improved AI Copilot reliability and consistency

Fixed

  • Several Copilot operations were not consistently using the primary AI provider. All Copilot functions now use a unified provider order with automatic fallback, ensuring a more reliable experience.
v1.4.56patchMay 29, 2026

Fixed AI provider routing for Analyze and Verify workflow operations

Fixed

  • Analyze Workflow, Pre-verify, and Post-verify Copilot operations were not routing through the primary AI provider. They now follow the same provider order as all other Copilot functions.
v1.4.55patchMay 29, 2026

fix: AI toggle now routes to local templates when off; online prompts go to AWS first

Fixed

  • Turning off the AI Copilot toggle now uses local template-based responses instead of a disabled message. Workflow generation and answers still work without calling any cloud AI.
  • When AI is enabled, all prompts (chat questions and workflow generation) are now sent to AWS Bedrock first. Azure OpenAI is only used as a fallback if AWS is unavailable.
v1.4.54patchMay 29, 2026

fix: AI Copilot toggle now blocks requests; copilot no longer reveals underlying model

Fixed

  • AI Copilot toggle was only cosmetic - requests still went through when paused. Toggle now blocks all AI calls and shows a clear message instead.
  • AI Copilot will no longer mention or confirm the underlying AI model or provider when asked. It identifies itself only as Genzbots AI Copilot.
v1.4.53patchMay 29, 2026

ui: compact AI Copilot header - remove title text, keep icon and toggle

Improved

  • AI Copilot header is now more compact - the title text has been removed to free up space, keeping the icon, Preview badge, and status toggle. Hover over the toggle to see the current status.
v1.4.52patchMay 29, 2026

fix: AI Copilot toggle now saves correctly

Fixed

  • Turning off the AI Copilot toggle had no effect - the setting reverted to enabled every time Studio restarted. The off state (false) was being silently dropped during save because it matched the boolean default value. Fixed so the toggle state is always written to disk.
v1.4.51patchMay 29, 2026

fix: machine ID mismatch now prompts re-login instead of showing offline warning

Fixed

  • After a security update changed how the device identity is stored, Studio showed a recurring offline grace-period warning instead of asking the user to sign in again. The device identity mismatch is now detected correctly and Studio prompts for re-login immediately, which re-registers the device and resolves the issue permanently.
v1.4.50patchMay 29, 2026

fix: offline license warning no longer appears at startup when online

Fixed

  • Studio showed an offline grace-period warning at every startup even when the license server was reachable. A previous transient failure had activated an internal retry delay that was being restored from disk on startup, silently skipping the server check and returning the cached offline result. Startup validation now always contacts the server directly, bypassing any stored delay.
v1.4.49patchMay 29, 2026

fix: auto-update no longer blocked by missing code-signing certificate

Fixed

  • Auto-update could not install because the installer was not yet digitally signed. Update now proceeds normally - signature enforcement will be re-enabled once a code-signing certificate is in place.
v1.4.48patchMay 29, 2026

fixes: startup license check, AI Copilot toggle, errors panel hidden by default

Fixed

  • Studio no longer shows an offline warning at startup when the license server is reachable - added a short retry before any grace-period dialog is displayed.
  • AI Copilot toggle in the toolbar is now clickable - switching the copilot on or off is saved across sessions.
  • Errors panel is now hidden by default and only appears when an activity error chip is clicked or a runtime error occurs.
v1.4.47patchMay 29, 2026

security: training data encryption (Gap 4) + per-robot agent identity (Gap 5)

Security

  • Training DB prompt and workflow fields are now encrypted at rest with AES-256-GCM (a configuration setting env var). Backward compatible - existing plaintext rows decrypt transparently. POST migrates legacy rows.
  • Agent self-registers its API key in the heartbeat payload. Dashboard stores it per-robot and sends it as an API header on every execute call, giving each robot a separate revocable identity. Agent logs every request with the resolved caller identity (dashboard vs operator).
  • Agent middleware now accepts a distinct DashboardApiKey (optional) separate from the operator ApiKey, with per-caller audit logging on every inbound request.
v1.4.46patchMay 29, 2026

security: rate limiting, CORS hardening, auth logging

Security

  • Added per-caller rate limiting on AI endpoints to prevent cost abuse.
  • Tightened CORS policy to production origins only (desktop Studio unaffected).
  • Unauthenticated requests on protected routes now generate audit log entries with caller IP.
v1.4.45patchMay 29, 2026

security: network transport hardening (NT-C1, NT-H1, NT-M1)

Security

  • All external traffic to AI and Dashboard services now requires HTTPS. Services no longer expose ports directly.
  • Agent API key is now mandatory. The agent refuses to start without a configured key. All endpoints enforce authentication unconditionally.
  • RPA recorder compatibility endpoints now require authentication, consistent with all other API routes.
v1.4.44patchMay 29, 2026

security: tightened network configuration and audit logging

Security

  • Tightened validation of backend URL configuration.
  • Added audit logging for sensitive operations.
v1.4.43patchMay 29, 2026

security: minor internal hardening

Security

  • Minor internal hardening and code documentation improvements.
v1.4.42patchMay 29, 2026

security: hardened package handling and process execution

Security

  • Hardened project package extraction against malformed archives.
  • Strengthened credential storage with improved encryption.
  • Improved safety of system command execution.
v1.4.41patchMay 28, 2026

security: internal code hardening

Security

  • Internal code hardening and documentation of security controls.
v1.4.40patchMay 28, 2026

security: strengthened telemetry integrity and input handling

Security

  • Strengthened telemetry integrity with batch-level verification.
  • Improved input handling to prevent header injection.
v1.4.39patchMay 28, 2026

security: hardened file handling and data protection

Security

  • Improved data protection for stored credentials.
  • Hardened file extraction and download handling.
v1.4.38patchMay 28, 2026

security: resolved security issues

Security

  • Resolved security issues.
v1.4.37patchMay 28, 2026

security: resolved security issues

Security

  • Resolved security issues.
v1.4.36patchMay 28, 2026

security: resolved security issues

Security

  • Resolved security issues.
v1.4.35patchMay 28, 2026

ui: simplified settings page

Improvements

  • Simplified the Settings page - removed internal configuration fields not relevant to end users.
v1.4.34patchMay 28, 2026

security: resolved security issues

Security

  • Resolved security issues.
v1.4.30patchMay 28, 2026

security: fix C1/C2/H1/H2 - hardcode beta expiry, activity feature gate, cross-process robot limit, DPAPI entropy

Security

  • Beta expiry date is now hardcoded in C# (2026-07-16) - editing license-tiers.json cannot extend the beta window.
  • WorkflowExecutor enforces feature gates on activities before execution - activities with RequiredFeature throw FeatureNotLicensedException even when Studio UI is bypassed.
  • InvokeWorkflowActivity now declares RequiredFeature = studio.invoke_workflow - calling it from Robot or a custom app requires Standard+ tier.
  • Concurrent robot limit now uses a named system Semaphore (cross-process) - launching multiple Robot.exe instances no longer bypasses the per-tier limit.
  • DPAPI entropy now includes the Windows MachineGuid - renaming a PC is no longer enough to replay a stolen license blob on another machine.
v1.4.29patchMay 28, 2026

security: harden licensing - hardcode feature grants, encrypt grace-period clock, disable beta fallback

Security

  • Tier feature grants are now hardcoded in C# - editing license-tiers.json no longer changes which features a tier unlocks.
  • Last-validated timestamp is now stored inside the DPAPI-encrypted license blob instead of a plain-text file or registry value - resetting it no longer extends the offline grace period.
  • Missing license-tiers.json now falls back to beta=disabled (Starter tier) instead of silently granting Pro access.
v1.4.28patchMay 28, 2026

fix: 8 File/Panel/Run menu fixes

Bug Fixes

  • Save Project As dialog now shows 'Save Project As' header instead of 'Create New Project'.
  • Errors panel is now hidden by default - shown automatically when errors are detected or via View menu / Ctrl+Shift+E.
  • Activities, Variables, Project, AI Copilot, Properties and Output panes are now mandatory (CanHide=False) - they cannot be hidden from the pane header, only auto-hidden.
  • Run From Selected now correctly runs from the selected activity (was falling back to running the entire workflow).
  • Pause/Resume shortcut (F6) now works for both regular runs and debug runs.
  • Pause / Resume Execution moved from Run section to Debug section in the Run menu.
  • Stop Recording shortcut (Shift+F2) now acts as the Done button - stops capture and imports recorded actions into the workflow.
  • Removed duplicate F5 binding for Pause (F5 is used for Record toolbar button).
v1.4.27patchMay 28, 2026

feat: panel shortcuts always activate panel, never hide; full shortcut coverage for all View menu panels

New Features

  • Panel shortcuts (Ctrl+Shift+A/P/V/O/E/I) now always show and activate the panel instead of toggling visibility. Use View menu checkboxes to hide panels.
  • Added shortcuts for all remaining panels: Execution Trace (Ctrl+Shift+T), Healing (Ctrl+Shift+H), AI Usage (Ctrl+Shift+U), Locals (Ctrl+Shift+L), Watch (Ctrl+Shift+W), Immediate (Ctrl+Shift+M).
  • All new shortcuts shown in View menu as InputGestureText hints.
v1.4.26patchMay 27, 2026

fix: Type Into with ClearFirst=false now appends instead of replacing

Bug Fixes

  • Type Into activity with Clear Before Typing unchecked now correctly appends text instead of replacing the existing field value. Root cause: (1) Playwright FillAsync always replaces the entire field value - fixed by skipping FillAsync when clearFirst=false and using TypeAsync (append) instead. (2) TypeViaNativeSetterAsync JS set el.value = text unconditionally - fixed to append: el.value = existingValue + text. (3) Desktop UIA ValuePattern.SetValue(text) also replaced unconditionally - fixed to read current value and concatenate. (4) VerifyTypedValueAsync now accepts actual.EndsWith(expectedText) as success when clearFirst=false.
v1.4.25patchMay 27, 2026

fix: highlight overlay now stops immediately when Selector Editor is closed

Bug Fixes

  • UIA element highlight overlay is now cancelled and closed the moment the Selector Editor dialog is closed. Root cause: HighlightOverlayWindow was created with a fire-and-forget Task.Delay(2s).ContinueWith() that had no cancellation path - closing the dialog called Dispose() which only disposed the FlaUI automation object, leaving the overlay window alive until the 2-second timer fired. Fix: SelectorValidationService now tracks the active overlay and a CancellationTokenSource; Dispose() calls CancelActiveHighlight() which cancels the timer and immediately closes the overlay on the UI thread.
v1.4.24patchMay 27, 2026

fix: Select Item card no longer shows all dropdown values after selection

Bug Fixes

  • Select Item action card now shows a clean label (HTML id / CSS selector) instead of all option values, both before and after selecting. Root cause: ElementName for <select> captures InnerText (all options concatenated); UIElement.Name was set from this raw value so CommitSelectItem re-read and displayed all options. Fixed by computing a clean display name from AutomationId or CssSelector and writing it to both action.Details and UIElement.Name before the picker is shown.
v1.4.23patchMay 27, 2026

fix: Select Item card shows clean name not all dropdown values; Debug Run enabled after recording

Bug Fixes

  • Select Item action card in recorder now shows a clean element name (AutomationId or first word of name, capped at 40 chars) instead of all dropdown option values concatenated. After the user picks an option, the card shows 'elementName -> selectedValue'.
  • Debug Run button was not re-enabled after a recording session ended. Root cause: _isRecording property was missing NotifyCanExecuteChangedFor for DebugWorkflowCommand, DebugOrPauseResumeCommand, and StepIntoCommand. When IsRecording flipped to false, only RunWorkflowCommand was notified. Now all three are notified so Debug Run and Step Into enable correctly after recording.
v1.4.22patchMay 27, 2026

feat: Select Item activity in recorder - indicate dropdown, pick option from inline panel

New Features

  • Select Item is now available in the recorder left panel activity list (dropdown icon).
  • Flow: click Select Item in left panel -> indicate the dropdown/combobox element -> recorder window reappears showing an inline dropdown picker populated with the available options -> choose or type the option -> press Enter or click confirm.
  • For web SELECT elements: options are automatically fetched from the element and shown in the picker.
  • For desktop ComboBox/ListBox: same flow; options shown if captured during indication.
  • If no options were fetched (dynamic/AJAX-loaded dropdown), user can type the value freely.
  • The generated Select Item activity has Value pre-filled with the chosen option and SelectBy=value.
  • Select Item is a manual insert-menu activity only - automatic detection on click has been removed to avoid confusion.
  • Previous v1.4.21 auto-detection of combobox clicks has been reverted.
v1.4.21patchMay 27, 2026

feat: recorder now generates Select Item activity for combobox/dropdown/select elements

New Features

  • Recording a click on a ComboBox, List, or ListBox (desktop) now produces a Select Item activity instead of a Click.
  • Recording a click on a <select> element (web) now produces a Select Item activity instead of a Click.
  • The generated Select Item activity has an empty Value field - fill it in the designer with the option to select.
  • SelectBy defaults to 'value' (matches by option value/text). Change to 'index' or 'label' in properties if needed.
  • All other click targets continue to produce Click activities as before.
v1.4.20patchMay 27, 2026

fix: self-healing uses captured ImagePath regardless of ImageEnabled checkbox

Bug Fixes

  • ImageEnabled=false no longer clears ImageTemplate from the target. ImageEnabled only gates the CV image-matching strategy (options.EnableImage). The captured element image (ImagePath) is always available to self-healing as visual context.
  • Previously: if ImageEnabled was unchecked, ImageTemplate was nulled out in ResolveTarget() so self-healing had no image to send to the AI, even though one was captured at indication time.
  • Now: self-healing reads target.ImageTemplate.Path unconditionally - if an image was captured it is sent to the AI regardless of whether the CV matching strategy is enabled.
v1.4.19patchMay 27, 2026

feat: self-healing now sends captured element image to AI for visual identification

Improvements

  • When SelfHealing is triggered and the activity has an ImagePath configured (captured during element indication), the element screenshot is now sent to the AI as a base64-encoded image alongside the broken selector.
  • The AI (AWS Bedrock / Azure GPT-4o) uses the image visually to identify the correct element in the current UI and propose a repaired selector - more accurate than selector text alone.
  • Image is attached in both primary AI backend path (TryBackendHealAsync) and the Direct Azure fallback path (TryDirectAzureHealAsync).
  • If the image file cannot be read (missing, locked, etc.) the repair still proceeds using text context only - non-fatal.
  • No UI changes required - ImagePath is already captured automatically at indication time and stored with the activity.
v1.4.18patchMay 27, 2026

fix: all targeting methods (Fuzzy, Anchor, Image, NativeText, SelfHealing) now strictly controlled by their checkboxes

Bug Fixes

  • All 6 targeting method flags are now correctly honoured: if unchecked they do NOT run; if checked they DO run. Previously FuzzyMatch, Anchor, and SelfHealing had hardcoded defaults that ignored the user's checkbox settings.
  • FuzzyMatch=False now disables fuzzy scoring - previously ForExecution() hardcoded EnableFuzzy=true so fuzzy always ran regardless of the checkbox.
  • AnchorEnabled=False now disables anchor-narrowing in the resolver - previously ForExecution() hardcoded EnableAnchor=true.
  • SelfHealing=True now actually triggers AI self-healing - was never wired to options.WithSelfHealing(); the checkbox had no effect.
  • ImageEnabled=True now correctly enables image/CV matching using the ImagePath captured at indication time. ImageConfidence from Targeting Methods overrides the value stored at indication.
  • NativeText and StrictMatch were already correctly wired (unchanged).
  • Added AnchorEnabled and SelfHealingEnabled properties to UnifiedTarget with JSON serialization.
  • ResolveOptions.ForExecution() defaults for EnableFuzzy and EnableAnchor changed to false - activity params are now the sole source of truth.
  • All changes apply consistently across: ResolveElementViaResolverAsync (used by all 8 unified activities), WaitForElement desktop resolver paths, and UnifiedAutomationService.FindAndValidateAsync.
v1.4.17patchMay 27, 2026

fix: StrictMatch no longer blocks fuzzy/image/self-healing fallbacks

Bug Fixes

  • StrictMatch now correctly controls only HOW the selector is matched (exact [attr=value] CSS vs flexible partial-class), not WHETHER fallback strategies run.
  • Previously v1.4.16 incorrectly called WithStrict() which disabled fuzzy, image, alternates, and self-healing when StrictMatch was on.
  • Now: StrictMatch=on tries the exact CSS selector first (no partial-class fuzzing in the CSS), then if that fails, falls through to Fuzzy -> Image -> SelfHealing as normal - user controls each independently.
  • Removed .WithStrict(target.StrictMatchEnabled) from ResolveElementViaResolverAsync and both WaitForElement desktop resolver calls.
v1.4.16patchMay 27, 2026

feat: complete StrictMatch and NativeText wiring across all unified activities

Bug Fixes and Improvements

  • StrictMatch now fully works on desktop too - sets StrictMatchEnabled on UnifiedTarget, which disables fuzzy/image/self-healing in ResolveOptions via WithStrict(). Previously only injected strictmatch=true into webctrl XML (web-only effect).
  • NativeText now fully works on desktop too - wires target.NativeTextEnabled into ResolveOptions.WithNativeText() for all resolution paths including the SelectorResolver fuzzy desktop strategy.
  • Fixed WaitForElement activity desktop paths that were creating UnifiedTarget.FromSelector() fresh, discarding all targeting method params (StrictMatch, NativeText, FuzzyMatch). Now correctly reuses the activity target which already has all params applied.
  • Added StrictMatchEnabled and NativeTextEnabled properties to UnifiedTarget model with JSON serialization.
  • All 8 unified activities now consistently propagate StrictMatch and NativeText from Properties panel through to the element resolver.
v1.4.15patchMay 27, 2026

feat: implement NativeText - OS accessibility text in desktop fuzzy matching

New Features

  • NativeText parameter is now fully implemented (was previously dead code - UI showed it but had no effect).
  • When NativeText=True (default), desktop fuzzy matching now also scores against HelpText (tooltip/description) and TextPattern text content alongside the primary UIA Name property.
  • This helps locate elements that lack stable IDs or class names but have readable accessible labels or tooltip text.
  • Full wiring: NativeText param -> UnifiedActivity.ResolveTarget -> UnifiedTarget.NativeTextEnabled -> ResolveOptions.EnableNativeText -> SelectorResolver.TryFuzzyDesktop -> HelpText + TextPattern scoring.
v1.4.14patchMay 27, 2026

fix: move StrictMatch above FuzzyMatch in Targeting Methods

Bug Fixes

  • Moved StrictMatch to the top of the Targeting Methods section (above FuzzyMatch) in the Properties panel.
v1.4.13patchMay 27, 2026

fix: Strict Match parameter now visible in Properties panel for Click activity

Bug Fixes

  • Fixed 'Strict Match' not appearing in the Properties panel for Click Element activity.
  • Root cause: StrictMatch was added to ClickElementActivity.GetParameters() (old browser-specific activity) but the Properties panel uses UnifiedClickActivity routed through AddTargetingMethodsGroup() in PropertiesViewModel.
  • Fix: Added StrictMatch boolean to the Targeting Methods group in PropertiesViewModel (same section as FuzzyMatch, AnchorEnabled, etc.) and moved StrictMatch injection to UnifiedActivity.ResolveTarget() so it applies automatically for all unified activities.
v1.4.12patchMay 27, 2026

feat: StrictMatch option for browser Click Element activity

New Features

  • Added 'Strict Match' option to Click Element (browser) activity.
  • When enabled, all selector attributes must match the element exactly using CSS [attr=value] equality.
  • Example: class='btn btn-primary' will NOT match an element with class='btn btn-primary spin' - useful for waiting until animations complete before clicking.
  • Implementation is fully dynamic - any attribute on the selector (id, class, type, data-*, aria-*, etc.) is enforced exactly without hardcoding per-attribute logic.
  • When Strict Match is on, no fallback selectors are tried - the selector must match as recorded.
v1.4.11patchMay 26, 2026

fix: browser re-indication attaching to wrong browser (Edge instead of Chrome)

Bug Fixes

  • Fixed re-indication of browser child activities attaching to the wrong browser (e.g. Edge) when Chrome was the Playwright session.
  • Root cause: LiveBrowserBridge.IsAttached was calling IsSessionActive() which blocks on the UI thread, deadlocks, and removes the live Playwright session as a side-effect. Subsequent CDP discovery then attaches to whichever browser responds first.
  • Fix: IsAttached now uses GetActiveSessionIds().Contains() - the same non-destructive check used elsewhere - so the Playwright Chrome session is preserved and re-indication always targets the correct browser.
v1.4.10patchMay 26, 2026

fix: recorder action ordering and blank container regression

Bug Fixes

  • Removed incorrect timestamp-based reordering of recorded actions - action order now matches exactly what the user recorded.
  • Fixed blank Launch Browser/Application container in designer - child activities (click, type, get text) now correctly nest under the parent container.
v1.4.9patchMay 26, 2026

fix: recorder and general activity improvements

Improvements

  • Replay button hidden from recorder overlay.
  • Added 'Wait for Element' and 'Get Cell Value' to the recorder Add Step list.
  • MaximizeWindow now defaults to true for Launch Browser and Launch Application.

Bug Fixes

  • Recorder now closes the browser it opened when the user clicks Done.
  • Fixed action ordering bug where recorded type-then-click appeared as click-then-type in designer.
  • Fixed blank Launch Browser card appearing when child actions were not captured.
  • Get Text no longer returns an XPath expression - extracts the actual text value instead.
v1.4.8patchMay 25, 2026

fix: Pro beta tier no longer gets Pro+ features

Bug Fixes

  • Beta mode now correctly restricts features to the user's actual tier. Pro beta users only get Standard and Pro features.
  • Pro+ exclusive features (voice input, agentic AI, dark theme, git integration, etc.) are now properly locked for Pro and Standard beta users.
  • IsFeatureAvailable and CanUse no longer bypass feature gates for beta accounts.
v1.4.7patchMay 25, 2026

fix: send button disabled when empty input; no upgrade warning on empty send

Bug Fixes

  • Send button in Copilot is now disabled when the input field is empty, preventing accidental clicks.
  • Clicking send with an empty input no longer shows a Pro+ upgrade warning. The button is simply disabled until the user types something.
  • Send button re-enables immediately as soon as the user starts typing.
v1.4.5patchMay 23, 2026

fix: Talk to AI correctly gated to Pro+ tier, account panel feature lists corrected

Fixes

  • Talk to AI (voice input in Copilot) is now correctly restricted to Pro+ and above. Pro users no longer see the Mic icon in the send button and clicking send with an empty message shows an upgrade prompt instead of trying to start voice.
  • Account panel Pro feature list no longer incorrectly shows 'AI Voice Mode (Say)' -- voice is a Pro+ exclusive feature.
  • Account panel Pro+ feature list now shows 'Talk to AI (voice input in Copilot)' as a key differentiator from Pro.
  • Standard to Pro upgrade hint no longer lists AI Voice Mode -- that belongs to Pro+ not Pro.
  • Pro to Pro+ upgrade hint now correctly shows Talk to AI as the main new unlock.
v1.4.4patchMay 23, 2026

feat: accurate AI provider tracking in telemetry (Bedrock vs Azure fallback)

Improvements

  • All AI telemetry events now report the actual provider used for each request. When AWS Bedrock is unavailable and Azure OpenAI handles the request as a fallback, the admin portal correctly records azure_openai instead of aws_bedrock.
  • Backend routes detect the actual provider per-call across all 5 AI features: workflow generation, copilot chat, decompose, verify, and selector repair.
  • docker-compose now correctly passes AWS Bedrock credentials into the container so Bedrock can be attempted before falling back to Azure.
v1.4.3patchMay 23, 2026

fix: AI Usage By Feature list now updates correctly after prompts and Refresh

Fixes

  • AI Usage By Feature breakdown now persists correctly across Refresh. Previously, clicking Refresh wiped the per-feature rows because the credits server does not yet return a feature breakdown in its balance response. The local accumulated counts are now preserved when the server omits that field.
  • By Feature rows update immediately after each AI prompt via local credit deduction tracking, with no need to wait for a full server fetch.
v1.4.2patchMay 23, 2026

fix: AI Usage Refresh button applies updated values immediately

Fixes

  • AI Usage Refresh button now applies the latest server values immediately. Clicking Refresh shows 'Refreshing...' then updates all rows and the By Feature breakdown in the same frame as the server response, with no queuing delay.
  • BalanceChanged event handler upgraded from BeginInvoke to InvokeAsync so background credit deductions also update the dashboard at normal dispatcher priority instead of background priority.
v1.4.1patchMay 23, 2026

fix: unified send/stop/mic button, AI usage voice tracking, TTS workflow summary

Improvements

  • Unified action button in Copilot: shows Mic icon when input is empty (click to start Talk to AI), ArrowUp when typing (click to send), and Stop when processing or voice is active (click to cancel immediately at any stage).
  • Talk to AI button removed from toolbar. Mic functionality merged into the send button to reduce clutter.
  • Voice language setting in Settings now applies instantly to Copilot without requiring a restart.
  • AI speaks only a brief summary when a workflow is generated, suggested, fixed, or repaired. The full written response is still shown in the chat bubble. Plain conversational answers are spoken in full.
  • Voice Recorder now deducts credits correctly, so Talk to AI interactions appear in the AI Usage By Feature breakdown.
  • AI Usage panel Clear button removed. Usage data is server-side and not user-clearable.
  • AI Usage last-refreshed timestamp updates automatically whenever the balance loads.

Fixes

  • Voice STT and TTS usage now tracked in ai.usage telemetry events sent to the admin portal.
  • Talk to AI voice credit deduction was missing. Each interaction now correctly deducts 2 credits.
  • AI Usage By Feature rows were always showing 0 for Voice Recorder. Fixed.
  • Do-release script parsing error caused by special characters in string literals. Fixed.
v1.4.0minorMay 23, 2026

feat: AWS Bedrock AI, Google Cloud voice, AI usage telemetry, Talk-to-AI UX overhaul

Backend

  • AWS Bedrock integration: added provider abstraction layer supporting runtime switch between AI providers.
  • Workflow generation and selector repair route through the new provider abstraction. Azure OpenAI preserved as an alternative provider.
  • Chat endpoint response extended with token usage and cost tracking fields.
  • Switch AI provider with one configuration change and backend restart. No code deploy needed.

Improvements

  • Talk to AI status bar is now clean: shows only Listening / Processing / Speaking states. The cramped Heard: ... truncation is gone.
  • If the user edits the transcript in the input box during the 2.2s silence window, their edited version is sent to AI instead of the raw STT output.
  • Stopping Talk to AI clears any unsent voice transcript from the input box automatically.
  • Chat response from the backend now includes token usage and cost fields for accurate per-call tracking.
  • Azure OpenAI fallback service now tracks usage on every successful call.
  • Voice provider name is included in every usage telemetry event.

New Features

  • AWS Bedrock (Claude Sonnet / Haiku) is now the default AI provider for all workflow generation, selector repair, chat, and copilot features. Switch back to Azure OpenAI at any time via backend configuration.
  • Google Cloud Speech-to-Text (gRPC streaming) is now the primary voice provider for Talk to AI and Dictate. Azure Speech SDK is retained as fallback and for always-on wake word detection.
  • AI usage telemetry: every AI call now emits a usage event to the admin portal with provider, model, token counts, cost, and latency.
  • Talk to AI UX overhaul: live speech transcript now appears directly in the chat input box (same as dictate mode). Interim words show in gray while you speak; confirmed words turn black. You can edit the text before the silence timer auto-submits it.
  • Voice usage is tracked separately per provider with character count and language for cost estimation.
v1.3.80patchMay 22, 2026

fix: recorder Talk to AI shows Starting... while Google STT stream auto-restarts between utterances

Bug Fixes

  • After the first voice command is processed, the Talk to AI button now shows 'Starting...' while Google STT auto-restarts its stream (500ms gap) instead of staying stuck on 'Listening...'.
  • Once the mic is active again (first partial audio arrives), the button immediately switches back to showing the live transcript.
  • Added StreamRestarting event chain: GoogleVoiceClient -> VoiceService -> RecorderVoiceService -> RecorderOverlayWindow so the UI is notified of every restart cycle.
v1.3.79patchMay 22, 2026

feat: recorder UI cleanup - remove hamburger, move browser picker, voice transcript in button, reduce spacing, remove timer

New Features

  • Browser picker moved into Launch Browser in Add Step list - select Chrome, Edge, Brave or Firefox before launching. Defaults to Chrome.
  • Voice transcript now shows directly in the Talk to AI button label (Starting... / Listening... / Processing... / spoken text) instead of a separate row - gains space for more activities.
  • Settings hamburger icon removed - it was unused.
  • Timer (00:00) removed from the Recorded Actions header - not needed for manual recording.
  • Bottom buttons and top controls row padding reduced to gain more vertical space for the activity list.
v1.3.78patchMay 22, 2026

fix: recorder Talk to AI status - shows Starting then Listening, clears on stop

Bug Fixes

  • Recorder Talk to AI: status now shows 'Starting...' while the mic is opening (1-5s for Google STT), then switches to 'Listening...' once audio capture is active.
  • Recorder Talk to AI: clicking Stop now immediately clears the transcript text so status does not remain stuck on 'Listening...' after stopping.
v1.3.77patchMay 22, 2026

feat: Check for Updates on Account page, progress status messages, fix recorder Talk to AI stop

New Features

  • Account page now has a Check for Updates button - no need to open a project to check for the latest version.
  • Progress status now shows 'Analysing your request...' while intent is being classified.
  • After classification: Question/Undo shows 'Processing your request', Edit shows 'Editing your workflow', Generate shows 'Building your workflow'.

Bug Fixes

  • Recorder Talk to AI: clicking Stop now immediately discards any in-flight utterance - voice processing stops even if an AI call was already in progress.
  • Added _isVoiceListening guard at the start of OnUtteranceCompleted and after each await to prevent stale utterances from being processed after stop.
v1.3.76patchMay 22, 2026

fix: AI fallback chain - AWS Bedrock primary, Azure GPT-4o secondary, Template last resort

Bug Fixes

  • Root cause: CopilotService only used AIRouter (Azure/Template) when no backend was configured at all - if backend was set but offline, it would just fail with no fallback.
  • Fixed: when AWS Bedrock backend is down/unavailable, CopilotService now falls through to AIRouter which tries Azure GPT-4o first, then Template.
  • Fixed: FetchAndApplyBackendCredentials now reads AZURE_OPENAI_ENDPOINT/KEY from local .env as fallback when AI_PROVIDER=aws (backend returns no Azure creds) - ensures Azure GPT-4o is always available as secondary.
  • AI chain is now: AWS Bedrock (primary) -> Azure GPT-4o (secondary, if backend down) -> Template (last resort).
  • Voice chain: Google Speech (primary) -> Azure Speech (secondary, if Google fails) -> disabled (no offline fallback).
v1.3.75patchMay 22, 2026

fix: Google Cloud voice works correctly; beta includes ProPlus features (voice recorder unlocked)

Bug Fixes

  • AIBootstrapper now checks VoiceService.IsEnabled directly instead of re-reading env vars - no more false NOT CONFIGURED log when Google Cloud Speech is active.
  • Beta license now includes ProPlus features (ai.voice_recorder, ai.agentic_automation, etc.) so voice recorder and other ProPlus features work during beta.
  • Google Cloud Speech setup: place service-account JSON in AppData\Genzbots and add VOICE_PROVIDER=google, GOOGLE_APPLICATION_CREDENTIALS, and GOOGLE_CLOUD_PROJECT to AppData\Genzbots\.env
v1.3.74patchMay 22, 2026

fix: Studio fails to launch when Google Cloud Speech assembly cannot be loaded

Bug Fixes

  • Root cause: VoiceService.Create called new GoogleVoiceClient() eagerly even with no Google credentials; a missing native gRPC dependency caused FileNotFoundException which propagated up and crashed startup.
  • Fixed: GoogleVoiceClient construction is now wrapped in try/catch in VoiceService.Create so a load failure gracefully falls back to Azure or no-op voice mode.
  • Fixed: IVoiceService DI registration now catches all exceptions from VoiceService.Create and returns a disabled no-op instance instead of crashing the application.
v1.3.73patchMay 22, 2026

feat: thread user identity (email + machine ID) through all AI service calls

Improvements

  • User email and machine ID now passed to GenerateDecomposedAsync, ClassifyIntentAsync, and ProcessMessageAsync.
  • SingleThreadStrategy forwards user identity to GenerateWorkflowAsync so all AI backend calls are attributed.
  • Training examples and intent classification logs now include the originating user and machine for better analysis.
v1.3.72patchMay 22, 2026

fix: beta license no longer bypasses server validation - force-kickout and tier changes now take effect immediately

Bug Fixes

  • Root cause: beta license short-circuited before calling the server, so SESSION_DEACTIVATED was never received.
  • Fixed: when beta is active and a JWT token exists, Studio now validates with the server first before applying the beta tier.
  • Admin force-relogin and tier upgrades now take effect within 15 minutes (the background validation interval).
  • No-session users (not logged in) still get beta immediately without a server call as before.
v1.3.71patchMay 22, 2026

fix: license validation improvements - HKCU registry, offline grace period, signed-out sentinel

Bug Fixes

  • License storage migrated to user-level storage so sign-out always succeeds without admin rights.
  • Transient server errors during license validation now apply an offline grace period instead of immediately clearing the session, preventing spurious logouts.
  • Only explicit deactivation events clear the license cache and require re-login.
  • A sign-out marker prevents stale license data from being re-read after explicit sign-out.
  • One-time migration: existing license data moved to user-level storage on first launch.
v1.3.70patchMay 21, 2026

feat: wire thumbs up/down feedback to backend training DB with user identity

Features

  • Thumbs up/down feedback now sent to AI backend training database (previously only stored locally).
  • Training records now include user identity for per-user analysis.
  • Workflow generation result ID is now surfaced through the generation pipeline for tracking.
  • Backend database schema updated to support user identity on training records.
v1.3.69patchMay 21, 2026

fix: pass user dropdown selection to intent classifier for accurate admin portal analytics

Bug Fixes

  • User Choice column was null in admin intent classification log. ClassifyIntentAsync was called without passing SelectedRequestType.
  • Fixed: CopilotViewModel now passes SelectedRequestType to ClassifyIntentAsync so keyword_wrong, gpt_wrong, and user_overrode flags are computed correctly on the server
v1.3.68patchMay 20, 2026

perf: UI element tree loads instantly via lazy on-expand loading; fix AI credit % display

Performance

  • UI element tree in Selector Editor now loads in under 1 second instead of 30-60+ seconds
  • Root cause: same O(N) desktop.FindAllChildren + Process.GetProcessById per window pattern that caused the 52s highlight hang
  • Fix: O(1) window lookup via Process.GetProcessesByName + automation.FromHandle instead of desktop walk
  • Fix: Only first 3 levels loaded eagerly (~50-100 elements); deeper nodes load on-demand when user expands
  • Fix: Chrome render widget search now uses bounded BFS (3 levels) instead of FindAllDescendants (full subtree)
  • Expanding a node with a placeholder child fetches its real children instantly from UIA on-demand

Bug Fixes

  • AI credits % in Copilot tab now shows actual Azure cost budget percentage (matching the AI Usage dashboard) instead of integer credit count percentage
  • Fixed garbled arrow characters in v1.3.53 release notes history
v1.3.67patchMay 20, 2026

fix: highlight no longer walks entire UIA desktop tree before Playwright (52s hang eliminated)

Bug Fixes

  • Root cause identified: HighlightElementAsync called ResolveBrowserWindowHandle which uses FindWindowElement (UIA desktop.FindAllChildren + Process.GetProcessById per window) taking 52+ seconds on busy desktops
  • Removed UIA-based foreground resolution from highlight path. Playwright BringToFrontAsync and tab activation handle browser foreground internally, same as the validate path which was already fast (22ms)
  • Added lightweight Process.MainWindowHandle foreground switch (O(1) vs O(N) desktop walk) inside the Playwright highlight path as a quick best-effort OS foreground nudge
v1.3.66patchMay 20, 2026

fix: SelectorEditorDialog and SelectorValidationService now use real Serilog loggers instead of NullLogger

Bug Fixes

  • SelectorEditorDialog and SelectorValidationService were created with NullLogger, so all timing and diagnostic logs from v1.3.65 were silently discarded
  • Both now resolve ILoggerFactory from DI to get real Serilog-backed loggers, so all Highlight timing logs will now appear in the log file
v1.3.65patchMay 20, 2026

diag: comprehensive timing logs across entire highlight path; reduced highlight timeouts

Performance

  • Reduced Playwright highlight element timeout from 5s to 2s (highlight is UI feedback, not automation)
  • Reduced tab activation timeout from 2s to 1s for highlight path
  • Result: worst-case Playwright highlight path now bounded to ~3s instead of ~10s

Diagnostics

  • Added Stopwatch timing logs to every stage of the highlight pipeline: button click, browser foreground, Playwright session lookup, CDP auto-attach, tab activation, ResolveLocatorAcrossFramesAsync, ScrollIntoView, JS inject, UIA fallback resolve
  • SLOW operation warnings (>500ms) logged for individual CountAsync and visibility checks inside the resolver loop
  • Every log line prefixed with [Highlight] or [Resolve] for easy log filtering
v1.3.64patchMay 20, 2026

fix: highlighter no longer blocks UI for 30+ seconds

Bug Fixes

  • PlaywrightBrowserService.HighlightElementAsync no longer awaits Task.Delay(safeDurationMs + 80) after injecting JS outline - returns immediately, JS setTimeout handles cleanup
  • SelectorEditorDialog Highlight button now uses a short 300ms delay for both UIA and Playwright paths instead of 2.2s asymmetric wait
  • Result: Highlight returns in ~300ms after element resolution; red outline still appears for full 2 seconds via browser JS
v1.3.63patchMay 20, 2026

fix: robust variable/expression resolver; Copilot file attachments (Excel, Word, drag-drop)

New Features

  • Copilot: attach Excel (.xlsx/.xls) and Word (.docx/.doc) files - content is extracted and sent as context to AI
  • Copilot: drag-and-drop files from Explorer onto the chat input to attach them
  • Excel attachments show first 3 sheets as pipe-delimited text preview (up to 50 rows per sheet)
  • Word attachments extract paragraph text for AI context
  • File dialog now shows Excel/Word categories alongside images and text files

Bug Fixes

  • DataRow column lookup now trims whitespace and normalizes spaces - fixes 'Last Name ' (trailing space) not matching 'Last Name' in Excel data
  • Calling .ToString(), .Trim(), .ToLower(), .ToUpper() on null values now returns empty string instead of throwing (null-safe method calls)
  • Assign activity: when expression evaluates to null (e.g. column not found), returns empty string instead of the raw expression text as a literal value
  • Bare variable references that don't exist in context now resolve to empty string instead of passing the variable name as literal text (e.g. Type Into no longer types 'strFirstName' literally)
v1.3.62patchMay 20, 2026

fix: strip dollar-brace variable format from all activity params; auto-register variables on edit path

Bug Fixes

  • Strip dollar-brace variable format from all activity parameters in NormalizeParameters - affects Type Into Text, Read Excel FilePath, and any other param where AI wraps a variable reference in dollar-brace notation
  • Edit/Add path now auto-registers output variables from newly added activities (mirrors Generate path's EnsureOutputVariablesRegistered). Applies to Get Text, Read Excel, Read CSV, HTTP Request, Execute Query, Assign, For Each, For Each Row, Launch Browser, Input Dialog, etc.
  • Added rule 10b to edit AI prompt: variable names must be bare (no dollar-brace wrapping). WRONG: dollar{downloadedFile}, CORRECT: downloadedFile. String literals use quotes. Concatenation: 'My Name is ' + strName
v1.3.61patchMay 20, 2026

Copilot: improved intent classification with user feedback loop and positional context awareness

Improvements

  • Intent classification now logs keyword result, AI result, AND user's dropdown selection - all three compared for mismatch tracking
  • New mismatch flags: user_overrode (user corrected AI), keyword_wrong, gpt_wrong - pinpoints which classifier needs improvement
  • AI classifier prompt rewritten: no longer blindly maps 'I want to...' to Generate. Now reads full sentence for positional signals
  • New CRITICAL RULES in AI prompt: 'after X activity', 'use the downloaded file', 'then read', 'for each row' all correctly route to Edit
  • Keyword classifier: added positional continuation patterns ('after the', 'use the downloaded', 'then read/loop/process', 'continue with', 'for each row')
  • Removed over-broad 'read excel and', 'for each row in', 'iterate over' phrases from new-workflow detection (were causing false Generate classification)
  • Removed post-AI 'I want to...' force-Generate override - user dropdown is now the only final authority on routing
v1.3.60patchMay 20, 2026

fix: request type dropdown now correctly overrides AI classification for Edit/Add requests

Bug Fixes

  • Request type dropdown (Generate / Edit+Add / Question) now correctly overrides the AI intent classifier. Selecting Edit/Add routes to the edit pipeline even when AI would have classified as Generate.
  • Undo intent from AI classifier is still always honoured regardless of dropdown selection.
v1.3.59patchMay 20, 2026

Copilot: request type dropdown, voice language in Settings, Talk to AI ProPlus-only

Improvements

  • Talk to AI feature is now completely hidden (not just disabled) for tiers below ProPlus - voice language label also hidden
  • Intent classification documentation generated at docs/intent-classification.md covering all keyword rules, AI prompt, override logic, and DB logging

New Features

  • Copilot chat now has a request type dropdown (Generate / Edit+Add / Question) - user explicitly tells AI what to do, bypassing auto-classification entirely
  • For Generate requests, compact workflow tree is now sent as context so AI knows what already exists on canvas and avoids duplicating steps
  • For Edit/Add requests, full workflow context (all properties, variables, conditions) is sent for maximum accuracy
  • Voice input language moved to Home > Settings > Voice input language - one global setting applies to Copilot and Recorder, no more per-panel selector
v1.3.58patchMay 20, 2026

AI Usage tab unlocked for all + recorder drag-to-reorder + visible scrollbar

Improvements

  • AI Usage panel is now accessible to all users - no longer gated behind a license tier

New Features

  • Recorder recorded steps can now be drag-to-reorder - grab the handle on the left of any step and drag it up or down to change execution order
  • Recorder actions pane now shows a visible scrollbar when steps overflow the visible area
v1.3.57patchMay 20, 2026

Fix highlight freeze - cancellation token propagated into CDP HTTP probes

Bug Fixes

  • Fixed highlight freezing for 30 seconds - CancellationToken now propagated all the way into HttpClient.GetStringAsync inside TryDiscoverCdpEndpointAsync so HTTP probes abort immediately when no CDP port is open
  • Fixed Task.Run wrapping anti-pattern - passing async delegate to Task.Run returned Task<Task<string>> not Task<string>, so WaitAsync only cancelled the outer wrapper while inner probes kept running for full 30s
  • Added cancellation checks between each CDP discovery strategy so port scan stops immediately after token expires
v1.3.56patchMay 20, 2026

Fix highlight freeze and first-time browser element screenshot

Bug Fixes

  • Fixed highlight freezing for up to 30 seconds - validation path auto-attach (CDP probe) now runs on background thread with 3s hard timeout, matching the highlight path fix in v1.3.55
  • Fixed first-time browser element indication capturing entire browser window - when CDP bounds are not yet available on first indication, cursor-region screenshot is used instead of falling through to UIA container bounds
v1.3.55patchMay 20, 2026

Fix highlight freeze and download filename

Bug Fixes

  • Fixed Highlight button causing Studio to freeze (Not Responding) - WMI scan and CDP HTTP probes now run on background thread with 3s hard timeout
  • Fixed downloaded file saved as 'download' instead of original filename - download attribute now set to filename extracted from href (e.g. challenge.xlsx)
  • Fixed anchor href filename extraction for relative URLs like ./assets/file.xlsx
v1.3.54patchMay 20, 2026

Save Browser Download fixes: element screenshot, highlight freeze, multi-strategy download

Improvements

  • Save Browser Download now uses 4 strategies to handle all download types: Playwright download event, direct anchor href with forced download attribute, JS fetch/blob for XHR-triggered downloads, and graceful error when all strategies fail
  • Download file extension auto-detected from MIME type when using JS fetch fallback (PDF, XLSX, CSV, ZIP, JSON, TXT supported)
  • Save path now auto-appends file extension from suggested filename when savePath has no extension

Bug Fixes

  • Fixed element screenshot capturing whole browser window instead of the indicated element - now uses CDP accurate bounds for browser elements
  • Fixed selector editor Highlight button freezing for up to 8 seconds when no Playwright debug port is open - auto-attach CDP scan now times out after 3 seconds
  • Fixed Save Browser Download highlight not working when no active recorder session is present
v1.3.53patchMay 20, 2026

Activity search shows category path + Save Browser Download fixes

New Features

  • Activity search results now show category path (e.g. UI Automation > Browser) below each result so you know where the activity lives
  • Save Browser Download added to activities panel under UI Automation > Browser
  • Save Browser Download added to recorder left panel activity list
  • Save Path field on Save Browser Download now has a folder browse button

Bug Fixes

  • Fixed indication result writing to wrong parameter name - activities with non-standard selector param names (e.g. TriggerSelector) now receive the selector correctly
  • Fixed Save Browser Download TriggerSelector not being captured on indication
v1.3.52patchMay 20, 2026

Recorder redesign: left activity panel + Launch Browser step

New Features

  • Recorder redesigned with always-visible activity list on the left panel - no more popup menu
  • Click any activity type (Click, Type Into, Get Text, Hover, etc.) to start element indication immediately
  • New Launch Browser activity in recorder - type/paste a URL and hit Enter to open browser and begin recording
  • Auto-detects existing Playwright browser session when indicating elements - prepends Launch Browser step with current page URL automatically
  • Record/Pause buttons hidden - indication-based flow is now the primary recording method

Bug Fixes

  • Removed click-blocking mouse hook (v1.3.50/51) - indication flow does not need it
  • Fixed duplicate Launch Browser containers when both URL and selector were captured
  • URL from Launch Browser step is now correctly applied to the workflow container
v1.3.51patchMay 19, 2026

Recorder: fix click blocking - recorder toolbar buttons (pause/stop) now work correctly

Bug Fixes

  • Recorder pause, stop and toolbar buttons were blocked by the new click-suppression hook - now fixed
  • Click blocking now uses the window HWND to detect recorder overlay clicks, not unreliable bounds
  • Clicks on any part of the recorder window (buttons, scroll, activity cards) pass through normally
  • Clicks on the target application being recorded are still blocked as intended
v1.3.50patchMay 19, 2026

Recorder: clicks are now blocked during recording so target app is not affected

Bug Fixes

  • Recorder now blocks left and right clicks at the OS level during recording. Buttons, links and form submits no longer trigger while recording
  • Click is captured and recorded as normal, but the action is swallowed so the page or dialog does not change
  • Pause recording first if you need a click to actually execute (e.g. to navigate to the next page to record)
  • Keyboard input and scroll are unaffected - only mouse clicks are suppressed
v1.3.49patchMay 19, 2026

UI polish: Copilot default panel, usage decimal, recorder click-through fix

Bug Fixes

  • Recorder: indication no longer passes click through to target app (blocking hook installed before window hides)
  • Studio launch now shows AI Copilot panel by default instead of AI Usage
  • Healing panel is hidden at startup and auto-shows only when healing events are present after execution
  • AI usage percentage in Copilot header now shows 1 decimal place (e.g. 1.6% instead of 2%)
v1.3.48patchMay 19, 2026

Scheduler: execution history recorded, status colors, hidden console window

Bug Fixes

  • Scheduled runs now appear in Execution History (TriggerSource=Scheduled recorded in DB)
  • Scheduler task list Status column now shows clean labels: Scheduled / Running / Success / Failed / Partial / Disabled
  • Status column colored: green=Success, red=Failed, blue=Running/Scheduled, grey=Disabled, amber=Partial
  • Last Run column now shows blank when a task has never run (was showing Jan 01 date)
  • Robot Console window is now hidden when launched by Task Scheduler (no console flash)
v1.3.47patchMay 19, 2026

Scheduler: force-update Robot Console on release to fix MSI skipping same-size DLLs

Bug Fixes

  • Release script now force-copies all Robot Console DLLs to the install directory after MSI build
  • MSI was silently skipping Robot DLL updates when file size happened to match (no version resource in .NET EXE stub)
  • GenzbotsRobot.dll (122880 bytes) and Genzbots.Core.dll are now guaranteed to be updated after each release
v1.3.46patchMay 19, 2026

Scheduler execution confirmed working - IAssetManager fix verified

Bug Fixes

  • Robot executor now correctly runs scheduled workflows end-to-end (tested locally: 2/2 activities completed)
  • IAssetManager (ProjectAssetManager) registered in Robot DI container - was causing crash before first activity ran
  • Asset manager SetProjectPath called with project directory so workflow assets resolve correctly
  • Browse dialog in Home panel Scheduler tab now shows Genzbots Project (*.gproj) filter first
v1.3.45patchMay 19, 2026

Scheduler: fixed execution - IAssetManager registered; top navbar menu removed

Bug Fixes

  • Robot executor crashed with 'Unable to resolve service for type IAssetManager' when running any scheduled workflow - IAssetManager (ProjectAssetManager) is now registered in the Robot's DI container
  • Asset manager SetProjectPath now called with project path so workflow assets resolve correctly
  • Removed Scheduler top-level menu from title bar (was added in v1.3.44) - scheduler stays in Home panel left sidebar only
  • Scheduler items restored to Tools menu
v1.3.44patchMay 19, 2026

Scheduler: fixed Browse dialog to show .gproj first; added Scheduler top-level menu

New Features

  • Added 'Scheduler' top-level menu in title bar (between Run and Tools)
  • Scheduler menu has 'Schedule Workflow...' and 'Manage Scheduled Tasks...' items
  • Removed scheduler items from Tools menu (now in dedicated Scheduler menu)

Bug Fixes

  • Browse dialog in Home panel Scheduler tab now shows 'Genzbots Project (*.gproj)' filter first (was always showing *.gbw - wrong dialog was being patched in previous releases)
  • Selecting a .gproj file now correctly sets ProjectPath and passes --project arg to GenzbotsRobot.exe
  • Section label updated to 'PROJECT / WORKFLOW', placeholder updated to 'Select a project or workflow...'
v1.3.43patchMay 19, 2026

Scheduler: fixed .gproj filter always shown first; auto-detects .gproj as project arg

Bug Fixes

  • Browse dialog now always opens with 'Genzbots Project (*.gproj)' filter selected (FilterIndex=1)
  • TaskSchedulerService auto-detects .gproj extension and uses --project arg even if ProjectPath not set explicitly
  • Prevents the old bug where selecting a .gproj via Browse generated --workflow flag instead of --project
v1.3.42patchMay 19, 2026

Scheduler: Browse now shows .gproj files first (forced clean rebuild)

Bug Fixes

  • Fixed Browse dialog in Schedule dialog not showing .gproj filter (cache issue in previous build)
  • Clean rebuild ensures Genzbots.Scheduler.dll correctly includes the .gproj file filter
v1.3.41patchMay 19, 2026

Scheduler dialog updated to project mode - shows 'PROJECT / WORKFLOW' section

Improvements

  • Schedule dialog title updated to 'Schedule Project / Workflow'
  • Section label changed from 'WORKFLOW' to 'PROJECT / WORKFLOW' to clarify project scheduling
  • Placeholder text updated to 'Select a project or workflow...'
  • Browse button now defaults to .gproj filter first, then .gbw
v1.3.40patchMay 19, 2026

Scheduler runs by project; AI Usage panel shows monthly budget only

Improvements

  • AI Usage panel removed credit quota bar - shows only Monthly AI Budget (actual spend vs tier cap)
  • Monthly AI Budget % now matches what admin portal shows under Monthly AI Budget card

New Features

  • Scheduler now schedules by project (.gproj) - runs the project main workflow via --project arg
  • Robot.Console accepts --project flag: resolves mainWorkflow from .gproj and executes it
  • Schedule dialog defaults to project mode when a project is open; Browse accepts .gproj or .gbw
v1.3.39patchMay 19, 2026

AI Usage panel: feature bar chart + two separate credit and cost-budget meters

New Features

  • AI Usage panel now shows a feature breakdown bar chart (Generate / Edit / Chat / Self-Healing / Voice / Home Chat) with credits used this month
  • Two separate progress bars: credit quota used (credit count %) and cost budget used (actual USD spend % of tier cap)
  • The two percentages match what admin portal shows - credit % is credits consumed, cost % is actual Azure OpenAI spend

Bug Fixes

  • Clarified why Studio showed 0.89% while admin showed 1.6% - they are different metrics (credit count vs actual USD cost)
  • Studio now shows both so users can see both dimensions of their usage
v1.3.38patchMay 19, 2026

AI Usage panel shows percentage with 2 decimal places

Improvements

  • AI Usage panel now shows credit usage as e.g. '1.65% used' instead of '1% used'
  • Useful when each request uses a small fraction of monthly credits
v1.3.37patchMay 19, 2026

Fixed Copilot tree-format response - orchestrator was never wired, all actions collapsed into one container

Bug Fixes

  • Fixed Copilot always showing one-liner response instead of full tree structure
  • Root cause 1: GenerationOrchestrator (SingleThreadStrategy) was never wired into CopilotService - was always null, so the proper tree-summary path was completely bypassed
  • Root cause 2: BuildWorkflowTreeSummary only checked ThenActions/BodyActions but Sequence/container actions store children in ChildActivities
  • Fix: orchestrator is now initialized in CopilotService constructor when AI backend is available
  • Fix: tree builder now correctly recurses into ChildActivities, ThenActions, BodyActions, and TryActions
v1.3.36patchMay 19, 2026

Dynamic version on login screen + consistent tree format in Copilot responses

Bug Fixes

  • Login screen now shows the actual installed version instead of hardcoded 'Version 1.0 Beta'
  • Copilot responses now always use tree-format summary with root nodes and branch lines
  • Previously the fallback/template path reported plain 'Workflow generated.' instead of the structured tree - now consistent with AI backend path
v1.3.35patchMay 19, 2026

Fixed sign-out re-login on machines with legacy HKLM registry entry

Bug Fixes

  • Fixed Studio still re-logging in automatically after sign-out on upgraded machines
  • Old versions wrote license to HKLM registry which cannot be deleted without admin rights - surviving token caused auto re-login
  • Fix: added signed-out sentinel file that blocks any surviving HKLM entry from being used after sign-out
  • Old HKLM data is migrated once to HKCU on first load, then ignored permanently after sign-out
v1.3.34patchMay 19, 2026

Fixed sign-out immediately re-logging in

Bug Fixes

  • Fixed Studio re-logging in automatically after sign-out
  • Root cause: license storage used HKLM registry which requires admin rights - delete silently failed, leaving old token behind
  • Fix: moved registry storage to HKCU (current user) so sign-out always clears the token correctly
v1.3.33patchMay 19, 2026

AI Usage panel live credits + Copilot generate fix

New Features

  • AI Usage panel now displays live credit balance (credits used, % remaining, reset date, feature breakdown)
  • Credits tracked per feature - generate, edit, chat, voice - shown in AI Usage tab

Bug Fixes

  • Fixed AI Usage panel always showing 0%.
  • Fixed authentication token verification silently failing on production due to a whitespace issue in the server configuration.
  • Removed self-healing stats from AI Usage panel - tab now shows only AI Credits meter
  • Fixed Copilot generate request modifying existing open workflow instead of creating a new Sequence
  • New workflow requests now wrap generated activities in a named Sequence appended to the canvas
v1.3.30patchMay 19, 2026

Scheduler and panel improvements

New Features

  • Scheduler status column is wider (160px) with full-text tooltip on hover
  • Scheduler minute picker now shows all 00-59 values for easier time selection

Bug Fixes

  • Fixed Windows Task Scheduler registration - interactive-token tasks now pass credentials correctly
  • Healing panel and AI Usage panel now visible by default when a project is open
  • AI Usage and Healing panels split into separate vertical panes (like Locals/Immediate layout)
v1.3.29patchMay 19, 2026

Improved Scheduler and AI Usage panel

New Features

  • Redesigned Scheduler page - full-width layout with header, inline form, and task list
  • AI Usage panel now shows Healing metrics on top and credit usage at the bottom

Bug Fixes

  • Fixed scheduled tasks failing to run - corrected Windows Task Scheduler registration
  • Healing and AI Usage panels now visible by default when a project is open
v1.3.28patchMay 19, 2026

Scheduler and AI improvements

New Features

  • Inline scheduler on the Scheduler page - schedule workflows without popup dialogs

Bug Fixes

  • Fixed scheduler task creation failing on standard Windows accounts
  • AI Usage tab now updates immediately after every AI request
  • Fixed one-time scheduled tasks silently not running when set to a past time
v1.3.27patchMay 19, 2026

Fix Studio auto-relaunch after update

Bug Fixes

  • Fixed Studio not relaunching automatically after an update completes
v1.3.26patchMay 18, 2026

Fix Studio relaunch after update and restore full release history

Bug Fixes

  • Fixed Studio not relaunching after update - relaunch now handled by C# process watcher instead of broken WiX custom action
  • Restored full release history (all 36 versions) to releases.json
v1.3.25patchMay 18, 2026

Add Check for Updates menu item and fix silent auto-update check

New Features

  • Added Check for Updates option in Help menu to manually trigger update check at any time

Bug Fixes

  • Auto-update check now correctly shows update dialog regardless of CheckForUpdates setting state
  • Manual update check shows You are up to date message when already on latest version
v1.3.24patchMay 18, 2026

Fix Studio not relaunching after update

Bug Fixes

  • Fixed Studio not relaunching after update using MSI commit action that fires for both fresh installs and upgrades
v1.3.23patchMay 18, 2026

Fix Studio not launching after update

Bug Fixes

  • Fixed Studio not relaunching after update install using MSI custom action instead of external bat watcher
  • UAC cancel during update no longer crashes Studio
v1.3.22patchMay 18, 2026

Fix update installer UAC cancel crash and post-update relaunch

Bug Fixes

  • Fixed Windows error dialog when UAC prompt is cancelled during update install
  • Studio now stays open if update install is cancelled instead of shutting down
  • Fixed post-update relaunch failing due to leading whitespace in bat file path variable
v1.3.21patchMay 18, 2026

Fix Studio not launching after update install

Bug Fixes

  • Fixed Studio not relaunching after update install on machines with spaces in Program Files path
  • Increased post-install wait to 3 seconds before relaunch to ensure file writes complete
v1.3.20patchMay 18, 2026

Fix AI intent classification for Pro/ProPlus users

Bug Fixes

  • Fixed AI copilot intent classifier not activating for Pro/ProPlus users after login
  • CopilotService now receives Azure credentials from backend on login (same as CloudAIService)
  • gpt-4o-mini classifier now correctly routes edit requests for installed builds
v1.3.19patchMay 18, 2026

Fix AI copilot intent classification accuracy

Bug Fixes

  • Fixed AI copilot incorrectly generating new workflow instead of editing existing one
  • Natural language edit requests now correctly detected even without explicit keywords
  • References to existing activities like in the for loop or after the read excel now route to edit
v1.3.18patchMay 18, 2026

Fix AI usage panel display and false quota warnings

Bug Fixes

  • Fixed AI quota warning incorrectly showing when there is no usage
  • Removed credit terminology from AI Usage panel
  • AI Usage panel now correctly shows usage percentage without false alerts
v1.3.17patchMay 18, 2026

Fix startup crash, session stability, and post-update relaunch

Bug Fixes

  • Fixed startup crash caused by invalid icon reference in AI Usage panel
  • Studio now reliably comes to foreground after post-update relaunch
  • Fixed session not being remembered across restarts in some scenarios
v1.3.16patchMay 18, 2026

AI Usage dashboard and improved session stability

Improvements

  • Added AI Usage panel with credit usage progress and feature breakdown
  • Improved session stability and reduced unexpected sign-in prompts on startup
  • AI copilot now uses smarter workflow context for more accurate responses
v1.3.15patchMay 17, 2026

Fix AI Copilot always connecting to localhost instead of production server

Bug Fixes

  • AI Copilot now always uses the production server in installed builds.
  • Local development configuration files no longer silently break AI for end users in installed builds.
  • A warning is logged when a development override is detected and corrected at startup.
v1.3.14patchMay 17, 2026

Full step-by-step diagnostic trace for AI Copilot connectivity

Improvements

  • Logs now written to %AppData%\Genzbots\Logs\ (was failing silently due to protected Program Files directory)
  • Each stage of the AI request flow tagged with a unique code (CP-01 through AB-06)
  • Error codes shown directly in chat bubble (ERR-01 through ERR-07) for instant identification
  • Full exception type, message, and stack trace logged at every failure point
v1.3.13patchMay 17, 2026

Diagnose AI Copilot connectivity: detailed error logging + startup health check

Improvements

  • AI Copilot now shows exact error details instead of generic temporarily unavailable message
  • Detailed connection diagnostics logged at startup (check Studio logs for AI health check results)
  • AI health check now fires at startup so availability is known before first use
v1.3.12patchMay 17, 2026

Fixed AI Copilot not working for beta and standard users

Bug Fixes

  • Fixed AI Copilot failing for users without a paid session token
  • AI Copilot now works for all license tiers including beta
v1.3.11patchMay 17, 2026

Diagnostic build to identify AI Copilot connection error

Improvements

  • Improved error diagnostics for AI Copilot connectivity issues
v1.3.10patchMay 17, 2026

Fixed AI Copilot session expired error

Bug Fixes

  • Fixed AI Copilot showing temporarily unavailable when session has expired
  • AI Copilot now shows a clear sign-in prompt when your session needs renewal
v1.3.9patchMay 17, 2026

Fixed Studio double-launch after update and AI Copilot not sending requests

Bug Fixes

  • Fixed Studio briefly opening then closing itself after auto-update
  • Fixed AI Copilot not connecting after sign-in on fresh installations
v1.3.8patchMay 17, 2026

Fixed auto-update relaunch and AI Copilot connectivity

Bug Fixes

  • Fixed Studio not reopening after auto-update completes
  • Fixed AI Copilot failing to authenticate after sign-in
v1.3.7patchMay 17, 2026

Fixed AI Copilot connectivity and error tracking

Bug Fixes

  • Fixed AI Copilot connecting to wrong endpoint on some installations
  • Improved error tracking for faster issue resolution
v1.3.6patchMay 17, 2026

Improved AI Copilot messaging and reliability

Improvements

  • AI Copilot shows clearer messages when the service is unavailable
  • Improved AI service connection reliability
v1.3.5patchMay 17, 2026

Fixed auto-update installation on Windows

Bug Fixes

  • Fixed update installer failing with an error on some Windows configurations
  • Fixed update prompt appearing repeatedly after updating
v1.3.4patchMay 17, 2026

AI Copilot now connects to cloud backend

Improvements

  • AI Copilot powered by cloud backend for faster and more reliable workflow generation
  • Improved stability and performance for AI features
v1.3.3patchMay 17, 2026

Improved AI Copilot authentication and security

Improvements

  • AI Copilot now uses your account session for all requests
  • Improved security for AI features
  • AI status indicator now reflects your sign-in state accurately
v1.3.2patchMay 17, 2026

Fix activities becoming unresponsive after moving between containers

Bug Fixes

  • Activities moved out of containers (Try/Catch, Sequence, If/Else) now remain fully editable
  • AI Copilot-generated activities also render correctly after being repositioned
v1.3.1patchMay 16, 2026

Update button shows busy state; Studio auto-relaunches after install

Bug Fixes

  • Update Now button now shows downloading/installing state immediately on click
  • Studio now relaunches automatically after the update installs
v1.3.0minorMay 16, 2026

Update button shows downloading state and Studio relaunches automatically after install

Improvements

  • Update Now button changes to Downloading... and greys out immediately on click
  • Button text updates to Installing... when MSI launches
  • Studio automatically relaunches after the installer finishes
  • Update dialog cursor changes to arrow when button is disabled
v1.2.9patchMay 16, 2026

Fix AI-generated activities becoming unusable after move/cut/paste in designer

Bug Fixes

  • Activities added by AI Copilot (Try Catch, Switch, Parallel, ElseIf) no longer lose their children when moved or cut/pasted
  • InitializeActivityDefaults was unconditionally overwriting CatchBlocks, SwitchCases, ParallelBranches and ElseIfBranches on every paste, wiping AI-generated content
  • Fix: collections are only initialized when empty/null - existing children are preserved
v1.2.8patchMay 16, 2026

AI Copilot workflow generation fix

Bug Fixes

  • AI Copilot now reliably generates workflows on fresh installs
v1.2.7patchMay 15, 2026

AI Copilot status indicator fix

Bug Fixes

  • AI Copilot status dot now correctly shows Online when AI is available
v1.2.6patchMay 15, 2026

AI Copilot connectivity improvements

Bug Fixes

  • AI Copilot now connects correctly on all machines
  • Version displays correctly as 1.2.6 throughout Studio
v1.2.5patchMay 15, 2026

Fix AI Copilot offline and version display after fresh install

Bug Fixes

  • AI Copilot now connects correctly after fresh MSI install
  • Encrypted credentials reliably bundled in installer (inline encryption fix)
  • Version displays correctly as 1.2.5 throughout Studio
v1.2.4patchMay 15, 2026

Bug fixes and improvements

Bug Fixes

  • Bug fixes and improvements
v1.2.3patchMay 15, 2026

Bug fixes and improvements

Bug Fixes

  • Bug fixes and improvements
v1.2.2patchMay 15, 2026

Bug fixes and improvements

Bug Fixes

  • Bug fixes and improvements
v1.2.1patchMay 15, 2026

Bug fixes and improvements

Bug Fixes

  • Bug fixes and improvements
v1.2.0minorMay 14, 2026

Initial public release

New Features

  • Genzbots Studio initial public release

Want to stay up to date? Follow our blog for deep-dives on every major release.

Read the docs →