OpenFrame v0.9.52 - Mingo Everywhere, Fae Customization & Time Tracking

Mingo is no longer confined to the ticket section - it now lives in a persistent sidebar accessible from anywhere in the UI with full page-context awareness. New 'Generate with AI' one-click shortcuts appear on every form and editor, and Mingo can now draft tickets, docs, runbooks, and client emails. An expanded API toolset lets Mingo take action across PSA, RMM, docs, and integrations, making it the primary way techs interact with the platform rather than a standalone feature.

MSPs can now brand and configure Fae on a per-tenant basis - covering custom assistant names, per-tenant theming, custom intents, and configurable guardrails. The same Fae engine can serve a law firm and a manufacturing client with completely different personalities, workflows, and end-user permissions, making the assistant feel like a native part of each MSP's service rather than a generic widget.

Technicians can now log time against tickets, projects, and clients directly inside OpenFrame. A global timer in the top navigation bar supports start, pause, and confirm controls with a dropdown showing recent entries. A dedicated 'My Time' page and a global Worktime table in Settings give full visibility across all technicians. Time entries feed directly into the native billing module, eliminating CSV exports and third-party tools like Toggl or Clockify.

A unified notification center is now built into the OpenFrame admin interface. Covers ticket status change history with inline status blocks visible to clients, unread message tracking in chats, and system-level alerts for admins about activity across tickets, devices, and the platform. Delivered via NATS for live UI updates and GraphQL for catch-up reads, with support for desktop OS notifications, a bell indicator, and approval flows across both OSS and SaaS distributions.

A self-serve portal where MSPs can open support tickets with the Flamingo team and walk through onboarding guides independently. Covers ticket submission, status tracking, a searchable knowledge base, and step-by-step setup walkthroughs for new customers - replacing the need to dig through Slack, Discord, or email for answers.

TacticalRMM is replaced by a Flamingo-native, open-source RMM agent built directly into OpenFrame. The agent owns the full endpoint stack - monitoring, scripts, patching, remote access hooks, and reporting - with no third-party vendor dependency. Ships open source so MSPs can audit, fork, and self-host.

Fixed a critical loop where closing the Fae prompt would immediately re-trigger it for multiple users on the same tenant. The only previous workaround was rebooting the machine. Dismissing Fae now persists correctly and does not re-prompt.

Fixed an issue where the message input field retained its content after a message was sent, requiring users to manually clear it before typing a new message.

Fixed a bug where clicking 'Resend Verification' on the Email Not Verified modal returned a 409 status error, preventing users from verifying their email address.

Fixed the connection lost message displaying inside the message input area, which blocked users from copying or editing their prompt. The message now appears as a toast notification. Also restored the colored connection status dot (green/red) next to the tenant name.

Fixed a bug on Logs and Scripts pages where characters typed during result loading would disappear. Also resolved laggy/stuck search behavior on Tickets and Monitoring pages where every keystroke caused a full page navigation.

Fixed the agent list in the device tab changing order randomly on each page load or navigation. The list now maintains a consistent, predictable order.

Fixed an issue where a newly created customer did not appear in the customer list until the page was manually refreshed. The list now updates automatically after successful customer creation.

Fixed missing filter options in the sort and filter modal on the Logs page.

Replaced the incorrect component used in the 'full info' block on the Log Details page with the correct design-specified component.

Fixed the tag dropdown on the article creation/edit page to fetch all available tags rather than only those already linked to articles, ensuring newly created or unlinked tags appear as selectable options.

Fixed an issue where changing the LLM Provider or Provider Model in AI Settings did not immediately update the model indicator in the Admin Chat window. The model now reflects the new configuration as soon as settings are saved.

Fixed a bug in Fae chat where the user's profile image was returned as null, causing a broken or empty avatar display.

Fixed missing hover preview behavior for quick actions in Fae. Hovering over a quick action now displays the prompt preview, consistent with behavior on all other chat clients.

Fixed the AI settings preview/example section not reflecting the updated assistant name as the user types it. The preview now updates dynamically.

Fixed a flash of default appearance (light theme and default pink accent) on chat desktop start or reload before the user's configured theme was applied. Appearance settings are now persisted locally and applied before first paint.

The preview screen on the AI customization page caused visual inconsistencies in Safari due to font rendering bugs. The preview section is now hidden for Safari users until the underlying font issues are resolved.

Fixed the auto-provisioning option being unavailable or not rendering correctly on the SSO Settings page.

Fixed a layout shift where the page title block would jump or shift position when navigating between pages or when content was loading. The title area now maintains consistent height and position regardless of loading state.

Removed the doubled spacing gap between the search bar and table on Scripts, Scripts v2, and Notifications pages. Standardized tab-page behavior across Scripts and Monitoring to match the rest of the platform, and ensured the page title block maintains consistent height whether or not action buttons are present.

The new Mingo sidebar is built by extending the existing Global AI Ask component as its foundation - no forking or duplication. Both Mingo Guido and OpenFrame Mingo surfaces share the same abstracted sidebar behavior and are pixel-perfect design-aligned across all states: collapsed, expanded, empty, and loading.

Users can now attach OpenFrame entities (Devices, Scripts, Tickets, Customers, Users, Knowledge Articles, Policies, Queries) to Mingo messages via @-mention picker. Mingo agent responses render these as rich entity chips - icon, resolved label, and clickable link - instead of raw @marker:id tokens.

A page context tag in the Mingo chat interface shows users exactly what page context is being passed to the assistant. As users navigate the app, the frontend automatically accumulates context items from visited pages and sends them alongside each chat message, giving Mingo richer situational awareness without manual input.

A dedicated Mingo section in AI Settings now covers LLM provider selection, model selection, answer style, and quick actions configuration - all in one place.

Admins can now configure independent quick action sets for Fae and Mingo. Each action includes a name and a custom AI instruction defining its behavior, giving MSPs precise control over the shortcuts available to end-users and technicians.

MSPs can now configure a distinct AI assistant profile for each customer organization - including custom instructions, tone, knowledge base scope, branding, and available tools/actions. Customers can use the default appearance or a fully custom one, visible on the Customer Edit and Customer Details pages.

Configuration changes made in the management portal (behavior toggles, permissions, appearance) are now correctly propagated and applied to the running OpenFrame/Fae application on endpoints in real time.

The Fae client app is renamed to 'OpenFrame' across all platforms and materials. Since the AI assistant name is now customizable per tenant, the app name is decoupled from the assistant identity.

The welcome screen copy is updated to dynamically use the configured AI assistant name: 'Meet [AI Assistant Name], your AI IT assistant. Fixes what it can right away, and hands off the rest to your technicians.'

Fae is now enforced as a single instance application, preventing multiple simultaneous instances from running on the same device.

The Fae Chat initialization screen now removes the default Quick Actions list on load and replaces it with skeleton loaders for Chat Lists, AI Model, and Quick Actions - providing a smoother, less jarring startup experience.

Users can now rename Mingo chat sessions with custom descriptive titles for better organization, and archive dialogs to keep the main list clean. Archived chats remain accessible; deleted chats are permanently removed.

When Mingo operates in device context, it now filters to only Active and Offline devices, excluding Archived, Pending, and other non-relevant statuses to reduce noise and irrelevant suggestions.

In the collapsed state of command execution blocks, raw command text is replaced with a human-readable description of what the command does, making Mingo responses more accessible to non-technical users.

On both global AI Settings and per-Customer AI settings pages, the avatar area now falls back to the default AI assistant image when no custom image has been uploaded, preventing broken or empty avatar states.

All main platform sections (Customers, Devices, Scripts, Script Schedules, Policies, Checks, Queries, Logs) now have extended empty state placeholders with section descriptions and an 'Ask Mingo about...' button that triggers Mingo's Guide Mode for contextual onboarding assistance.

A new Worktime section in Settings provides a global table of all time entries across all technicians, including a Customer column. The timer widget, manual entry modal, and edit entry modal all support customer field assignment, with entries automatically linked to the associated ticket's customer.

MSPs can now edit their organization profile - including company name, avatar, and website URL - directly from the Settings section.

Registration now validates email and domain fields in real time with debounced on-keystroke checks, showing inline feedback for already-registered emails or taken/reserved domains before the user attempts to submit.

The 'Add Device' flow now requires an organization to be selected before proceeding. The option is hidden or disabled when no organizations exist, preventing orphaned device registrations.

The header block now displays the MSP's company name instead of the raw tenant domain, providing a more professional and recognizable identity across the admin interface.

The welcome screen in the OpenFrame client app now displays organization information so users can confirm which organization they are logging into or have been invited to.

The Dashboard now shows status counter blocks for all device states - Active, Offline, Pending, and Archived - each visually distinct and enabling quick filtering or navigation to the corresponding device list.

The Device Details page is redesigned to surface all available Fleet data, including visual representations of assigned policies, checks, queries, scheduled scripts, tickets, and knowledge base articles associated with each device.

A new Assigned Devices tab on the Query Details page lists all devices the query is assigned to, consistent with the pattern used on Policy and Check detail pages.

OS platform values in the Policy List are now displayed with their respective OS icons (Windows, macOS, Linux) instead of plain text.

Three informational warning blocks are added to the New Device screen: an administrator privileges warning with platform-specific instructions, an antivirus false-positive warning, and a platform-specific installation note. Padding inside all warning blocks is corrected to match design specifications.

Ticket cards now display relative timestamps following a clear display logic: 'Just now' for under 1 minute, 'X min ago' for 1-59 minutes, 'X hour(s) ago' for 1-23 hours, and MM/DD/YYYY for 24+ hours. A full timestamp tooltip appears on hover.

The Settings page now shows whether the user's email is verified and includes a resend verification email button for unverified accounts.

Navigation buttons inside settings blocks now have proper hover states consistent with the design system.

Fields with no data now consistently display '-' in gray (#888888) across all tables and detail pages, replacing inconsistent placeholders like 'No Description' or 'N/A'.

The Help Center section in OpenFrame admin is updated to reflect the current structure on hub.openframe.ai, including Onboarding Guides, Development Roadmap, Bug-fixes and Enhancements, Product Releases, Help Center tickets, Privacy Policy, and Terms of Service.

The device window listing all running agents now displays a 'Signed by Flamingo' label/badge and the Flamingo fork URL link for each agent, improving transparency and trust verification.

A tags filter modal is now available on the mobile view of the Knowledge Base page, bringing mobile filtering to parity with the desktop experience.

The deprecated 'Tier' column has been removed from the Customers table following the deprecation of the tier parameter in the latest system update.

Monitoring, Scripts, Schedules, and Logs pages have been migrated from the legacy ListPageLayout table to the unified PageLayout component, standardizing layout behavior across the platform.

Incorrect paddings in the details info block on the Policy Details page are corrected. Device tables on the Policy Details page now include filter controls.

The logout flow in OpenFrame is now fully designed and implemented, covering the logout button placement in main settings, a confirmation modal, and the post-logout screen or redirect.

HubSpot synchronization is expanded to include all users of a tenant, not just tenant owners, ensuring complete CRM visibility across the customer base.

All Terraform modules are updated to Terraform 1.14 compatibility. Shared microservices Helm charts are consolidated into a single Shared Helm Chart with a fully restructured values schema, and an AppSet for Shared Microservices is introduced pointing to the GitOps fleet repo.

Services requiring zero downtime are now identified and protected from eviction during node scaling events, improving platform reliability.

IAM configuration for tenant runners is implemented, improving security isolation and access control across tenant workloads.

The mongodb-meshcentral application has been removed from the production environment as part of the ongoing migration away from MeshCentral dependencies.

RegSync is configured to mirror new Grafana Community images and Helm charts, keeping the internal registry current with upstream releases.

Fixed robots.txt returning an inaccessible response on openframe.ai host variants. The file now returns HTTP 200 with correct Content-Type on all host variants, resolving an Ahrefs-flagged indexing risk.