New Features

• A command palette (⌘K on macOS, Ctrl+K on Windows and Linux) opens from anywhere in the app for quick keyboard access to instances and bridges by name, plus the top-level pages. Start typing to filter, then press Enter to jump straight to the match instead of scanning the sidebar or lists

• An S7comm bridge sample preconfigured for a test simulator, for trying a Siemens S7 read flow without a physical PLC
• Behind the Historian and write-flow feature flags, a new Historian bridge template writes a Unified Namespace data contract into TimescaleDB. Pick it under the TimescaleDB vendor in Add Bridge, set the database host, credentials, and data contract name, then deploy. The bridge stores every UNS message matching that contract into TimescaleDB hypertables and drops the rest, so you only choose the contract — there is no topic regex to keep in sync. Requires PostgreSQL 16+ with TimescaleDB.

New Features

• Preview: set up a TimescaleDB Historian for an instance from the new Plugins tab. Add the Historian plugin, then enter the TimescaleDB connection — host, port, database, user, password, and SSL mode — to create, edit, or remove it. The tab appears on instances that support it.

Improvements

• In a standalone data flow, a data contract that doesn't exist on the instance no longer blocks Save & Deploy. The editor still underlines the unknown contract, and a dialog now names the missing contracts and asks you to confirm before deploying, so you can go ahead when you plan to create the contract later.

• The code editor now draws vertical indentation guide lines, so nested YAML and JSON are easier to follow and a deeply indented config no longer reads as empty.
• The code editor gained Collapse all and Expand all buttons, so you can fold every section of a long YAML or JSON config at once instead of clicking the gutter arrows one by one.

• The code editor now shows a breadcrumb bar with the path to wherever your cursor sits, so you can tell where you are in a long YAML or JSON config at a glance. Click any part of the path to jump straight to it. Collapsed sections open automatically. The bar stays in place while you edit, and list positions read as "1st item", "2nd item", and so on.

Fixes

• Creating a bridge from a template now uses that template's processing setup. Previously the Conditions tab could open with example rules the template never defined, and those examples were deployed with the bridge.

• On the home page's "Latest UMH Learn Articles" feed, the external link icon no longer floats away from the text when a post title wraps across multiple lines — it now sits right after the title's last word instead of centering against the whole wrapped block. The icon is also no longer shown on feed thumbnail images, where it didn't fit well.

• Deleting a second instance right after deleting a first one no longer shows an empty dialog that fails to delete it. Previously, the dialog carried over leftover state from the first deletion, so the next delete attempt required reloading the page to work.

New Features

• Preview: an instance can now store a connection to an external TimescaleDB or PostgreSQL historian database (host, port, database, login, and TLS mode), which can be created, viewed, updated, and removed. Storing the connection does not yet route any data to it

Improvements

• The full bridge and stream-processor configuration is no longer written to the umh-core log at INFO on every internal re-apply (moved to DEBUG), removing a source of log flooding and CPU pressure when a config keeps being re-applied. While a bridge's observed config differs from its desired config, its status in the Management Console now shows a bounded summary of what differs, and the umh-core log carries one warning per bridge per minute, so the cause is visible without enabling debug logging

New Features

• Node-RED JS and tag processor expose a protobuf namespace (protobuf.decode / protobuf.encode) to decode and encode protobuf messages inline using an embedded base64 descriptor set, including proto2 extension fields
• Sparkplug B input decodes proto2 extension fields from an inline schema, exposing them per metric as spb_ext_* and spb_metric_decoded metadata
• TimescaleDB Historian output that saves a UNS data contract into TimescaleDB under a dedicated umh schema. By default every metadata key is stored; metadata_keys_exclude drops selected keys by exact name or prefix_* while keeping the rest.

Improvements

• The Authentication and Authorization documentation now covers enterprise Auth0 setup: how a default connection routes all logins through your own identity provider, how to choose the Account Owner as a break-glass account, and how inviting UMH personnel via their @umh.app address works, including who is responsible for their access

Fixes

• Stopping umh-core (docker stop, Kubernetes pod termination, CTRL-C) now shuts down workers gracefully on the first SIGTERM. Previously the signal killed the internal control loop before the graceful drain ran, so the drain waited out every timeout with nothing left to process — 33 s measured against Docker's 10 s grace period, ending in a hard kill mid-shutdown with workers never running their stopping steps. The drain now keeps the control loop alive, sizes its budget to the depth of the worker tree, and reliably stops workers that were mid-action, blocked behind an unhealthy dependency, or racing the shutdown signal — so it finishes well within the grace period instead of timing out. A second SIGTERM still forces an immediate exit
• Sparkplug B input can nest device data under its edge node via include_edge_node_in_location, so identically-named devices on different edge nodes no longer collide at the top of the hierarchy

Improvements

• For users testing the Write flows preview, a bridges-table cell with no throughput to show now explains why, instead of a stale 0.00: an "add flow" prompt when the side is not configured, or a lock when the protocol cannot have that side (a read-only or write-only protocol). The table and bridge overview now agree on each side's state.

Fixes

• For users testing the Write flows preview, the bridges-table throughput cell no longer briefly shows 0.00 while a flow is stopping; it waits for the flow to actually stop, so the summed cell matches the per-side cells during the transition.
• When a bridge update fails (for example, a config that references a template variable that doesn't exist), the "Continue Editing" button now returns you to the editor to fix the config. Previously it did nothing — the button linked to the edit page you were already on, so clicking it left you stuck on the failed-deploy dialog.
• After deploying a newly created bridge, "Continue Editing" again lands on the tab you deployed from (Read Flow, Write Flow, or General) instead of always the General tab.

Improvements

• Modbus addresses now require a Slave ID. The address editor's Slave ID dropdown no longer offers "All", so you have to pick a single slave from the configured slaveIDs. Rows missing a Slave ID show a red warning icon in the address table and Save & Deploy stays blocked until every row has one. Previously "All" silently subscribed to every configured slave, which duplicated tags when more than one slave was configured.

• When using the write flow preview, you can now switch between read and write log types. This is the same behavior, which already exists for the metrics.

Fixes

• Cards across the app now share a consistent look and layout. Previously, different pages used card components directly, producing variations in header spacing, title size, description placement, and border treatment. All cards now render through a single unified component, so padding, shadow, and typographic hierarchy are uniform regardless of where a card appears.
• Clicking into and enabling a code editor now places the cursor exactly where you clicked on the first click. Previously the first click was swallowed and you had to click a second time in the same spot to position the cursor. Scrolling the page while hovering an unfocused editor still works as before.

New Features

• Timeseries Data models can now reference timeseries-boolean as a payload shape, alongside the built-in timeseries-number and timeseries-string. Previously only number and string shapes existed, so a model referencing timeseries-boolean was rejected at deploy time and boolean process values like machine-running, alarm-active, or valve-open had to be faked as 0/1 numbers, losing type fidelity.

Improvements

• On a screen too small for the console, a message now points you to a desktop browser on every page, including sign-in and registration, instead of a broken layout

• In a bridge the code mode, tag processor and the address table now underline any data contract that doesn't exist on the instance, and Save & Deploy stays blocked until you create it or fix the name. Previously a mistyped or never-created data contract deployed silently and was only flagged after deployment.

• In a standalone data flow, the code editor now underlines any data contract that doesn't exist on the instance, and Save & Deploy stays blocked until you create it or fix the name. Previously a mistyped or never-created data contract deployed silently and was only flagged after deployment.

Improvements

• Enterprise companies that sign in through their own SSO connection can now invite UMH value engineers to help manage their instances. Previously, an invitation sent to a @umh.app address was tied to the company's own connection, where the engineer has no account, so the invitation could not be accepted. Now, invitations to @umh.app addresses use UMH's Google Workspace login, which is enabled on your organization automatically, so your UMH value engineers can accept the invite.

Fixes

• Classic instances showed Docker stop, pull, and run commands in the update instructions dialog — instructions that don't apply because Classic runs on Kubernetes, not as a standalone Docker container. The dialog now explains the correct upgrade path and links directly to the instance details page. UMH Core update instructions (Docker Run and Docker Compose) are unchanged.

• Naming a standalone flow the same as one that already exists on the instance is now caught as you type. The name field is highlighted and Save & Deploy is blocked. Previously the deploy appeared to start and then failed on the instance, because two flows with the same name resolve to the same identity.

• After deploying a standalone flow, the "Continue Editing" button now opens the flow's editing view as intended. Previously it only closed the dialog and left you on the create page.
• For users testing the Write flows preview, add selection between time-series and relational data processing for read and write flows.

New Features

• For users testing the Write flows preview, the bridges table now shows read and write throughput separately in one cell, each clickable to open its own flow. Previously a single combined number always linked to the read flow, so write throughput was neither visible nor reachable from the table

Improvements

• The instance name on the Agent card is now read-only text alongside the other details; a Rename button on the card opens a dialog to change it, so the card no longer carries an inline input and a separate Save

• Memory and disk usage on the instance overview now lead with the amount in use, shown as 10.9 GiB / 16.0 GiB (68.3%) instead of 68.3% of 16.0 GiB, so the actual consumption is the first value you read

• The Agent card now shows the instance location as plain read-only text instead of greyed-out fields, so a multi-level location reads clearly; the note that location is fixed after setup moved to an info icon on the label

• The Agent card no longer shows a 0.00 ms latency when no measurement has been reported; the latency row is hidden until a real reading is available
• Status cards now show the reason a component is degraded in a tooltip on its status badge instead of a banner inside the card, so cards no longer resize as health changes

Fixes

• After deploying a standalone flow, the "Continue Editing" button now opens the flow's editing view as intended. Previously it only closed the dialog and left you on the create page
• Safari now shows a clear page recommending Chrome or Edge, with download links, instead of a blank screen

Improvements

• Bridge deploy, edit, and delete progress messages and the bridge status feed now say "Bridge" instead of "Protocol converter", matching the Management Console terminology

Improvements

• The TimescaleDB Historian bridge template now creates all of its tables, types, and functions in a dedicated umh schema (for example umh.value_<contract> instead of value_<contract>), keeping the historian's objects separate from anything else in your Postgres database. New deployments use the umh schema automatically, and the database setup no longer needs a grant on the public schema. If you already deployed an earlier version that used the public schema, the bridge editor now includes a short one-time migration guide (a few ALTER ... SET SCHEMA umh statements) to move your existing tables into umh before redeploying, so your historical data is preserved.
• Deploying a bridge now saves your full configuration in one step. If the deployment does not reach a running state in time, your configuration is kept instead of being discarded and you are taken to the bridge's editing view, where you can adjust it and deploy again. Previously a failed first deployment could leave you without a saved bridge.

Fixes

• During sign-up, choosing a company name that was already taken showed "You are already onboarded" and redirected you to the home page, even though your account was never created. The onboarding form now shows "Please use a different company name" and keeps you on the page so you can pick another name and finish signing up.

Improvements

• Added Snowflake PUT bridge template with batching tuned for cost-efficient Snowpipe ingestion (6h period, ~150 MB compressed files).

Fixes

• After deploying a newly created bridge, the "Continue Editing" button always reopened the editor on the General tab. It now returns you to the tab you deployed from — General, Read Flow, or Write Flow — so you stay where you were working.
• TimescaleDB Historian bridges deployed with the historian data contract now store tag metadata; a SQL error in the template's database setup previously stopped the metadata table from being created, so metadata was lost
• Copying a write-only bridge now works correctly. Previously, copying a bridge failed with a "No Protocol Configuration" error because the action only checked for an existing read flow configuration. Now also write-flow configurations are recognized and copied correctly.

Improvements

• OPC UA: encrypted connections (Basic256Sha256, both Sign and Sign & Encrypt) now complete reliably, including signing-only mode which previously failed to connect
• When a bridge, flow, or stream processor config fails to render because the result is invalid YAML, the error now includes the failing lines of the rendered output. Previously the error named only a line number (for example yaml: line 25: did not find expected key) while the rendered output was discarded, so the failure could not be diagnosed from logs
• Editing a bridge whose new configuration keeps failing to render now fails after three identical render failures (a few seconds) and rolls back to the previous configuration automatically, naming the render failure as the root cause; while the edit is still retrying, each progress update also names the render failure. Previously such an edit retried for the full 30-second timeout before rolling back, and progress updates gave no reason

Fixes

• OPC UA: reading or subscribing to large tag sets (roughly a thousand or more) over an encrypted connection no longer drops the connection with an EOF error
• OPC UA: subscriptions now survive a reconnect, so data collection resumes automatically after a network interruption instead of stalling until the input is restarted
• OPC UA: connecting to servers that present their certificate as a chain (leaf plus intermediates), such as Siemens WinCC Unified, now works
• OPC UA: a single unreadable or invalid node ID no longer blocks monitoring the rest of a subscribed batch
• OPC UA: browsing certain servers no longer crashes the input

New Features

• TimescaleDB Historian bridge template in the Add Bridge wizard: Save data from the UNS into TimescaleDB for a single data model. Values and metadata are stored in TimescaleDB hypertables with compression enabled on deploy, keeping high-volume time-series data compact and fast to query. By default it archives all tag metadata except high-churn transport keys (OPC UA and Sparkplug timestamps, sequence numbers, Kafka record times, and similar) that change on nearly every message and would otherwise bloat the metadata table; you can instead set an explicit list of metadata keys to keep. It is maintained and supported by the UMH team, with the required database setup documented inline in the bridge editor.
• Stream processors now have an overview page, and the editor has been reworked. Opening a stream processor from the data flow list shows a read-only summary with live monitoring (status, throughput, plus Logs and Metrics) instead of dropping you straight into the editor. A new "Edit Config" button opens the editor. In the editor, the general settings and the UNS location are now split into separate cards, and the code editor fills the available height and scrolls internally instead of stretching the page.

• Standalone data flows now have an overview page and a refreshed editor. Opening a Standalone Flow from the data flow list shows a read-only summary with a live monitoring view. A new "Edit Config" button opens the editor when you want to make changes. The editor now matches the rest of the app with the same cards and spacing. Activating or pausing a flow is now a clear "Activate Flow" toggle in its own card, the same control bridges use, instead of an "Active/Stopped" dropdown.

Improvements

• Data flow type selection (Bridge, Stream Processor, Standalone) now lives only in the Data Flows sidebar sub-navigation. The duplicate tab bar above the data flow list has been removed, since it switched between the same views the sidebar already offers.

• CPU usage in the instance details row now shows usage against the container's cgroup limit when one is set (for example, 1850 / 2000 mCPU), instead of as a percentage of all host cores.

• Hovering the CPU row opens a metrics tooltip with usage, limit, and throttle breakdown.

Fixes

• Sparkplug B bridges now show every metric in the Topic Browser, including ones that never change after startup. A Sparkplug device announces each metric's initial value once and then only reports later changes, but the default template ignored those initial announcements, so any metric that stayed constant never appeared; initial values now come through and seed each topic, with later changes updating it in place

Improvements

• After a successful deployment, the dialog previously pushed you out of the editor: the primary "Continue" button returned you to the data flow overview, and when updating an existing flow there was no option to keep editing. The primary action is now "Continue Editing", which keeps you in the editor. A "Back to Overview" link still takes you back to the data flow list, or to the bridge's overview page when write flows are enabled.

Fixes

• When deploying or editing a bridge or standalone data flow failed because of an invalid or rejected configuration, the dialog still offered a "Save Anyway" button. For these errors that button could not help, and for an invalid config it would re-apply the broken configuration and report success — leaving the component crash-looping with no rollback. "Save Anyway" now appears only for errors where ignoring them is safe (such as a timeout); for configuration errors the dialog asks you to fix the configuration and try again.

Improvements

• Consistent spacing between tabs and the content below them in the bridge editor, matching the rest of the page.

Fixes

• Interface elements that are links or images can no longer be dragged loose from their position. The browser let you pick up the back button on detail pages, sidebar tabs, breadcrumbs, logos, link-styled buttons such as "Add Bridge", table row-action menu entries, your profile picture, and decorative images and drag them around the page; everything now stays put while clicking still works as before.

Fixes

• Sparkplug B output: a standard tag_processor → sparkplug_b flow now publishes its tags. The output read the value from a payload field named after the tag, but tag_processor emits it under value, so every message was silently dropped — leaving only an empty NBIRTH on the broker. The output now extracts the value the same way the Sparkplug B input does (value/val/data/measurement), makes virtual_path optional, and warns instead of dropping silently when a payload can't be turned into a metric
• OPC-UA input now preserves string values exactly; previously, numeric-looking strings like serial codes were emitted as numbers and lost precision. For tags processed by tag_processor, also set msg.meta.datatype = "string" so auto-detection does not convert them back
• Sparkplug B input: metric datatypes now survive from BIRTH to DATA. Per spec, NDATA/DDATA carries only alias + value while the BIRTH certificate defines name and datatype — but the input cached only the name, so DATA messages lost their spb_datatype metadata and signed integers decoded as their unsigned two's-complement wire value (an Int32 of -12 surfaced as 4294967284). The alias cache now restores the datatype alongside the name, and Int8/Int16/Int32/Int64 wire values are reinterpreted as signed

Improvements

• After deploying a standalone data flow or a Classic protocol converter, the success dialog's button was labeled "Data Flows" while the bridge deploy flow labeled it "Continue". All three now use "Continue".

• Clicking "Continue" after deploying a bridge dropped you onto that bridge's edit page, while deploying a standalone data flow returned you to the data flow overview. Both now return to the overview.

Fixes

• Condition cards and their clause rows in the Tag Processor can no longer be dragged loose from anywhere on the card. They now move only when you grab the drag handle, the way reordering was always meant to work.
• Previously, when configuring a bridge's processing before choosing an instance, the address table section showed an "Upgrade Required" message — even though the real problem was that no instance had been selected yet. The notification now shows a clear "No Instance Selected" prompt pointing you to the General tab. In addition the notification has been moved out of the processing section address tab, to the top of the read flow to be better and quicker recognizable.

Fixes

• Previously, a standalone data flow that writes to more than one destination (switch, broker, or fallback) showed zero or incomplete throughput in the Management Console even though it was processing data normally. The throughput, error, and connection counts now include every destination

Preview: Write Flows

• The write flow schema has been updated to use structured fields. The previous flat fields input_topics and output are replaced by source.topics, destination.protocol, destination.code, processing, and extra. Write flows configured before this release must be re-created from the UI; read flows and connection settings are not affected.

New Features

• Modbus bridge addresses can now be edited through a guided dialog. Previously, each address was a single string such as temperature.holding.100.INT16:scale=0.1 that you had to assemble and edit by hand in the read flow table. Clicking an address cell, which holds a 'unifiedAddress' now opens a dialog with a separate field for each part (name, register type, address, data type, and optional modifiers like scale and slave ID), and only shows the fields that apply to your selection (for example, length appears only for string types). Hovering an address cell shows its parts at a glance.

Improvements

• After migrating your account to Auth0, entering your email on the login page now sends you straight to the sign-in method you migrated with. Previously, you landed on Auth0's hosted login UI and had to pick your provider (Google, LinkedIn, SAML, ...) before being prompted for credentials. This applies to both community and enterprise accounts; enterprise users are additionally routed directly into their company's Auth0 organization.
• The sidebar tab that opens data models is now labeled "Data Models", matching the heading on the page it opens. Previously the tab read "Models" while the page itself was titled "Data Models".

Fixes

• Previously, opening a bridge and leaving the page warned about unsaved changes even when nothing had been edited, while real edits to the Read Flow's Input section went undetected and could be lost on navigation. The bridge page now captures a clean baseline once it finishes loading and compares the Input configuration directly, so the unsaved-changes prompt appears only when you have actually made a change.

Improvements

• Sparkplug B input: request_birth_on_connect now defaults to true, so secondary_active/primary bridges proactively rebirth newly seen nodes on connect. Set it to false to keep the prior behavior (ignored under secondary_passive)
• Sparkplug B input: field descriptions and the primary-role startup log now clarify that identity.edge_node_id is the Sparkplug v3.0-compatible host_id in the STATE topic (spBv1.0/STATE/<host_id>)
• New snowflake_put output: ports the warpstreamlabs/bento Snowflake output into benthos-umh for writing batched messages to Snowflake stages with optional Snowpipe ingestion. Supports user/password and key-pair auth, all gosnowflake compression modes, and per-message stage/Snowpipe interpolation

Fixes

• When an FSMv2 transport worker (auth, push, or pull) was stuck in a transient-retry state, the supervisor heartbeat printed the same stale reason set on entry — for example "Desired state is running, transitioning to Starting" — instead of the live per-tick reason the worker already composes, such as "auth backoff: 42 errors (cloudflare_challenge), delay 60s". The heartbeat now surfaces the live reason within one ~10s cycle.
• Sparkplug B input: identity.group_id now filters the MQTT subscription by default. Previously an empty subscription.groups subscribed to every Sparkplug group on the broker (spBv1.0/+/#) regardless of identity.group_id. To restore the old behavior, set subscription.groups: ["+"] explicitly
• Sparkplug B input: bridges now request a rebirth when DATA references aliases the cache hasn't seen (typically after a bridge restart with no retained NBIRTH/DBIRTH on the broker). Previously tags surfaced as …/_historian/alias_<n> until something external triggered recovery

Preview: Write Flows

• Write flows (preview) now take their input topics directly on the flow, with Go template support for values like {{ .location_path }}. Write flows configured on this preview before this release are not carried over and need to be re-created from the UI; read flows and connection settings are unaffected

Fixes

• Tag names and virtual paths entered in the address table are now preserved when reopening a bridge. Previously, reopening a configured bridge showed an empty table, and re-deploying would overwrite the saved values with empty ones.

Fixes

• S7 bridges now connect using the configured port. Previously, the port entered during setup was ignored and the connection always used the default port 102
• Location and virtual path columns in the address table no longer show a required field indicator

Improvements

• Links across the app now share a single visual style. Previously, links used different colors, underline treatments, and hover effects depending on where they appeared, producing an inconsistent look across pages. They now follow a unified style, and external links — those pointing outside the app, such as to documentation — automatically open in a new tab and display a small indicator icon next to the text.

• We now display a reload dialog when the app needs to be refreshed.

Fixes

• Previously, the log viewer re-fetched the full 24-hour window on every refresh and kept all returned lines in memory. On instances with busy logs, this grew continuously and could crash the browser tab. The log viewer now fetches only new entries since the last refresh and caps the in-memory log at 10,000 lines.
• Previously, picking certain protocol templates when adding a bridge (for example http_server) produced a default name like Generic-http_server-bridge-1 that could not be saved, with an error about invalid characters. The suggested name now uses only valid characters (e.g. Generic-http-server-bridge-1), so you can deploy without renaming it first.
• Previously, the Data Contracts page showed a dead-end empty state on fresh instances, with no indication that data contracts are created automatically when a data model is deployed. The empty state now explains this and links directly to the data model creation page. The stream processor editor's contract selector also shows a link to add a data model when no contracts are available.

Fixes

• Editing or deploying a bridge with an invalid config could hang the Management Console in "executing" indefinitely — umh-core's rollback path crashed mid-execution, the instance restarted, and the UI never received a final reply. The bridge often ended up needing delete + recreate to recover. The edit/deploy now ends with a clear failure when the rollback succeeds. Affects v0.44.19
• An unexpected error in umh-core (e.g. editing or deploying a bridge) used to restart the umh-core instance and leave the operation stuck in "executing". Affected operations now end with a clear failure message (Internal error: <type> action failed unexpectedly. UMH engineering has been notified.) while the rest of the instance keeps running

New Features

• Community accounts can now migrate from email/password login to Auth0. A banner on the home page opens a short wizard that links your account to an Auth0 identity. After linking, you sign in to UMH through Auth0, which adds Google and LinkedIn sign-in options and supports MFA. You still use your existing password to unlock your data after sign-in. Email/password login keeps working for accounts that haven't migrated yet, so you can take the migration at your own pace.

• The login page now starts with a single email field and routes you to either the password screen or to Auth0 based on which login method your account uses. If you previously linked an Auth0 identity but kept signing in with email/password, you are now routed through Auth0.

Improvements

• The bridge setup wizard, dynamic protocol forms, and enhanced read flow UI were previously shipped as previews enabled by default, and are now generally available.

Fixes

• Previously, bridges created with Copy Template showed "Not configured" on the Read Flow card even while data was flowing through. This showed a misleading Configure Read Flow button on bridges that were already set up. Both new copies and existing affected bridges now display their Read Flow correctly.
• Previously, using the breadcrumb navigation for some bridges and standalones would result in showing a missing overview page. This is now fixed.

Preview: Enterprise login migration

• Enterprise organizations can opt in to a migration that provisions an Auth0 organization for the company, lets the owner pick a default sign-in connection (Google, LinkedIn, or a custom Auth0 connection ID), and walks each member through a short migration once the owner has finished. To opt in, contact your UMH point of contact to prepare the migration (i.e. setup the custom SAML connection in Auth0 and designate a company owner who is then able to run through the company migration wizard).

Improvements

• Previously, the component tasked with communicating with the UI could stop working if it rebuilt its HTTP connection (which happens after persistent network failures) while a message was being sent or received. Connection rebuilds and in-flight requests are now coordinated so they cannot interfere.
• Config backup is now enabled by default. Previously, automatic config.yaml backups required setting ENABLE_CONFIG_BACKUP=true. After validating the feature in customer environments for over two months, every umh-core instance now writes timestamped copies of config.yaml to /data/config-backups/ on every config write, retaining the 100 most recent versions (roughly 5–20 MB). To opt out, set ENABLE_CONFIG_BACKUP=false. To roll back a bad config, use the second-most-recent backup — the most recent one mirrors the bad write
• Previously, docker stop and Kubernetes pod terminations killed umh-core immediately, dropping in-flight messages and connections without a clean shutdown. umh-core now handles SIGTERM and lets its background services (e.g., the communicator) finish their in-flight work before exiting, typically well within Docker's 10 s and Kubernetes' 30 s default grace periods.

Fixes

• A data flow (bridge, standalone, or stream processor) could get stuck in a starting state when its on-disk service directory was left in an inconsistent state — for example after a container restart or OOM-kill interrupted setup. umh-core now detects the inconsistency and rebuilds the directory automatically. Previously, recovery required restarting umh-core or recreating the data flow under a different name.
• OPC-UA input could get stuck while browsing when a configured NodeID did not exist on the server, requiring a manual restart. Browse failures now trigger a clean reconnect
• OPC-UA input no longer spams Variant is nil errors when a node sends a status update without a value. These are harmless and now logged at debug level with the NodeID and status code
• Modbus TCP input now reconnects immediately on any transport-level error (timeouts, resets, network failures), not just broken pipes. Previously these stuck the connection for up to 10 seconds
• Modbus TCP input now recovers automatically from transaction-ID mismatches. Previously, when a slow PLC reply arrived after its read timeout, the next poll picked up the stale frame and failed with modbus: response transaction id 'X' does not match request 'Y'. The connection thrashed (reconnect, mismatch, reconnect) and reads stalled until conditions cleared or the input was restarted

Improvements

• The delete confirmation for data flows and data models now includes the instance name in the description. Previously, if the same flow or data model name existed on multiple instances, the dialog gave no way to tell which instance was being targeted.

Fixes

• Previously, using "Copy Bridge" on an existing bridge and clicking "Save & Deploy" failed with an "Action Failure" error. Cloning and deploying bridges now works correctly

Preview: Disable Read Flow

• Improved metrics view on overview page for different states of read and write flows.

Delete dialogs, sidebar data flows grouping, and several polish fixes: alert text wrapping, version tooltip/dialog cleanup, button cursors, and bridge condition meta keys.

Improvements

• Delete dialogs for instances, data flows, and data models now use the same dialog shell as the rest of the app. Escape and outside clicks are blocked while a deletion is running, the title updates to reflect progress, and a "Continue" button appears when the operation completes.

• The sidebar now groups Bridges, Stream Processors, and Standalone Flows under a single collapsible "Data Flows" section. Jumping between data flow types takes one click from anywhere in the app, and the section remembers whether you left it open or collapsed.

Fixes

• Previously, error messages, identifiers, and URLs without spaces could overflow the alert box and appear cut off. Alert text now wraps at any character boundary, keeping the full message visible.

• Cleaned up the version update tooltip in the instances table: removed the low-contrast divider and washed-out text on the dark tooltip background, evened out the spacing, and clarified the call-to-action so it's clear the badge itself is what you click to open the update instructions.
• Update instructions dialog now uses the standard dialog look and feel (matching other dialogs in the app), with a "View changelog" action in the footer instead of a buried link in the description.

• Disabled buttons now show a not-allowed cursor instead of the default arrow cursor, and loading buttons show a busy cursor. Previously, pointer-events-none on disabled buttons suppressed all pointer events, so the cursor never changed.

• Conditions in the bridge processing step now match incoming messages. Previously, the default field selector for OPC UA and S7 bridges generated YAML referencing meta keys (msg.meta.nodeId, msg.meta.address) that the inputs do not emit, so conditions silently never fired. The selector now defaults to the actual runtime keys (msg.meta.opcua_attr_nodeid, msg.meta.s7_address) while keeping the human-readable labels "Node ID" and "Address" in the dropdown.

Improvements

• Tag processor now supports msg.meta.datatype to override value type auto-detection. Set to "string", "number", or "bool" to force the output type

Fixes

• Dialogs no longer disappear when clicking outside of them
• Code block text selection in sidebar no longer appears dark in light mode

Enables FSMv2 communicator by default.

Improvements

• The FSMv2 communicator is now enabled by default. Previously, enterprise networks with firewalls that dropped idle connections could leave instances appearing offline in the Management Console until the container was restarted. The new communicator uses separate connections for status and commands, so a dropped pull connection no longer prevents heartbeats from reaching the backend, and it recovers automatically when the connection comes back. The memory cleanup routine it depends on is also enabled by default. If you previously disabled FSMv2 by setting useFSMv2Transport: false in config.yaml, that setting has had no effect since v0.44.8 -- to keep the legacy communicator you must set USE_FSMV2_TRANSPORT=false as an environment variable. To roll back, set USE_FSMV2_TRANSPORT=false and USE_FSMV2_MEMORY_CLEANUP=false.

Preview: Write Flows

• Previously, write flows used raw input for benthos. We now define a user variable UMH_TOPICS and automatically generate the proper input in umh-core, which provides an improved user experience and less surface for errors.

Fixes

• When creating a Bridge with the newest version of umh-core you previously got a warning for not existing schemas. These are now added.
• Depending on the screen resolution, the toggle-switch appeared slightly misaligned. This issue has been resolved, so the button now appears consistent regardless of the screen resolution.
• Previously, standalone flow names accepted spaces and other characters that cause deployment failures. Names are now validated the same as bridges and stream processors: letters, numbers, dashes, and underscores only

Preview: Disable Read Flow

• Now uses UMH_TOPICS list as input for write flows. Breaks all existing write flows.

Improvements

• Cache API for JavaScript processors: new cache.set(key, value), cache.get(key), cache.exists(key), and cache.delete(key) methods for tracking state across messages. Previously, state management required complex Benthos branch/request_map/result_map configurations. Now you can store any JSON-compatible value (strings, numbers, objects, arrays) directly from JavaScript. Use cache.exists(key) before cache.get(key) to handle missing keys. Available in both nodered_js and tag_processor. Currently in-memory only (lost on restart), persistent backend planned
• Updated Go dependencies, includes security fixes for OIDC and JOSE authentication libraries

Preview: Disable Read Flows

• Supports extracting metrics for write flows for bridges.

Improvements

• Cards in the read flow editor now group related settings behind tabs, so you only see what you need for the step you're on. Long sections that used to force a lot of scrolling now fit in one view, and switching between settings takes one click instead of hunting through the page.

• Each category now has a single Code/Visual toggle instead of one per section. Switch the whole card at once when you want to drop into raw config, and flip it back without having to redo the choice on every tab. Your view stays consistent while you work through a card. - Metrics will display separate metrics for read and write flows in bridges, once umh-core supports it. - We replaced the editor engine behind the tag processor rules, bridge configuration, data models, stream processors, and other configuration screens. Editors now open faster, handle typing and scrolling more smoothly on large configurations, and offer easier and faster implementation for future features.

Fixes

• When confirm dialog is loading, pressing Escape or clicking outside no longer cancels the running action. Dialog stays locked until the operation is finished.
• Previously, navigating to “Add standalone flow” showed a Monitoring card carrying over status and throughput data from a previously viewed component. The Monitoring card now only appears when editing an existing component

Security updates and more templates.

New Features

• Add more templates and ability to show a verbose vendor name.

Adds relational field type, per-direction flow metrics, config backup with post-write fix, and retry-resilient status delivery for FSMv2.

New Features

• Data models now support a new _relational field type for defining flat relational data alongside timeseries fields. Each inline field contains named columns with a type (string, number). This is mutually exclusive with _payloadshape and _refModel on the same field

Improvements

• Previously, a Flow only exposed a single health and throughput metric. Starting with this version, umh-core exposes separate throughput metrics and health status for read and write flows.

Preview: Config Backup

• ENABLE_CONFIG_BACKUP=true saves a timestamped copy of config.yaml to /data/config-backups/ every time umh-core writes the config (Management Console actions, and startup after a manual edit). umh-core keeps the 100 most recent backups and deletes older ones. Use these to roll back a bad config
• Fixed: the first config change after a restart was not backed up, because the pre-write file matched the latest backup and the duplicate check skipped it. Backups now run after the write. One caveat: the most recent backup mirrors the most recent write, so to recover the state from before a bad write, use the second-most-recent backup

Preview: FSMv2 Communicator

• Previously, healthy instances could appear offline in Management Console when earlier status messages failed to send. New status messages now continue to reach Management Console while previous ones are being retried. Only affects instances with USE_FSMV2_TRANSPORT=true

Reworked install dialog with Docker Run/Compose/Manual tabs and download support; fixed orphaned identity records on user deletion, OPC UA browser crash loop, and sidebar icon hover contrast.

New Features

• Reworked install dialog. The install dialog now shows three tabs: Docker Run, Docker Compose, and Manual Setup. They cover different deployment environments. Closing without copying or downloading the command now triggers a confirmation prompt instead of navigating away silently. A prominent warning highlights that the auth token can not be retrieved after closing. Download buttons let you save the command as a .sh or .yml file. After completing setup, the instances view shows a brief notice that the instance will appear offline until the install command runs.  

Fixes

• Previously, deleting a user left orphaned identity records in the database, so re-inviting the same email address could fail during signup. User deletion now removes all associated identity records in the same operation
• Previously, opening the OPC UA browser updated the form in the background (syncing checked node IDs), which the UI misread as a protocol change and reinitialized the browser — creating a crash loop. The OPC UA browser now opens without crashing.
• Previously, the rendered icon did not inherit the hover text color, so it always had the same color regardless of the hover state. Fixed sidebar user dropdown icon contrast. The Settings, User Management, and Logout icons stayed a fixed color on hover. They now inherit the highlighted text color, matching the rest of the menu item

Fix issue when redeploying a bridge.

Fixes

• Previously, if a read flow had a stale error, a new deployment would always fail, even if the current state is healthy. Now, we first always stop the flow, redeploy it and only check the current state, not any stale states.

Schema, theme, and editor fixes.

Fixes

• Map fields in the bridge editor JSON schema incorrectly produced "type": "string" instead of "type": "object" with additionalProperties. Component reference types (input, output, processor, scanner) and unknown types had the same issue -- all now map to the correct schema types
• Previously, if two bridges shared the same internal ID, the bridge selection list would fail to render. Duplicate entries are now filtered out automatically
• Make the background of the user management input fields respond to the selected theme

• Fixed the data model code editor being capped at a small fixed height, leaving most of the screen empty -- it now expands to fill the available window height, consistent with other editor instances

Improvements

• Unified Address Field for Modbus: introduces unifiedAddresses as a single-string alternative to the existing address object list. Format: name.register.address.type[:key=value]* (e.g., temperature.holding.100.INT16:scale=0.1). The legacy addresses object list continues to work with a deprecation warning. Both fields are mutually exclusive

Fixes

• CPU health stayed Degraded permanently after any brief throttle event because the ratio used cumulative counters since pod start. Now uses a sliding window, so health recovers when throttling stops
• ADS symbol downloads failed in certain configurations -- bumped ADS plugin to v1.0.8 which fixes the issue

Preview: FSMv2 Communicator

• Brief network interruptions during authentication (DNS failures, server errors) no longer produce Sentry warnings. If authentication keeps failing for five consecutive failures, a single persistent_auth_failure warning is logged. Permanent errors like invalid credentials or a deleted instance still warn immediately on first occurrence. Only affects instances with USE_FSMV2_TRANSPORT=true
• When authentication permanently fails (invalid credentials or deleted instance), the transport worker now enters a dedicated failed state instead of retrying indefinitely. It re-attempts authentication automatically when the configuration changes (new auth token, relay URL, or instance UUID). Only affects instances with USE_FSMV2_TRANSPORT=true

Fixes for location level naming and condition editing.

Fixes

• Previously, typing in a condition value field lost focus after every character, requiring you to click back into the field after each keystroke. Condition editing now keeps focus as expected
• Previously, location level inputs in the instance setup form would automatically reformat text as you typed -- lowercasing letters, replacing spaces with hyphens, and removing special characters. Location inputs now accept text exactly as entered

Fixes

• Updated gRPC to v1.79.3 to patch CVE-2026-33186, a critical authorization bypass where path-based access control policies could be circumvented via malformed HTTP/2 requests

Topic browser, table filters, and bridge settings now persist across navigation and page reloads. Fixed OPC UA bridge defaulting to wrong port, "Save Anyway" not bypassing validation errors, and special characters breaking bridge deployments. Preview: new "Disable Read Flows" toggle for bridges.

Improvements

• Previously, the topic browser reset to its default state whenever you navigated away -- your selected topic, expanded/collapsed nodes, search term, and scroll position were all lost. Now the topic browser remembers all of these across navigation and page reloads, and resets on logout
• Previously, new instances defaulted to the Enterprise release channel, which lagged behind in versions. Now new instances default to Stable, so they start on the latest recommended version
• Previously, the filter of a table was reset whenever you navigated away. Now the state of all table filters is persisted throughout the entire UI.
• Table filters now support fuzzy finding.

Fixes

• Previously, creating an OPC UA bridge pre-filled port 3000 instead of the OPC UA standard port 4840, requiring manual correction every time. The bridge setup assistant now defaults to port 4840
• Fixed "Save Anyway" button not bypassing validation errors during bridge updates -- clicking it re-triggered the same deployment without ignoring errors, causing the same failure again. It now correctly saves despite benthos linting or configuration errors

• Previously, special characters like double quotes in address mappings or metadata fields broke bridge deployments because they were not escaped in the generated YAML. Now values are properly escaped before sending to the backend and unescaped when loaded back into the UI

Preview: Disable Read Flows

• Introduced the new, experimental feature flag "Disable Read Flows" -- When activated you get a new toggle in the Read Flow section of the bridged. Using the toggle, you can temporarily disable reading data. This feature is already present for Standalones and is now being ported to the read flows.

Improvements

• Starting with this version, umh-core supports stopping and starting bridges. Previously, bridges were always active and could only be removed entirely. A Management Console update is required to expose this in the UI

Fixes

• The container previously restarted hundreds of times per minute when config.yaml was missing or invalid and now waits 60 seconds before retrying, giving you time to fix the configuration
• Previously, memory monitoring read host-level values even inside containers, while CPU monitoring already used cgroup-aware values. Memory metrics now read cgroup v2 limits and usage, so dashboards show correct container memory utilization
• Fixed debug logging settings being lost when editing bridges or data flows -- the debug_level flag is now preserved across configuration changes

Preview: FSMv2 Communicator

• Instances could appear permanently offline in the Management Console even though the pod was running and healthy -- only a restart would fix it. This happened because token re-authentication briefly caused child workers to enter a stopping state from which they could not recover, leaving the communicator stuck indefinitely. Workers now always complete the stop and automatically recover when the parent is healthy again. Requires USE_FSMV2_TRANSPORT=true

We revised how locations work in the Management Console. The instance setup form now matches what UMH Core already supports -- more than five location levels, and location is locked after the initial setup to prevent accidental redeployment of all your data flows.

Improvements

• Previously, the instance setup form was limited to exactly 5 fixed location fields (enterprise, site, area, line, work cell), even though UMH Core already supports deeper hierarchies. Core instances running v0.44.9 or later now support up to 10 levels in the Management Console, so you can define location hierarchies that match your organization's structure

Fixes

• Fixed instance location being editable after setup -- changing it would silently redeploy all data flows and rename all tag names, which is almost never intended. Location is now set once during setup and clearly marked as permanent. If you need a different location, create a new instance. Changing the location of a bridge is still possible
• Fixed the configuration editor and log viewer being capped at a small fixed height, leaving most of the screen empty -- both now expand to fill the available window height

• Fixed feature flag toggles in Settings taking effect silently with no confirmation -- toggling now shows a brief notification

Improvements

• Previously, internal timing metrics grew continuously in memory the longer your instance ran, consuming over 500 MB on busy systems with many bridges and data flows. These metrics now use fixed-size histogram buckets that no longer grow with uptime, reducing steady-state container memory by roughly 30%

If you run instances with a large number of topics, the Management Console could feel sluggish -- slow clicks, delayed page updates, and high latency shown in the status bar. This was most apparent with many instances, on older laptops, or alongside resource-heavy software like antivirus scanners. The browser now handles incoming data much more efficiently during updates. This does not eliminate all sources of latency, but removes the largest one.

Improvements

• Condition input fields now suggest values directly from your CSV address table -- clicking a suggestion auto-filters the table to matching rows

Fixes

• Fixed tag names falling back to a raw system identifier when no name was configured -- this could cause deployment errors

This release simplifies S7 addressing and fixes three edge cases in the Management Console editor and S7 data type handling.

Improvements

• S7 addresses for PE, PA, MK, C, and T areas no longer require a block number that served no function. You can now write PE.X0.0 instead of PE0.X0.0. The old format still works but logs a deprecation warning and will be removed in a future version. Data Block addresses (DB1.DW20) are unchanged

Fixes

• The S7 DateAndTime data type crashed due to an incorrect buffer size and now reads correctly
• Fields with children that already have default values were incorrectly marked as required when editing bridge configurations -- they are now correctly treated as optional
• Fields marked as deprecated in bridge plugin definitions were not flagged in the Management Console editor -- they now correctly appear as deprecated

Preview: FSMv2 Communicator

• Status updates and incoming commands now use independent connections, so a connection dropped by an enterprise proxy or firewall no longer prevents the instance from sending heartbeats. Requires USE_FSMV2_TRANSPORT=true

The interface no longer stutters periodically during your session -- a background security check was running on the main thread every 10 seconds and has been moved off it.

Fixes

• Fixed the error details panel showing no content when a validation error occurred -- it now displays the specific errors correctly
• Fixed a bug where the instance deletion dialog would get stuck after deleting an instance -- the confirmation input stayed visible and the dialog could not be closed

Fixes for bridge configuration saving and condition builder defaults that were introduced alongside the CSV Address Tables release.

Fixes

• Fixed bridge save discarding deleted addresses and custom meta fields -- removing an address row or a custom metadata column in the address table now persists correctly on save
• Fixed the second conditional clause defaulting to "equals" when its value was empty -- new clauses now start without a pre-selected operator, matching the first clause behavior
• "Unit" has been removed from the default CSV table columns. A warning indicating column duplicates was added

Configuring OPC-UA and S7 bridges no longer requires writing YAML by hand. The Read Flow processing section now uses a visual address table where you can add, edit, and search addresses directly, or import them in bulk from a CSV file. You can also duplicate any existing bridge as a starting point for a new one. Previously available as a preview, this is now the default experience for all OPC-UA and S7 bridges on instances running UMH Core v0.44.6 or later.

New Features

CSV Address Tables - The Read Flow processing section for OPC-UA and S7 bridges is reorganized into three sections: Default, Addresses, and Conditional Formatting. The Addresses section replaces the raw YAML input with a spreadsheet-style table. Each row represents one node ID (OPC-UA) or address (S7), with columns for data contract, location path suffix, virtual path, tag name, and unit. You can add custom metadata columns for any additional fields your setup needs. Support for more protocols is planned for upcoming releases.

Import your address list from a CSV file to populate the table in bulk, or export the current configuration back to CSV. The table supports pagination (20, 50, or 100 rows per page) and live search across all columns, so working with hundreds of addresses stays manageable.

Each section has its own UI/Code toggle. Switch to code mode to see and edit the generated YAML directly, then switch back to continue in the visual editor. Changes sync both ways.

The Conditional Formatting section now uses structured dropdowns for building conditions: select a meta tag, pick an operator (equals, starts with, contains, and others), and enter a value. Chain multiple clauses with AND/OR, and drag to reorder conditions. Previously, conditions required typing the full JavaScript expression manually.

Create From Existing Bridge - You can now create a new bridge by copying an existing one. From the Data Flow list, click the three-dot menu on any bridge and select "Copy Bridge", or use the "From Existing Bridge" card in the creation wizard. The new bridge is pre-filled with the source bridge's connection settings, protocol configuration, and full Read Flow processing setup. The name is automatically suffixed with "-copy" to avoid conflicts. Edit any field before saving. Currently, each copy is independent. Shared bridge templates where changes propagate to all referencing bridges are already supported via config.yaml. A future release will bring this to the Management Console UI as well.

Improvements

• The Data Contract dropdown in the processing section now lists custom data contracts from your instance alongside the built-in _historian and _raw options
• Protocol configuration forms no longer show deprecated fields, reducing clutter in the input section

Fixes

• Fixed the Advanced Settings section disappearing when a protocol had no auto-generated advanced fields, which blocked access to custom configuration. Custom YAML fields now also show inline validation errors
• Fixed Read Flow processing silently dropping all messages when address mappings were configured -- re-saving your bridge after this update corrects it automatically

This release adds config backup as an opt-in preview and includes stability fixes for FSMv2 communicator preview users. All changes require feature flags -- if you haven't enabled any preview features, this release does not affect you.

Preview: Config Backup

• Previously, config.yaml could be permanently lost during upgrades or crash loops with no way to recover. Timestamped backups are now created before every write, with deduplication to prevent churn during restarts, retaining the last 100 versions in /data/config-backups/

To enable, set the ENABLE_CONFIG_BACKUP=true environment variable and restart your container:

docker run -e ENABLE_CONFIG_BACKUP=true ...

To verify, check that backup files appear in /data/config-backups/ after a config change. If no files appear, confirm the environment variable is set and the container was restarted.

Preview: FSMv2 Communicator

• The cleanup routine retained state history for 24 hours, allowing hundreds of thousands of entries to accumulate in RAM -- the retention window is now 1 hour, reducing steady-state memory by roughly 95%. Requires USE_FSMV2_MEMORY_CLEANUP=true

• The supervisor logged an error on every tick when the communicator temporarily had no workers during a restart cycle -- it now skips the tick and self-heals when workers return. Requires USE_FSMV2_TRANSPORT=true

• A nil-pointer panic could occur in ResetTransportAction under specific timing conditions -- the action now safely handles nil state. Requires USE_FSMV2_TRANSPORT=true

IO-Link bridges that fetch device descriptions from the IODD registry now connect correctly. Previously, the Management Console's IODD proxy did not trust the certificate authority used by ioddfinder.io-link.com, which prevented IO-Link bridges from loading device descriptions needed to decode sensor data.

Improvements

• Updated Go runtime to 1.25.6 and Node.js to 22.22.0 with security patches

Fixes

• Fixed the "Detected Changes" confirmation dialog appearing on Read Flow configuration pages when no changes had been made

Preview Improvements

These improvements apply to features currently available as a preview.

• Addresses now correctly load when opening an existing bridge, creating a new one, or switching between protocols (Enhanced Read Flow UI)
• Fixed deleting newly added custom columns from the address table not working (Enhanced Read Flow UI)

The right-click menu on Data Flow pages no longer shows redundant labels. Previously, the menu item repeated the flow type name (e.g., "Standalone Flow" or "Stream Processor") even though you were already viewing that tab. Now it simply says "Configure" with a cog icon, which is clearer and less cluttered.

Fixes

• Fixed the "Go back to safety" button on error pages not navigating anywhere when clicked. The button now performs a full page reload to take you back to the home page.
• Fixed buttons with tooltips (like the "Full screen" button in code editors) not responding to clicks
• Fixed dropdown menu items showing a default cursor instead of a pointer cursor on hover
• Fixed inconsistent font styling in the audit logs table

If you enabled the FSMv2 communicator preview (introduced in v0.44.5), this release fixes three issues that could cause your instance to use a lot of memory, go offline, or restart unexpectedly. We strongly recommend upgrading.

If you have not enabled the FSMv2 preview, this release does not affect you.

Action Required (FSMv2 Preview Users Only)

You need to add a new environment variable to get the memory fix.

Regardless of how you enabled FSMv2, set both of these environment variables and restart your container:

• USE_FSMV2_TRANSPORT=true — enables FSMv2 (unchanged from before)
• USE_FSMV2_MEMORY_CLEANUP=true — new in v0.44.8, enables the periodic memory cleanup

docker run -e USE_FSMV2_TRANSPORT=true -e USE_FSMV2_MEMORY_CLEANUP=true ...

If you enabled FSMv2 via useFSMv2Transport: true in config.yaml, that method no longer works — FSMv2 is silently not running. Remove that line from your config.yaml and use the environment variables above instead.

To verify FSMv2 is running after the change:

docker exec <container-name> grep "Using FSMv2 communicator" /data/logs/umh-core/current

If this returns no output, FSMv2 is not running. Double-check that both environment variables are set and that the container was restarted.

Fixes

• Memory no longer grows unbounded over time. Previously, long-running instances with FSMv2 enabled would slowly consume more and more memory because internal state changes accumulated without ever being cleaned up. One customer reported 13 GB of RAM usage at only 4 messages per second. Eventually, the system would kill the container to free memory, causing a brief data gap while it restarted. A periodic cleanup routine now keeps memory usage flat regardless of how long the instance runs. This fix requires USE_FSMV2_MEMORY_CLEANUP=true (see above).

• Instances no longer crash when a worker hits an unexpected error. Previously, if one of the internal FSMv2 components encountered an unexpected error, it could crash the entire process, taking your instance offline until the container restarted. The system now catches these errors and automatically recovers. If the same error keeps recurring, the affected component is stopped to prevent crash loops while the rest of the system keeps running.

• Instances no longer get stuck after configuration changes. When you changed a configuration and the system restarted an internal component, the new component could inherit a stale "shut down" signal from the old one and immediately stop itself. This could make the instance appear stuck or unresponsive after config changes. The shutdown signal is now properly cleared before starting a replacement.

The crash fix and the configuration fix are automatically active for all FSMv2 users and do not require additional configuration.

Other

• Updated container base image (Alpine) with security patches.
• Updated Go dependencies.

Setting up new instances is now more reliable, and we've added helpful context throughout the console -- from changelog links in the update dialog to descriptions on the Location card. Non-admin users also now get clear explanations for disabled actions instead of hidden buttons.

Improvements

• The umh-core update dialog now includes a link to the changelog so you can see what changed before upgrading

• The Location card now displays a brief description explaining what it is used for, with a link to the documentation for further details

• The login page now displays an ISO 27001 certification badge alongside the existing G2 badges, with all badges slightly smaller for a cleaner appearance

• Non-administrator users previously saw no "Add Instance" button at all. Now the button is always visible but disabled for non-admins, with a tooltip explaining "Only company administrators can add instances"

Fixes

• Fixed the install command dialog silently deleting your instance when closed without using the copy button -- closing the dialog now keeps your instance intact
• Fixed Docker Run and Docker Compose tabs incorrectly appearing for UMH Classic instances, which only have a single install command
• Fixed the disabled "Continue" button in the install dialog not explaining why it was disabled -- it now shows a tooltip
• Fixed the condition editor showing duplicate sections and broken layout when viewing Read Flow configurations
• Fixed the upgrade alert showing "v0.44.4 or higher" when the correct required version is v0.44.6
• Fixed rows in the Topic Browser's "Last Message" section having inconsistent heights when info tooltips were present
• Fixed the Data Flow Component editor using a hardcoded Benthos schema version instead of loading the correct version dynamically

Preview Improvements

These improvements apply to features currently available as a preview.

• You can now hover over template variables like location_path to see the actual resolved value based on your instance and bridge location (Enhanced Read Flow UI)

• Fixed conditions using strict JavaScript operators (===, !==) not being recognized by the condition editor (Enhanced Read Flow UI)
• Fixed users who registered through Auth0 being unable to perform certificate operations on V2 instances (V2 Certificates)

This release removes the 20-address limit for S7 bridges and adds per-slave register mapping for Modbus gateways.

New Features

• You can now assign specific registers to individual slaves when reading from a Modbus TCP/RTU gateway. Previously, all configured addresses were read from every slave, when different slaves expose different registers. This forced you to create separate connections per slave, which could exhaust the limited TCP connections available on many industrial gateways. Now you can add a slaveID field to each address entry to target a specific slave. Addresses without a slaveID are still read from all configured slaves, so existing configurations work without changes.

Improvements

• S7 bridges no longer have a 20-address limit. S7 connections now negotiate the actual PDU size with the PLC during connection and automatically split addresses into optimally-sized batches. If you previously had to create multiple connections to the same PLC because of the 20-address limit, you can now configure all addresses in a single connection. The batchMaxSize configuration field is deprecated and ignored — PDU size is handled automatically. Existing configs with batchMaxSize continue to work with a deprecation warning.

Fixes

• Updated container base image to Alpine 3.23.3, which includes security patches for OpenSSL vulnerabilities.

• Updated Go dependencies with security fixes for the Gin framework (GraphQL API) and other dependencies.

This release fixes container startup failures on fresh deployments and enables advanced variable types for bridge template configurations.

Fixes

• Containers now start reliably on fresh deployments. Previously, containers could fail to start with "permission denied" errors when services tried to write to temporary directories. This only affected fresh installations without Docker cache - upgrades and cached deployments were not impacted. The issue was caused by a known Docker BuildKit limitation that doesn't preserve directory permissions during multi-stage builds.

Improvements

• Bridge templates now support array and object variables. You can now use complex variable types in your protocol converter YAML configurations, not just simple text values. This enables bulk configuration patterns like defining a list of PLC addresses with their tag names, units, and data contracts as a single variable, then iterating over them in your template with {{ range .AddressMappings }}. This is particularly useful for S7 and OPC UA bridges where you need to map many addresses to tags.

Example usage in YAML:

templateInfo:
  isTemplated: true
  variables:
    - label: "AddressMappings"
      value:
        - Address: "DB1.DW20"
          TagName: "temperature"
          Unit: "celsius"
        - Address: "DB1.DW24"
          TagName: "pressure"
          Unit: "bar"

Then in your template: {{ range .AddressMappings }}{{ .Address }} -> {{ .TagName }}{{ end }}

Opening the Management Console with the same account in multiple browsers previously caused unpredictable behavior—messages would get routed to the wrong browser, causing buttons to stop responding and data to silently get lost. Now, logging in from a new browser automatically signs out your previous session. You'll see a clear message explaining what happened, so you always know which browser is active.

New Features

• Instance selectors now show offline instances grayed out with an "(offline)" label. Online instances appear at the top of the list, and offline instances cannot be selected. This prevents confusing behavior that occurred when trying to configure unavailable instances.

• The Management Console now detects which benthos-umh version your instance is running and loads the matching configuration forms. Previously, forms were hardcoded to one version, which could show incorrect fields if your instance version differed. If your version doesn't have an exact match, you'll see a banner explaining that some options may only be available in the YAML editor.

• When creating an account or accepting an invite, you now see a real-time checklist showing password requirements (12+ characters, uppercase letter, digit, and symbol). The checklist updates as you type, showing green checkmarks for requirements you've met. This replaces the previous approach of only showing errors after you tried to submit.

Improvements

• Privacy Policy links now point to the correct URL on signup, invite, and onboarding pages
• The page heading now correctly says "Data Flows" (plural) to match the sidebar menu
• The checkmark in dropdown menus is back on the left side where it was before, with improved contrast when highlighted

Fixes

• Fixed the X button in search/filter fields not being vertically centered
• Fixed sticky header in configuration pages that stopped working on pages with long content
• Fixed an issue where users were incorrectly signed out on certain API errors—now only authentication failures trigger sign-out
• Fixed profile pictures disappearing after page refresh and appearing squished when the sidebar is collapsed

We've made infrastructure improvements to help instances stay connected to the Management Console more reliably.

Some users have experienced instances appearing "offline" in the Management Console even though the instances were running correctly and processing data. This was caused by network interruptions that the legacy communicator couldn't recover from automatically.

What We've Done

• Backend servers are now more stable, reducing connection interruptions.

Preview: Try the New FSMv2 Communicator

If you've been experiencing connection issues (instances showing offline when they're actually running), you can try our completely rewritten communicator which has better recovery from network interruptions.

To enable, add to your config.yaml:

agent:
  communicator:
    useFSMv2Transport: true

Or via environment variable:

docker run -e USE_FSMV2_TRANSPORT=true ...

To verify it's working, exec into your container and check the logs:

docker exec -it <container-name> sh
grep fsmv2 /data/logs/umh-core/current

You should see log entries with fsmv2 and states like Syncing or TryingToAuthenticate.

Note: To disable later, you must explicitly set useFSMv2Transport: false in config.yaml.

We'd appreciate feedback on how this works for you. Once validated, FSMv2 will become the default communicator.

This release fixes issues affecting OPC UA data collection and JavaScript-based data processing.

Fixes

• OPC UA connections now properly rediscover nodes after reconnecting. Previously, if an OPC UA connection dropped and reconnected, the system would report zero detected nodes even though the server was available. Connections now correctly rediscover all nodes after network interruptions.

• JavaScript processors now correctly handle numbers from JSON data. When using the Tag Processor or Node-RED JS processor with structured JSON data, numeric values were incorrectly treated as objects instead of numbers. Calculations and comparisons involving numbers now work as expected.

You can now switch between dark, light, and system themes using the new theme toggle in the sidebar. The editor and all UI components automatically follow your selected theme, and your preference is remembered between sessions.

New Features

• When creating a new bridge, you can now copy the configuration from an existing bridge. The setup wizard shows all your bridges with their connection status, instance, and health indicators, making it easy to find the right one to copy.

• Instances now show an "Update" badge when a newer version is available. Click the badge to see update instructions for both Docker and Docker Compose deployments.

Improvements

• Protocol templates now show a "Verified" badge for curated protocols that have been tested and validated. Verified protocols are sorted to the top of the list, and all protocols include better descriptions to help you choose the right one.

• Green and orange status colors are now more readable in light mode
• Tooltips remain dark with high contrast in both light and dark modes
• Improved dialog scrolling for update instructions on smaller screens

You can now deploy UMH using Docker Compose as an alternative to single-container Docker or Kubernetes. This deployment option is perfect for production setups that need persistent storage and visualization without the complexity of Kubernetes.

The Docker Compose stack includes umh-core, TimescaleDB for time-series storage, PgBouncer for connection pooling, and Grafana for visualization. An optional Nginx configuration is provided for SSL/TLS termination.

View the Docker Compose documentation

Improvements

• Redpanda now runs with the --overprovisioned flag enabled by default, improving performance in containerized environments where CPU resources are shared

Fixes

• Fixed confusing documentation about Docker volume permissions when upgrading from versions before v0.44.0. Previously, the guide showed a chown command that didn't match the volume path in the docker run command. The guide now correctly explains how to fix permissions for both named volumes (using a temporary container) and bind mounts (using chown directly).

Creating new Data Flows is now guided and faster. When you select a protocol like OPC UA, S7, or Modbus, the form is automatically prefilled with common settings—you only need to enter your device's IP address and port. The new vendor/protocol selection cards make it easy to find your device type.

Improvements

• Protocol forms now display helpful examples below field descriptions, showing valid values at a glance (e.g., "Examples: ns=2;s=MyNode, ns=3;i=1001")
• The metadata panel in Topic Browser can now be resized by dragging the divider
• When creating bridges for Modbus or S7, the port field is now prefilled with the standard port (502 for Modbus, 102 for S7)
• Data Flow editors now only prompt "Unsaved changes" when you've made actual changes—formatting-only YAML changes no longer trigger false warnings
• Loading spinners now behave consistently across all pages, including when you only have umh-core instances
• The sidebar correctly remembers whether you left it open or closed between sessions

Fixes

• Protocol forms properly wait for prefilled data before rendering, preventing a flash of empty forms
• Fixed loading placeholder colors that changed after a recent update, restoring the original subtle appearance

We've upgraded our UI framework to deliver a more polished experience. Input fields are now more distinguishable with cleaner styling and proper backgrounds. Buttons have improved hover states for better feedback. Dark mode colors are more vibrant and consistent. Alert and error messages have a refined appearance with improved spacing and alignment throughout.

Code Editor Search

You can now search within YAML configurations using Ctrl+F (or Cmd+F on Mac) in the code editor. The Monaco editor search widget helps you quickly find and navigate to specific settings in large configurations.

Profile Picture Support

Your profile picture from Auth0 now appears in the sidebar and user menu. If no picture is available, you'll see your first name initial instead of email initial for a more personal touch.

Improvements

• App filters are now properly cleared when you sign out, so the next user sees a clean state

Fixes

• Fixed incorrect label in Topic Browser that showed "Physical Path" instead of "Location Path"

Action Required Before Upgrade

This version runs as a non-root user (UID 1000) instead of root for better security. Your existing data directory was created by root, so you must change its ownership before starting the new container:

sudo chown -R 1000:1000 /path/to/umh-core-data

If you skip this step, the container will fail to start and show an error message with the exact command to fix it.

For new installations, we recommend using Docker named volumes, which handle permissions automatically:

docker volume create umh-core-data
docker run -v umh-core-data:/data ...

New Features

• umh-core now runs all processes as a regular user instead of root. This limits potential damage if a vulnerability is ever exploited—even if an attacker gets into the container, they can't gain root-level access.

• New documentation at docs/production/security/ explains our security approach including threat model, shared responsibility, and compliance mapping for OWASP, NIST, and IEC 62443 standards.

Improvements

• You can now use uppercase letters in data flow names, giving you more freedom to match your existing naming conventions
• Delete buttons (like "Remove Level" in location forms) now use a warning color with appropriate hover states, making destructive actions more visually distinct
• The Docker install command for local umh-core now uses Docker named volumes instead of bind mounts, working consistently across Docker Desktop (Mac/Windows), native Linux, and Podman

Fixes

• Validation error messages in protocol configuration forms now display below input fields instead of beside them on narrow screens
• Fixed a visual glitch where the sidebar would briefly "pop in" when loading pages in Chrome
• Fixed inconsistent padding and hover states on icon buttons in side panels
• Fixed the chevron/arrow icon not displaying correctly when adding a stream processor
• Fixed the height and hover area of the sidebar collapse/expand toggle button

You can now access logs and metrics directly from any data flow component page. Click the "Logs" or "Metrics" button to open a side panel without leaving your current context. The panel appears on-demand instead of cluttering the interface with permanent tabs. Works across bridges, stream processors, and all data flow types.

Cleaner JavaScript Debugging Output

Bridge configuration debugging is now significantly easier. Previously, console output from JavaScript processors contained excessive escaped characters, making it difficult to read. The output now uses a Node.js-style representation that's more readable.

OPC UA Subscriptions Without Tag Hierarchy

Fixed an issue where subscribing directly to OPC UA child nodes (without a parent folder hierarchy) would cause validation errors or malformed topics with double dots. Topics are now properly constructed without the optional virtual path segment.

Improvements

• Status alerts and warnings now have a more consistent look with clearer explanations and collapsible technical details
• Buttons across the platform now follow a consistent design pattern with proper visual hierarchy

Fixes

• Delete dialogs now properly close when clicking the X button

You can now add debug_level: true to your bridge or data flow configurations to enable detailed debug logging when troubleshooting connection issues. This is especially helpful for diagnosing OPC UA problems, as it automatically enables verbose protocol-level debugging.

Improvements

• You can now use uppercase letters (A-Z) in bridge and component names, in addition to lowercase letters, numbers, dashes, and underscores.

Fixes

• Fixed a critical security vulnerability where malicious service names could potentially be used to access sensitive files on the system. Component names are now strictly validated to prevent path traversal attacks.

• Fixed production build configuration to properly disable CGO, improving binary performance and security.

When enabled via the Dynamic Protocol Forms feature flag, you can now configure S7, Modbus, and OPC UA bridges using a form-based UI instead of writing YAML. You'll see an "Input (UI)" / "Input (YAML)" toggle in the bridge configuration page. Click on any field to edit it directly, add addresses with the "+ Add" button, and the system automatically generates valid YAML. This is a preview feature—enable it only for testing.

SparkplugB Template Documentation

The default SparkplugB template now includes comprehensive documentation for all available metadata fields. You'll see inline comments explaining pre-sanitized fields that let you build custom hierarchical topic structures without writing manual sanitization functions. The template also now processes both DDATA and NDATA messages.

Improvements

• Your sidebar collapse/expand preference is now remembered across sessions
• The back button now works more like your browser's back button, making navigation feel more natural

Fixes

• Form validation errors no longer appear immediately when opening forms—they only show after you interact with a field
• Fixed an issue where flat NDATA metrics could cause filter problems

You can now duplicate conditions with a single click using the new duplicate button in the condition header. Previously, you had to manually copy and paste condition content. The delete button now includes a confirmation dialog to prevent accidental removals, and both buttons have helpful tooltips.

Improvements

• The code/UI toggle button in stream processors now matches the design of toggle components used elsewhere
• Fixed the tag processor editor incorrectly flagging console and other standard JavaScript APIs as errors

Fixes

• Instance names now appear immediately after you change them, with automatic cache refresh

Browse operations now use a global worker pool instead of creating separate pools for each starting node, preventing worker exhaustion. Previously, browsing 300 nodes with MaxWorkers=20 created 6,000 concurrent workers; now uses 20 workers total.

Improvements

• Getting started guide now uses "umh-core-data" folder name consistently across all documentation

• Automatic detection and optimization for Ignition, Kepware, Siemens S7-1200/S7-1500, and Prosys OPC UA servers with vendor-specific tuning

• Automatic detection when Siemens S7-1200 PLCs don't support DataChangeFilter (Micro Embedded Device Profile), preventing subscription errors

Fixes

• Fixed StatusBadFilterNotAllowed errors when subscribing to S7-1200 PLCs with data filters

• Resolved potential deadlocks during high-concurrency browse operations

Your production system no longer depends on Management Console connectivity to start. Previously, when Management Console was unreachable, the entire data infrastructure would fail to start. Now the system starts immediately and retries the connection in the background.

Faster Deployments with Large Configurations

Deployments complete 80-90% faster for customers with large configurations (e.g., 2,000+ OPC UA nodes). The system now logs configuration differences only at debug level instead of logging full structures on every change.

Fixes

• Fixed issue where location level fields (Level 2-5) showed placeholder text like "Your level 2 name" instead of actual saved values in Bridge edit views.

• When Management Console is unreachable, error logging is now limited to 1 error per second instead of 100 errors per second.

Fixed critical deadlock when browsing OPC UA servers with more than 100,000 nodes. Previously, browse operations would hang indefinitely and require container restarts. Now completes in under 1 second for 110k nodes.

OPC UA Performance Boost

Browse operations now complete up to 173x faster with 75-90% less memory usage. Connection times for large OPC UA servers are significantly faster, and CPU usage during browse is dramatically reduced.

Reduced Network Traffic

OPC UA bridges now automatically suppress duplicate value notifications (deadband filtering), reducing unnecessary network traffic and processing overhead. Enabled by default for all numeric node types.

Better JavaScript Debugging

Fixed console.log() and other console methods in JavaScript processors to work like browser environments. You can now properly debug JavaScript transformations by logging complex objects and seeing their actual structure.

Fixes

• Fixed timing mismatch that caused frequent EOF errors and connection failures in high-latency environments, particularly affecting sites in Japan and other regions with network latency above 200ms.

Fixed certificate generation issues that prevented connections to OPC UA servers with strict security requirements (like Ignition 8.3 using Basic256Sha256 security policy). Certificates now include all required security attributes. No configuration changes needed.

Fixes

• Fixed array handling to preserve whether values are numbers or text when passing data between systems. Previously, arrays like [1,2,3] could be confused with ["1","2","3"], causing data processing errors.

• Fixed crashes that could occur when restarting services with active Sparkplug B connections.

Resource limit blocking is now enabled by default to protect your system from overload. When your system is at capacity (high CPU/memory/disk usage), new protocol converter deployments will be blocked automatically with clear explanations.

You can disable this protection by setting agent.enableResourceLimitBlocking: false in your configuration, but we recommend keeping it enabled.

Improvements

• When deployment is blocked due to resource constraints, you'll now see the actual reason (like "CPU utilization critical - Reduce system load or disable resource limits") instead of a generic timeout message.

Fixes

• Your custom Redpanda resource limits (CPU cores and memory settings) now persist correctly across restarts. Previously, MaxCores and MemoryPerCoreInBytes were being overwritten to defaults.

• Protocol converters blocked by resource limits now show their configured connection details (IP address, port, flows) in the UI instead of "Connection Unavailable".

You can now use underscores in bridge and component names. Previously only lowercase letters, numbers, and hyphens were allowed. This makes it easier to match your existing naming conventions.

OPC UA Quality Tag Support

OPC UA data connections now capture quality/status codes from OPC UA servers, appearing in metadata as opcua_attr_statuscode. This helps verify data reliability from industrial equipment, especially useful for Ignition-based OPC UA servers.

Better Error Visibility

Stream processors now show clear warning messages when they receive incompatible data formats (like JSON objects instead of time-series data). Previously, data would silently disappear without explanation.

Better Health Status Display

Idle protocol converter states now correctly show as green/active instead of appearing unhealthy. This prevents false alarms when your data sources are legitimately inactive.

Fixes

• Container deployments no longer fail if Docker Hub is temporarily unavailable—the system now serves cached container images during outages
• The bridges interface now uses clearer terminology: "All" tab removed for simplicity, "Throughput" renamed to "Read Throughput", "Add Bridge" renamed to "Add Read Flow"
• Stream processors now generate smarter variable names for topics with the same name in different locations (e.g., area1Temperature instead of temperature_1)
• Form validation errors now clear immediately when you correct the input
• Company name now appears immediately after account registration
• Fixed critical issue where NULL-padded strings from S7 PLCs weren't being properly handled by the downsampler
• Fixed confusing IP address input instructions that made new users think they should type literal "{{ .IP }}" into form fields
• Components in idle states now correctly show as "active" health status instead of appearing degraded

Fixed login failures for enterprise customers using SAML-based single sign-on. Login now works regardless of email capitalization or verification status from SAML providers.

Multi-Organization Account Login

Fixed a bug that prevented users from accessing multiple organizations with the same email address. Messages are now routed correctly for all users regardless of how many organizations they access.

Instance Communication Stability

Fixed "All customer instances unreachable" errors for customers with many instances. The system now safely handles high-volume logging by limiting response sizes to 10MB and 200 messages per request.

Email Consistency in Onboarding

Email addresses are now normalized to lowercase throughout the entire onboarding process, preventing case-related mismatches.

The downsampler (report-by-exception) feature that's been automatically compressing your bridge data is now visible in the UI. You'll see a new "Advanced Processing" card in bridge configurations showing that compression is enabled by default. This feature reduces storage and network usage by filtering duplicate values—now you can see it's there and access documentation about how it works.

Faster Component Navigation

You can now click directly on component names to edit them, instead of using the three-dot menu every time. This works across Instances, Data Flows, and Models—just click the name in the table and you'll navigate straight to the configuration page. The three-dot menu is still available for additional actions.

Fixes

• Fixed misaligned "Remove Level" buttons in location configuration forms with consistent heights and proper spacing

Bridge Location Handling

• Fixed incorrect behavior when selecting or validating bridge locations
• Improved internal logic for bridge location validation to prevent mis-assigned or missing locations

User Management

• Added location validation when adding new users with direct feedback if the selected location doesn't meet requirements

UI Enhancements

• Aligned copy/paste buttons with better positioning for consistency across views
• Unified button styles throughout the interface for a cleaner look

Fixes

• Resolved minor visual inconsistencies in bridge configuration layout
• Cleaned up redundant validation messages

Topics now visually highlight when they receive new data, similar to MQTT Explorer. You'll see a brief highlight animation on the entire path whenever values update, making it easy to identify active data flows at a glance. Perfect for monitoring which topics are actively publishing without manually checking each one.