Release v0.115.0
Sep 21, 2026
Network devices now map themselves into a live topology diagram, alerting gets one unified delivery system, and disk cleanup finally frees real space — alongside billing profiles, custom sending domains, and a broad reliability and security pass.
Network Monitoring & Topology
- Network devices now assemble themselves into a live, logical site map built automatically from what your agents already see — interfaces, routes, and neighboring devices — no manual diagramming required.
- Shared configuration templates can be applied across many sites at once instead of configuring each site by hand.
- On-demand ping and traceroute-style diagnostics run right from the map.
- Needs an agent on v0.115.0 or newer for full effect — older agents keep working, they just won't feed the map.
Alerting & Monitors
- Alert delivery is now one unified system: a single Delivery page (Channels, Routing, and Escalation policies) replaces the old 'send to every channel that happens to be enabled' behavior, with a preview so you can see exactly who gets notified before you save a rule.
- Creating alert rules the old way is now frozen in favor of Monitors, with a one-click 'Create monitor' shortcut and a Recommended list to help you migrate. A converter can preview, check equivalence, and convert (or revert) your existing legacy alert rules to monitors.
- Fixed a network-check monitor that could raise a duplicate alert, or none at all if its alert device was offline — it now raises exactly one alert, reliably.
- Fixed an HTTP-check monitor ignoring its configured expected status code.
- Fixed a restart-service monitor action losing its restart configuration after being saved.
- Fixed monitors of kind 'offline' never firing.
- Fixed role and OS filters on monitor policy assignments being ignored.
- Fixed editing a script-check monitor sometimes silently losing the selected script.
- The failure ceiling before a service/process/network-check monitor stops retrying was raised from 20 to 100 consecutive failures.
Disk Cleanup
- Cleanup now permanently deletes files instead of moving them to an agent-local trash folder, so the space you reclaim is actually freed.
- Disk cleanup now scans and cleans any fixed volume on a device, not just the system drive.
- New OS-native cleaners: Windows Disk Cleanup and DISM component cleanup, macOS local snapshots and Homebrew cache cleanup, and Linux package caches and journal trimming — all reachable from the Disk Cleanup tab and from a new AI cleanup tool.
- The Disk Cleanup tab is now a full select, execute, and history workflow, and previews show exactly how many bytes will be reclaimed.
- Needs an agent on v0.115.0 or newer.
Billing Profiles
- One new Rates screen replaces scattered legacy category and organization rate fields.
- Time entries can now be tagged with a Work Type, and billed time reflects per-profile minimums and rounding rules.
- ⚠️ Every partner's existing labour-rate setup is automatically converted into billing profiles during the upgrade — this can change what's billable for existing customers. See the upgrade notes below before updating.
Partner Sending Domains & Email
- Send customer-facing email — quotes, invoices, portal invites, ticket notices — from your own domain instead of a shared Breeze address, with DNS verification, a test send, delivery health tracking, and automatic suspension if bounce or complaint rates spike.
- Quote, invoice, portal-invite, and ticket-notice emails now render as properly laid-out, branded HTML instead of plain-text stubs.
- Off by default — nothing changes for existing partners until a sending domain is configured.
Bare-Metal Recovery & Backup
- Restore-as-VM and disaster-recovery rehearsal/failover can now target a VHDX image directly, including Linux guest images.
- Fixed cross-snapshot incremental recovery failing partway through after the target disk had already been wiped.
- Whole-machine restore and rebuild jobs now get a 24-hour time budget instead of being killed at 60 minutes.
- Fixed a system-image rebuild that could report 'completed' without actually restoring any OS state.
- Fixed incremental backups silently running as full backups every time due to a scheduling bug — they now correctly run as small deltas.
- Fixed retention being able to delete a full backup that an active chain's differentials still depended on.
- Fixed backup jobs that fail server-side validation hanging in 'running' forever instead of failing visibly.
- Fixed larger file uploads (16 MB–100 MB) to MinIO/S3-compatible storage sometimes failing, and a partial backup run incorrectly verifying as fully passed.
- The file Restore Wizard is now reachable from the UI instead of requiring a raw API call.
- Windows bare-metal restores now preserve Hidden, System, and Sparse file attributes.
- DR execute failures now name the actual missing prerequisite instead of a generic error, and saving a DR plan can no longer half-apply.
- Closed two disaster-recovery race conditions that could resurrect a stale, already-finished recovery job.
Security & Access
- Moving a device to a new organization now requires an interactive login session, and, if you have two-factor authentication turned on, a fresh step-up verification. See the breaking-change note below.
- New partner accounts now require two-factor authentication by default. Existing partners are unaffected, but admins who have it off will see a reminder banner.
- Admin → Users now shows an MFA status column (Enrolled, Pending, Overdue) for every user.
- Script parameter values (in bash, PowerShell, and Python) are now always treated as inert data on the agent, never as executable code — closing a class of script-injection risk.
- Helper (device) authentication now properly enforces the device-credential lifecycle, and requesting a change to partner trust settings now requires partner-wide manage access.
AI & Automation
- 13 new read-only AI/MCP tools close gaps in what the AI assistant can see about your fleet, and an updated MCP protocol adds version negotiation and structured results.
- Fleet Designer can now be turned on with one click instead of a manual detour through Settings → AI Agents.
- Fixed a race that could let two concurrent AI requests both slip past a budget check before either was charged.
- Fixed an AI-agent-proposed fleet change, once approved by a technician, being mis-attributed to the AI instead of the approving human.
- Fixed a newly created patch schedule firing a backdated catch-up run before it was even created.
- Organization-scoped roles can now be granted a dedicated permission to open AI chat — previously some roles held every other AI permission but were still blocked from chat entirely.
Everyday Fixes
- Fixed site and organization picker dropdowns silently truncating around 50–100 results — most importantly in the 'move device' dialog, where a target site past that limit couldn't be selected at all.
- Fixed SNMP GetMulti hiding packet-level errors behind a blank result.
- Fixed network discovery falling back to SNMPv2c instead of using your configured SNMPv3 credentials.
- Changing an SNMP template now polls immediately instead of waiting for the next interval.
- Fixed the network proxy breaking out of the proxied session on root-relative and same-host links.
- The remote desktop launcher now explains why a session was revoked instead of silently dropping to idle.
- Login now fails closed with a clear error instead of a server error if Redis is briefly unavailable during the MFA check.
- Fixed the QuickBooks mapping picker not showing a newly created organization's name until reload.
- Fixed clicking an organization's name on the Organizations board sometimes navigating away instead of opening its details.
- Backup verification failures now show the actual reason instead of a bare 'failed'.
- Automations can no longer be created requesting a newly elevated run-as privilege; existing elevated automations are now flagged with a badge.
- Script library imports can now target 'All orgs' (partner-wide), not just a single organization.
- Fixed the report edit page being capped to a narrow width on wide screens.
- Fixed the remote tool launcher failing on devices with a numeric custom-field identifier.
Network devices now build their own map. Using nothing but what your agents already collect — interfaces, routes, neighboring devices — Breeze assembles a live, logical topology diagram for every site, with shared configuration templates you can push across many sites at once and on-demand ping/traceroute diagnostics right from the map. Alerting gets its long-overdue consolidation too: one Delivery page now controls Channels, Routing, and Escalation policies, with a preview that shows exactly who gets notified before you save anything — replacing the old behavior where an unconfigured alert quietly fanned out to every enabled channel. And Disk Cleanup finally does what it says: cleanup now permanently deletes rather than moving files to a trash folder, so the space you reclaim is real, and it covers every fixed volume plus new OS-native cleaners for Windows, macOS, and Linux.
Behind those headline features, this release also ships Billing Profiles (one Rates screen instead of scattered legacy rate fields), partner sending domains (customer email from your own domain instead of a shared Breeze address), and a substantial bare-metal recovery hardening pass — including a VHDX-based Restore-as-VM path and a fix for cross-snapshot incremental recovery that could previously fail after the target disk was already wiped. A long tail of monitor, SNMP, backup, and AI-assistant fixes rounds things out.
Two things to know before you upgrade, if you self-host: moving a device between organizations now requires an active login session and, with two-factor authentication on, a fresh step-up verification — this only affects direct API/script callers, not the web app, mobile app, or AI tools. And Billing Profiles automatically converts every partner’s existing labour-rate setup on upgrade, which can change what’s billable for some customers — run the dry-run report first. Full detail in the GitHub release notes.