# Meshery v1.0.80

2026-11-16 · [Original release notes](https://github.com/meshery/meshery/releases/tag/v1.0.80) · [Full summary](https://ratatosk.io/en/releases/meshery/v1.0.80)

Security 22 · Breaking changes 34 · Important fixes 84 · Enhancements 6 · Other changes 71

## Security (22)

### Kubernetes context credentials redacted from `GetContext` responses

- Affects: setups using the `GetContext` endpoint (`GET /api/system/kubernetes/contexts/{id}`)

Before this change, the endpoint exposed authentication and cluster values—including private keys, certificates, and bearer tokens—in plain text over HTTP. Responses now omit those credentials while retaining context identifiers and public endpoint metadata.

> - [Server] Redact sensitive auth credentials in GetContext HTTP response @Supersaiyan101 (#23797)

### Credential deletion now enforces ownership

- Affects: credential deletion through `DefaultLocalProvider`
- Additional action: a valid authenticated user

Previously, deletion matched only the credential ID, so a request targeting another user’s credential could remove it without checking ownership. Now, attempts to delete another user’s credential or delete without a valid identity return not-found and leave the credential intact.

> - [Server] Enforce user ownership validation in credential get and delete @Supersaiyan101 (#23661)

### Meshery SSE stream timeouts fixed

- Affects: Meshery load-test SSE stream, `/api/providers/stream`, and `/api/system/adapter/events`

Load-test runs of two minutes or more could lose their final result frame, while the providers and adapter-events streams were cut off every two minutes. These streams can now continue beyond two minutes, with a fresh 30-second write deadline for each frame.

> - [Server] Stop the HTTP server timeouts from cutting off SSE streams and slow uploads @SandstromPL (#23634)

### Token-file permissions tightened

- Affects: Windows `mesheryctl` token files and token files created on non-Windows platforms

On Windows, existing token files with broader access are restricted to the current user, and updates use a private temporary file. If access restrictions cannot be verified or an update fails, an error is reported and the existing token remains unchanged; newly created non-Windows token files use mode `0600`.

> - fix: restrict Windows token file permissions @0xmaster7 (#23380)

### Unsafe export paths fixed in `mesheryctl design export`

- Affects: designs exported with `mesheryctl design export`

Names containing `/`, `\`, or `../` could create subdirectories or write the export outside `--output`. Unsafe characters now become `_`, keeping the file directly inside `--output`; already-safe names such as `my-design` keep the same filename.

> - [mesheryctl] sanitize the design name in design export file paths @rakesh0x (#23557)

### Undefined token assignments rejected by mesheryctl

- Affects: `mesheryctl system token set` assignments using names absent from `tokens:`

Previously, `system token set <name>` saved names without a matching `tokens:` entry, so a subsequent request could send the default token to that context's server. Now the command returns an error and leaves the context unchanged.

> - [mesheryctl] Reject unknown tokens in token set and in-use tokens in token delete @Asheesh7298 (#23457)
> - [mesheryctl] Reject unknown tokens in token set and in-use tokens in token delete @Asheesh7298 (#23457)

### Windows token files restricted to the current user

- Affects: Windows `mesheryctl` token files

Previously, `Chmod(0600)` did not remove existing or inherited ACLs, so token files in shared directories could remain readable by other users. Writes and rewrites through `WriteTokenFile` now apply a protected ACL for the current user's SID, leaving the previous token intact if the restriction fails; Unix behavior is unchanged.

> - [mesheryctl] Enforce owner-only token file on Windows via explicit DACL @Pastoray (#23507)
> - [mesheryctl] Enforce owner-only token file on Windows via explicit DACL @Pastoray (#23507)

### Nil-provider and logger panics fixed in performance profile endpoints

- Affects: performance profile endpoints

A missing or typed-nil provider, or an unavailable handler logger, could cause a server panic when an endpoint handled a request. Missing providers now receive an `ErrNilProviderCode` 500 response, and guarded logger calls no longer panic.

> - [Server] Guard nil provider and nil logger in performance profile handlers @antcybersec (#23772)

### Docker Extension proxy authentication-token log exposure fixed

- Affects: Docker Extension proxy request and backend-error logging (Inferred)

Authentication tokens were exposed in request and backend-error logs. They are now redacted from those logs, including when query parameters are malformed; the delete path also no longer logs the empty token or channel.

> - [Docker Extension] Don't hang token store and logout when no websocket client is connected @CyberRik (#23765)

### Nil-pointer panics fixed in MeshSync endpoints

- Affects: the four MeshSync endpoints

Nil or uninitialized providers or persisters, and unavailable logger paths, could cause nil-pointer panics in these handlers. Missing providers or persisters now produce a 500 `ErrNilProviderCode` response, nil logger paths no longer panic, and malformed inputs or missing query parameters still receive validation errors.

> - [Server] Guard nil provider, persister, and logger across MeshSync handlers @antcybersec (#23770)

### Nil-provider and logger panics fixed in schedule and organization handlers

- Affects: schedule endpoints and `GetOrganizations`

When these handlers encountered a nil provider, including a typed nil provider, or made logger calls with an uninitialized logger, they could panic. Those paths no longer panic, while malformed schedule requests and missing schedule IDs retain appropriate error responses.

> - [Server] Guard nil provider and nil logger in schedule and organization handlers @antcybersec (#23768)

### Nil-dereference panics in `EvaluateRelationshipPolicy` fixed

- Affects: relationship policy evaluation

When a user or provider was missing, `h.SystemID` was nil, or `h.evalTracker` was uninitialized, the handler could panic on nil dereferences. It now returns HTTP 401 for a missing user and HTTP 500 for a missing provider, falls back to `uuid.Nil` for a nil system ID, and initializes a nil evaluation tracker.

> - [Server] Guard unchecked nil dereferences, unbounded body reads, and detached goroutine panics in EvaluateRelationshipPolicy @Amansingh0807 (#23664)

### `EvaluateRelationshipPolicy` request body limits enforced

- Affects: requests to `EvaluateRelationshipPolicy`

Previously, bodies over 4 MB were read without a size cap, risking memory exhaustion, and empty bodies were not specially rejected before JSON decoding. Now, oversized bodies receive HTTP 413 and empty bodies receive HTTP 400.

> - [Server] Guard unchecked nil dereferences, unbounded body reads, and detached goroutine panics in EvaluateRelationshipPolicy @Amansingh0807 (#23664)

### Credential GET ownership verification added

- Affects: credential GET requests through `DefaultLocalProvider`

Previously, `GetUserCredentialByID` returned credentials without checking whether `credential.UserId` matched the authenticated caller's `user.ID`, allowing another user's credential to be disclosed. Now, requests for another user's or an unowned credential, or requests without a resolved caller identity, return 404.

> - [Server] Enforce user ownership validation in credential get and delete @Supersaiyan101 (#23661)

### [Security status unclear] Nil-provider panic in `GetUsersKeys` fixed

- Affects: requests handled by `GetUsersKeys`

When provider initialization faults or a request reaches the handler during boot, a nil `provider` previously caused a nil-pointer panic that could take down the server. Such requests now receive HTTP 500 with `ErrNilProvider()` instead of crashing the process.

> - fix(server): guard nil provider and validate org ID in GetUsersKeys @hemv-857 (#23645)

### ValidationHandler oversized-request crash fixed

- Affects: the `/api/registry/validate` and `/api/meshmodels/validate` endpoints

Previously, oversized requests could be read without a limit and cause OOMKilled crashes. Requests over 4 MB now receive HTTP 413 instead of being read without a bound.

> - [Server] Bound request body size and guard against DoS in ValidationHandler @Supersaiyan101 (#23589)

### `ValidationHandler` returns HTTP 400 for invalid requests

- Affects: the `/api/registry/validate` and `/api/meshmodels/validate` endpoints

Client read failures previously returned HTTP 500. Standard JSON whitespace is trimmed before validation.

> - [Server] Bound request body size and guard against DoS in ValidationHandler @Supersaiyan101 (#23589)

### `DeleteConnection` UUID validation and nil-reference handling

- Affects: the `DeleteConnection` endpoint

Before this change, malformed or nil connection UUIDs lacked boundary validation, and missing user, provider, or system information could cause dereference panics. Now malformed or nil UUIDs return HTTP 400; nil-provider requests return HTTP 500, while missing user or system information no longer causes those panics.

> - [Server] Validate connection ID and guard nil dereference in DeleteConnection @Supersaiyan101 (#23404)

### Nil-provider handling fixed for extension operations

- Affects: extension install/remove handlers and capability persistence

When an unauthenticated request reached an install or remove handler during partial server bootstrap with no provider, the unchecked dereference could panic and terminate the server. The handlers now return HTTP 500 in that case, and capability persistence safely skips calls with a nil provider.

> - [Server] Guard nil provider dereference and unbounded body reads in extension install and remove handlers @Amansingh0807 (#23395)

### Extension handlers reject oversized bodies and avoid nil-provider panics

- Affects: extension installation and removal

Previously, reaching either handler without an initialized provider—including through unauthenticated access or during partial server bootstrap—caused an unhandled panic that terminated the server, and request bodies were read without a size bound. Now, bodies over 4 MB receive HTTP 413, and request bodies are closed.

> - [Server] Guard nil provider dereference and unbounded body reads in extension install and remove handlers @Amansingh0807 (#23395)

### mesheryctl token-file writes support long paths on Windows

- Affects: Windows setups writing token files to long paths

> - fix: restrict Windows token file permissions @0xmaster7 (#23380)

### [Security status unclear] MeshSync watched-resource access restored

- Affects: Meshery Operator Helm chart installations with MeshSync running in operator mode
- Additional action: `meshsync.serviceAccount.dedicated` enabled

Previously, MeshSync received `forbidden` errors on watched resources, breaking discovery. It now gets read access, including open-ended list/watch access for CRDs; with the default shared ServiceAccount, it still has the operator’s write permissions.

> - [Helm] Give MeshSync read-only RBAC, expose Broker service values, clean up CRs on uninstall @yi-nuo426 (#23259)

## Breaking changes (34)

### Cloned filters get fresh metadata

- Affects: filter cloning with the Local provider (Inferred)

When the Local provider clones a filter, the clone previously inherited the source’s `created_at` and `catalog_data`. It now gets its own creation time and no catalog data, while retaining the source’s file, resource, and location.

Evidence: “The copy +// keeps the source's file, resource and location, but gets its own creation +// time and no catalog data.” https://github.com/meshery/meshery/pull/23451

> - [Server] Clone designs and filters as fresh records @Asheesh7298 (#23451)

### Pattern-resource lookups exclude deleted records and report missing records as not found

- Affects: pattern-resource lookups by ID or attributes

Previously, missing records from ID and attribute lookups were not mapped to not-found, and ID lookups could include soft-deleted records. Other database errors are now reported as read errors.

Evidence: “Pattern resource lookups now exclude deleted records and distinguish missing records from database errors.” https://github.com/meshery/meshery/pull/23667

> - [Server] Harden pattern resource persistence @0xmaster7 (#23667)

### Pattern-resource deletion returns not-found for missing or deleted resources

- Affects: pattern-resource deletion

This occurs when a delete request targets a missing or already soft-deleted ID and the update affects no rows.

Evidence: “Deleting a resource that is missing or already deleted now returns a not-found error.” https://github.com/meshery/meshery/pull/23667

> - [Server] Harden pattern resource persistence @0xmaster7 (#23667)

### Kubernetes context deletion stops on snapshot-loading errors

- Affects: Kubernetes context deletion

Previously, if the Kubernetes context-list snapshot could not be loaded, the handler returned HTTP 200 and started the delete transition anyway. Now it returns HTTP 500 before deletion starts, allowing the caller to retry.

Evidence: “A failed Kubernetes context snapshot previously logged an error but returned 200 and started the delete transition without a usable list. The handler now reports the snapshot error before starting deletion, allowing the caller to retry.” https://github.com/meshery/meshery/pull/23613

> - [Server] Report context snapshot failures before deletion @0xmaster7 (#23613)

### Context deletion now returns explicit HTTP responses

- Affects: context deletion through `DELETE /api/system/kubernetes/contexts/{id}`

Previously, successful deletion sent no status or body; if machine initialization failed, the handler logged and persisted an error event but sent no response. It now returns HTTP 200 on success and HTTP 500 with `ErrFailToDelete` on initialization failure.

> - [Server] Write HTTP status 200 and error response in DeleteContext @nishant-devv (#23563)

### Request-body limits and validation for `SavePerformanceProfileHandler`

- Affects: performance profile save requests

Bodies over 4 MB could previously be silently truncated; they now receive HTTP 413. Empty, whitespace-only, and `null` payloads receive HTTP 400.

Evidence: “Enforced defensive checks against empty, whitespace-only, and `"null"` payloads with `ErrRequestBody` (`HTTP 400 Bad Request`).” https://github.com/meshery/meshery/pull/23561

> - [Server] Guard invalid UUID status codes, missing 404 responses, and unchecked token extraction in performance profile handlers @Amansingh0807 (#23561)

### Missing performance-profile deletions now return HTTP 404

- Affects: performance-profile deletion
- Action required: No error-code regeneration or schema change is required.

Previously, deleting an ID that was never stored locally could return HTTP 200 with a stub profile, falsely indicating success. A provider-reported missing profile also now returns HTTP 404.

> - [Server] Return 404 when deleting a performance profile that does not exist @Shikhyy (#23544)

### Preference persister input validation and missing-delete handling

- Affects: preference writes and deletes through `SessionPreferencePersister` (Inferred)

Before this release, writes could panic with an uninitialized database handler, an empty `userID`, or nil preference data; deleting a preference that did not exist could return silent `nil` and appear successful as HTTP 200. Those write inputs are now checked before database queries, and deleting a missing preference returns `ErrResultNotFound`.

Evidence: “`DeleteFromPersister` now verifies `res.RowsAffected == 0` and returns `ErrResultNotFound` when the target preference does not exist instead of returning silent `nil` (200 OK).” https://github.com/meshery/meshery/pull/23536

> - [Server] Guard nil database handler, validate userID, and check rows affected in preference persister @piyxsh31 (#23536)

### HTTP 404 handling for missing environments

- Affects: environment records

Previously, updating a nonexistent environment ID could silently create a row, while get, update, and delete requests for missing IDs could return 500 or 502. Now these requests return 404, and an update leaves the environment absent.

> - [Server] Honour Environment purpose property and align SystemDatabaseModels to v1beta3 @SachinyadavAug20 (#23429)

### 404 responses for credential updates and deletes with no live match

- Affects: credential update and delete requests

Previously, updating an already-deleted credential could return 200 without changing it, while repeating a delete returned 200 and changed `deleted_at`; an unknown ID could produce 500. Repeated deletes now leave the original deletion record unchanged.

Evidence: “On the local provider, `UpdateUserCredential` filtered its update on `deleted_at is NULL` but read the row back without it, so updating a deleted credential matched nothing and still returned the deleted row. The handler answered 200 and nothing changed. `DeleteUserCredential` matched by id only, so a second delete also answered 200 and moved `deleted_at`. An id that never existed came back as a bare `record not found`, which the handlers turned into a 500.” https://github.com/meshery/meshery/pull/23503

> - [Server] Return 404 when updating or deleting a missing credential @cortex36a (#23503)

### Connection actions now reject invalid IDs and handle missing data safely

- Affects: `POST /api/integrations/connections/{connectionId}/actions`

Previously, malformed or empty IDs were converted with `uuid.FromStringOrNil`, and absent provider, user, or system data could cause panics during MeshSync actions. Now malformed or nil IDs return HTTP 400, an unavailable provider returns HTTP 500, and actions with missing user or system data continue without a dereference panic.

Evidence: “Guards `provider == nil` before executing `setMeshsyncDeploymentModeAction` or `flushMeshsyncAction`, returning HTTP 500 with `"provider not initialized"` rather than panicking” https://github.com/meshery/meshery/pull/23417

> - [Server] Validate connection ID and guard nil dereferences in PerformConnectionAction @piyxsh31 (#23417)

### `mesheryctl perf result` uses the configured command output writer

- Affects: `mesheryctl perf result` setups with a configured command output writer (Inferred)

Previously, expanded results and their formatter wrote to raw stdout, even when the configured command output writer was different. After the change, that output goes to the configured writer instead.

Evidence: “Routed expanded result outputs and the output formatter to `cmd.OutOrStdout()` instead of raw `fmt.Printf` / `os.Stdout`.” https://github.com/meshery/meshery/pull/23729

> - [mesheryctl] Guard nil TestStartTime, fix dead check, and standardize timestamp in perf result (#23727) @muskan-creates352 (#23729)

### Remote design imports use URL-path basenames by default

- Affects: remote design imports without an explicit `--name`
- Action required: A remote URL whose path has no filename requires an explicit `--name`.

Query strings and fragments are excluded from the derived name, while the full URL is still sent with the request.

Evidence: “Derive the default name for remote imports from the URL path, excluding query and fragment while keeping the URL intact for the request.” https://github.com/meshery/meshery/pull/23703

> - [mesheryctl] Exclude URL query from imported design name @0xmaster7 (#23703)

### Log pod selection now prioritizes exact and cleaned-name matches

- Affects: `mesheryctl` log pod selection by specified pod name

Previously, a name matching multiple pods could silently select the first substring match, even if it was not the intended pod. Exact or cleaned-name matches can now select a different pod without an error; multiple matches at the highest priority and missing pods return errors.

Evidence: “Prefer exact pod names and exact cleaned names over substring matches when selecting logs. Report ambiguous matches instead of silently selecting the first pod.” https://github.com/meshery/meshery/pull/23666

> - [mesheryctl] Resolve log pod names unambiguously @0xmaster7 (#23666)

### Filter name searches now prompt on ambiguous matches and error on no matches

- Affects: filter name searches with `mesheryctl filter view`

Previously, multiple matches opened the first result automatically, while no matches logged a not-found message and returned success.

> - fix(mesheryctl): select filters by full name @0xmaster7 (#23580)

### Filter-name search now retrieves results across pages

- Affects: filter-name searches with `mesheryctl filter view`

Previously, the search selected from the API’s default-sized result page, and a search with no matches could silently succeed. Now it prompts when multiple filters match, returns a typed not-found error when none match, and reports errors if results cannot be fetched or are incomplete.

Evidence: “Searches can retrieve results across multiple pages, and report errors when results cannot be fetched or are incomplete.” https://github.com/meshery/meshery/pull/23580

> - fix(mesheryctl): select filters by full name @0xmaster7 (#23580)

### Method-changing redirects refused by `utils.MakeRequest`

- Affects: `mesheryctl` requests made through `utils.MakeRequest`

When a DELETE, POST, or PUT received a 301, 302, or 303, the client could resend it as a bodyless GET; a GET route at the destination could run and the command could report success without performing the original operation. These redirects are now refused: most return `ErrRedirectChangesMethod`, while refused 302 redirects return `ErrInvalidToken`; redirects that preserve the method remain followed.

Evidence: “`utils.MakeRequest` used a default `http.Client{}`. net/http answers a 301, 302 or 303 to a DELETE, POST or PUT by re-sending the request as a GET without its body. When the new path also has a GET route, the GET runs and the command reports success.” https://github.com/meshery/meshery/pull/23405

> - [mesheryctl] Report a method-changing redirect instead of following it @Asheesh7298 (#23405)
> - [mesheryctl] Report a method-changing redirect instead of following it @Asheesh7298 (#23405)

### `mesheryctl` refuses deletion of tokens used by contexts

- Affects: meshconfig setups with tokens referenced by contexts

When `system token delete <name>` is run for one of these tokens, the command lists the affected context names in sorted order and refuses deletion. Previously, it removed the token entry while leaving those contexts referring to a missing token; deleting an unused token continues to work as before.

Evidence: “`DeleteTokenFromConfig` refuses to delete a token that a context uses and lists those contexts in sorted order: `token "prod" is still used by context(s) prod, staging; set another token on them first with ...`.” https://github.com/meshery/meshery/pull/23457

> - [mesheryctl] Reject unknown tokens in token set and in-use tokens in token delete @Asheesh7298 (#23457)

### Design export filenames now reflect export type

- Affects: setups using `mesheryctl design export`, particularly `oci` or `original` exports (Inferred)

Previously, OCI exports had no extension, `original` Kubernetes-manifest and Docker Compose exports used `.tar.gz` despite containing YAML, and an unknown `--type` such as `helm` silently exported the current design as YAML. Now `oci` exports use `.tar`, `original` Helm-chart exports use `.tar.gz`, and other `original` exports use `.yaml`; `current` exports remain `.yaml`.

Evidence: “The Helm and `current` filenames don't change.” https://github.com/meshery/meshery/pull/23447

> - [mesheryctl] Name design export files by type and reject unknown --type @mayank609 (#23447)
> - [mesheryctl] Name design export files by type and reject unknown --type @mayank609 (#23447)

### Invalid `--type` values rejected by design export

- Affects: `mesheryctl design export` (Inferred)

A mistyped value such as `helm` used to silently export the current design as YAML. The command now returns `ErrInvalidArgument` for unsupported values.

Evidence: “`RunE` checks `--type` against `current`, `original`, `oci` before any network call and returns `ErrInvalidArgument` naming the valid values.” https://github.com/meshery/meshery/pull/23447

> - [mesheryctl] Name design export files by type and reject unknown --type @mayank609 (#23447)

### Consistent saved filenames for `view --save`

- Affects: `view --save` for components, connections, environments, filters, models, relationships, or workspaces

Previously, saved filenames were passed through or only partly sanitized, inconsistently across resource types. Now spaces and filesystem-sensitive characters become underscores, leading and trailing whitespace is trimmed, and unusable names fall back to an available ID or, for several item types, `unknown`.

Evidence: “When a name cannot produce a usable filename, the filename falls back to an item ID where available. For several item types, `unknown` is used if neither the name nor ID produces a usable filename.” https://github.com/meshery/meshery/pull/23436

> - fix(mesheryctl): sanitize filesystem-unsafe characters in view --save filenames @ThakurUjjwal1706 (#23436)

### Notification Center filters non-serializable UI payload values

- Affects: Notification Center integrations using `openNotificationCenter` (Inferred)

Previously, the events slice stored the payload as received, including an icon component, which caused a Redux warning. Non-serializable fields such as `Date`, `Map`, `Set`, class instances, and circular objects are now dropped; the icon is stored by name and resolved during rendering.

Evidence: “This screens the payload in openNotificationCenter before it reaches state. Only serializable values are kept, a Date, Map, Set, class instance or circular object gets dropped. The icon is stored as a name now, and the NotificationCenter resolves it to the component when rendering.” https://github.com/meshery/meshery/pull/23082

> - [UI] Keep non-serializable values out of Notification Center state @krishbuilds01 (#23082)

### Default-on pre-delete cleanup for Broker and MeshSync resources

- Affects: Meshery Operator Helm chart installations with release-owned Broker or MeshSync custom resources

Previously, `helm uninstall` removed the operator and these resources in one pass, leaving them stuck `Terminating` on finalizers only the operator removed. The hook now deletes them while the operator is running, waits 90 seconds, then removes finalizers if needed; cleanup can be disabled with `customResourceCleanup.enabled`.

Evidence: “**Symptom.** `helm uninstall` deletes the operator and the Broker/MeshSync resources in one pass. The resources then hang `Terminating` on `broker.meshery.io/finalizer` and `meshsync.meshery.io/finalizer`, which only the operator removes.” https://github.com/meshery/meshery/pull/23259

> - [Helm] Give MeshSync read-only RBAC, expose Broker service values, clean up CRs on uninstall @yi-nuo426 (#23259)

### Paginated `GetID` lookups find later design and filter matches

- Affects: Short-prefix `mesheryctl` lookups against servers with more than 10,000 designs or filters

Matches beyond the first page could be missed, causing `ValidId` to treat an ambiguous prefix as unique and risk unintended deletion, or fail to resolve a unique prefix. `GetID` now collects IDs across pages, skips duplicates, and reads both `totalCount` and `total_count`; if it cannot confirm the full result within 50 pages, it returns an incomplete-lookup or page-limit error.

Evidence: “When resolving short ID prefixes, `GetID` previously requested a single batch of `pagesize=10000` without inspecting total item counts. On servers with more than 10,000 designs or filters, matches past the first page were missed, causing `ValidId` to falsely conclude a prefix was unique (risking unintended deletions in `design delete`) or fail to resolve a unique prefix on subsequent pages.” https://github.com/meshery/meshery/pull/23206

> - [mesheryctl] Paginate ID prefix lookup across multiple pages in GetID @bodapatisaikrishna (#23206)
> - [mesheryctl] Paginate ID prefix lookup across multiple pages in GetID @bodapatisaikrishna (#23206)

### Go `Extensions` field renamed to `Account`

- Affects: Go code using the `Extensions` model field (Inferred)

The field was previously named `Acccount` and is now `Account`; its JSON tag remains `json:"account,omitempty"`, preserving wire and API compatibility.

Evidence: “The JSON struct tag remains `json:"account,omitempty"`, ensuring 100% wire and API backward-compatibility.” https://github.com/meshery/meshery/pull/23795

> - [Server] Fix misspelled 'Acccount' field in Extensions model @Amansingh0807 (#23795)

### Meshery provider validation follows the selected context

- Affects: `--context` targeting a context other than the saved `current-context`

Previously, a provider offered by the selected context could be rejected, while one offered only by the current context could be accepted and used to redeploy the selected context. An invalid context name now returns `ErrSettingTemporaryContext` before a request is made.

Evidence: “`system provider switch --context X` redeploys `X`, but `validateProvider` asked the current context's Meshery server for its providers, because `GetProviderInfo` uses `mctlCfg.GetBaseMesheryURL()`, which follows `CurrentContext`. A provider `X` offers could be rejected, and one only the current context's server offers was accepted and then used to redeploy `X`.” https://github.com/meshery/meshery/pull/23786

> - [mesheryctl] Validate the provider on the --context server in provider switch @Eternity0207 (#23786)

### Filter cloning rejects empty names

- Affects: filter cloning

Clone requests for an existing filter with an empty name receive a bad-request response.

> - [Server] Return 404 for missing filters and validate path UUIDs in filter handlers @antcybersec (#23726)

### Malformed workspace and environment path IDs now return HTTP 400

- Affects: workspace and environment routes that take path IDs

Previously, malformed IDs reached the Local provider, which returned HTTP 502 on 18 routes and HTTP 404 on three, where the ID was converted to the nil UUID. Now all 21 routes return HTTP 400 with `models.ErrInvalidUUID` before calling the provider; well-formed IDs that do not exist are unaffected.

Evidence: “A malformed ID in the path of a workspace or environment route used to reach the provider. With the Local provider that came back as `502 Bad Gateway` on 18 routes and as `404` on 3, where the ID was quietly turned into the nil UUID. All 21 routes now answer `400` with the existing `models.ErrInvalidUUID` (`meshery-server-1432`) before the provider is called.” https://github.com/meshery/meshery/pull/23637

> - [Server] Return 400 for a malformed ID on workspace and environment routes @himanishpuri (#23637)

### Schedule handlers reject empty inputs

- Affects: the schedule handler endpoints

Calls to `DeleteScheduleHandler` or `GetScheduleHandler` with an empty `scheduleID` now receive HTTP 400 before the provider is queried. A nil or empty decoded body sent to `SaveScheduleHandler` also receives HTTP 400 before the provider operation proceeds.

Evidence: “Added validation in `DeleteScheduleHandler` and `GetScheduleHandler` to ensure `scheduleID` is not empty before querying the provider, returning `http.StatusBadRequest` (400) if empty.” https://github.com/meshery/meshery/pull/23540

> - [Server] Guard nil provider and validate schedule ID across schedule handlers @piyxsh31 (#23540)

### Preference persister errors now use MeshKit constructors

- Affects: setups using the preference persister (Inferred)

Read, save, and create failures are returned using `ErrDBRead`, `ErrDBPut`, and `ErrDBCreate`, respectively.

Evidence: “Preference read, save, and delete operations now return more specific errors when they fail.” https://github.com/meshery/meshery/pull/23536

> - [Server] Guard nil database handler, validate userID, and check rows affected in preference persister @piyxsh31 (#23536)

### Configuration file close errors are now reported

- Affects: `mesheryctl` configuration-file creation or backup

Previously, close errors in these paths were ignored. Now, `CreateConfigFile` returns a close error, while the backup path passes it to `Log.Fatal`.

> - [mesheryctl] Fix HTTP response body and file descriptor leaks in utils @muskan-creates352 (#22645)

### Adapter deploy and validate waiters now handle response events directly

- Affects: adapter deploy or validate flows that wait for response events (Inferred)

If a waiter timed out before its event-reading goroutine finished, that goroutine could remain blocked trying to send its result. The wait loop now returns on matching success or error events and stops its timer on return.

> - fix: restrict Windows token file permissions @0xmaster7 (#23380)
> - fix: restrict Windows token file permissions @0xmaster7 (#23380)

### `mesheryctl filter delete` now requires exactly one argument

- Affects: the `mesheryctl filter delete` command

Previously, zero arguments could reach the handler, and multiple arguments were not rejected by argument validation. Now, either count produces a validation error before the handler runs.

Evidence: “This PR enforces exact argument validation (`cobra.ExactArgs(1)`) on `mesheryctl filter delete`, rejecting invocations with zero arguments or multiple arguments before executing the command handler.” https://github.com/meshery/meshery/pull/23519

> - [mesheryctl] Enforce exact argument validation in filter delete @AdityaChaudhary5412 (#23519)

### mesheryctl rejects context names containing `.`

- Affects: mesheryctl contexts whose names contain `.` (Inferred)

Previously, a dotted context could be created even though Viper could not read it back. The failed addition now leaves the existing configuration unchanged.

Evidence: “Adding a context name that contains a period now returns an error and leaves the existing configuration unchanged.” https://github.com/meshery/meshery/pull/23441

> - [mesheryctl] Reject dotted context names @0xmaster7 (#23441)

## Important fixes (84)

### False save events fixed for unsaved filter conversions and URL imports

- Affects: filter conversion and URL-import requests

When `Save` was false, filter conversions could record a “filters saved” event, and URL imports could record a save event despite not saving a filter. These operations now return normalized filter output without saving; URL imports with `Save` true still persist the filter and save event.

> - [Server] Prevent false "filters saved" events on unsaved filter conversions @antcybersec (#23711)

### MeshSync deletion now removes associated resource data

- Affects: MeshSync resource deletion or cluster-level flushing

Previously, deleting a resource could leave its metadata, spec, status, labels, and annotations behind; deterministic IDs let orphaned labels reattach when the resource was discovered again. Deletion now removes those associated rows in the same transaction, and a failed deletion rolls back instead of leaving a partial deletion.

> - [Server] Delete a MeshSync resource's child rows with it @Asheesh7298 (#23216)

### Fresh metadata and identity for cloned designs

- Affects: design cloning with the Local provider (Inferred)

Previously, clones carried over the source's creation time, usage counters, and catalog data, while their design files kept the source's `id` and `name` and advanced to its next version. Clones now start with zero counters, empty catalog data, and a design file set to the clone's identity and version `0.0.1`; the source design remains unchanged.

> - [Server] Clone designs and filters as fresh records @Asheesh7298 (#23451)

### Local `GetConnections` type filter fixed

- Affects: Local-provider connection-list queries using `?type=`

Previously, the filter matched the legacy `type` column, so server-saved connections populated only in `connection_type` were not returned. It now matches those saved connections, while registrant rows continue to match.

> - [Server] Filter local connections by connection_type @mayank609 (#23551)

### Connection `type` filters now match saved connection types

- Affects: connection-list filters

When filtering saved connections by `type`, the filter could miss them because their persisted type is stored as `connection_type`. Key/value and JSON-object filters are both accepted; utility filters support `connection_type` and `sub_type`.

> - [Server] Filter connections by persisted type @0xmaster7 (#23668)

### [Silent failure] Pattern-resource persistence and list error handling fixed

- Affects: pattern-resource save, delete, and list operations

Previously, some database errors during persistence or list operations were silently treated as success. These operations now return errors when the database handler is unavailable, the resource is nil, or a database operation fails.

> - [Server] Harden pattern resource persistence @0xmaster7 (#23667)

### `ReadCapabilitiesForUser` returns not-found for missing rows

- Affects: user capability reads

When a user's `user_capabilities` row was absent, the read returned `None` instead of the expected `meshery-server-1307` result-not-found error. `ProviderCapabilityHandler` continues to fetch capabilities from the remote provider when a read errors.

> - [Server] Guard nil capabilities payload and return ErrResultNotFound in UserCapabilitiesPersister @Yasas-Sri (#23644)

### State-machine cleanup preserves re-imported tracker entries

- Affects: Kubernetes contexts and connections

When a context or connection was re-imported during asynchronous teardown of its previous instance, the older teardown could remove the new state machine from the tracker. Now it removes the tracker entry only if it still points to the instance being torn down.

> - [Server] Introduce CompareAndDelete on StateMachine tracker to prevent re-imported connection eviction @nishant-devv (#23632)

### MeshSync cleanup fixed for deleted Kubernetes contexts and connections

- Affects: Kubernetes contexts and connections managed by MeshSync

Previously, cleanup could abort, leaving MeshSync records in the database and emitting a misleading context-selection event; disconnected contexts could also inflate the reference count. Now resources are flushed when the deleted context was the last active reference and retained when another active context uses the server, and context lists can be filtered by connection status.

> - [Server] Prevent MeshSync database leak on Kubernetes context deletion @nishant-devv (#23615)

### MeshSync cleanup reference counting fixed for context deletion

- Affects: Kubernetes context deletion with MeshSync cleanup

When cleanup counts references to a cluster, it now uses a complete snapshot that includes disconnected contexts. MeshSync data is retained while other contexts share the server and removed when the deleted context was the last associated context; failed or incomplete reads do not authorize cleanup.

> - fix(server): retain cluster context snapshot for cleanup @0xmaster7 (#23579)

### `POST /api/filter` response now includes decoded filters

- Affects: clients calling `POST /api/filter`

Successful save, convert-without-saving, and URL-import requests returned `[]` even when filters had been decoded. These responses now contain the filters, while a provider `null` response still returns `[]`.

> - [Server] Return the saved filters from POST /api/filter @antcybersec (#23600)

### Policy binding selector state isolation fixed

- Affects: designs with multiple bindings for the same relationship rule

Later iterations could overwrite earlier relationships’ `Match.From` and `Match.To` component IDs through shared selector state. Each identified relationship now retains isolated selector match IDs, so changes to one selector do not mutate another’s state.

> - [Server] Prevent in-place match selector mutation and cross-relationship state corruption in policy binding @nishant-devv (#23585)

### [Silent failure] Relationship cleanup now removes dotted Kubernetes keys

- Affects: relationship-mutated paths ending in Kubernetes annotation or label keys containing dots

Deleting a relationship whose mutated path ended in a dotted annotation or label key could silently leave the injected value in component configuration, with no error or log line. Cleanup now removes that value; ordinary nested paths and numeric array-index segments retain their stated behavior.

> - [Server] Escape literal keys in relationship cleanup paths @Karman580 (#23525)

### [Silent failure] SQLite environment migration adds and preserves `purpose`

- Affects: SQLite-backed system databases and environments created or updated through the local provider

Previously, the `v1beta1` migration could leave SQLite's `environments` table without `purpose`, causing later saves to fail, and updates could overwrite an environment's purpose with an empty value. The `v1beta3` migration adds the column; existing records resolve to user purpose unless administrative, new environments default to user purpose, and updates retain their stored purpose.

> - [Server] Honour Environment purpose property and align SystemDatabaseModels to v1beta3 @SachinyadavAug20 (#23429)

### MeshSync stored metadata now replaces stale labels and annotations

- Affects: resources handled by MeshSync update and store-update paths

Previously, updates to existing resources—including resyncs—and store-update creation could leave labels or annotations that were no longer present in the resource metadata. Those paths now remove absent keys and keep stored metadata aligned with the latest resource.

> - [Server] Replace stale MeshSync metadata on updates @0xmaster7 (#23440)

### [Silent failure] Best-effort, idempotent `utils.ForceCleanupCluster` cleanup

- Affects: Kubernetes cleanup using `utils.ForceCleanupCluster`, including system uninstall

Previously, a list failure stopped cleanup before later resource kinds, deletion failures could be dropped while the command reported success, and already-absent resources made repeat cleanup fail. Cleanup now continues across resource kinds, reports genuine list and deletion failures, and treats already-absent resources as successfully removed.

> - [mesheryctl] Make ForceCleanupCluster best-effort and idempotent @AnimeshPriyanshu (#23601)

### Filter name selection now matches whole names

- Affects: filter name searches with `mesheryctl filter view`

When results included partial or multiple matches, the command could select the first result instead of resolving the supplied name as a whole-name match. It now prefers exact-case matches, then case-insensitive exact matches.

> - fix(mesheryctl): select filters by full name @0xmaster7 (#23580)

### Reset target fixed for `mesheryctl system reset`

- Affects: use of `--config` with `mesheryctl system reset`

Previously, the supplied file was left unchanged while the default meshconfig gained the named context and switched `current-context`; the reset also failed if no default meshconfig existed. Now the named context in the supplied file is reset to defaults while retaining its platform, and the default meshconfig is left untouched.

> - [mesheryctl] Reset the meshconfig given by --config in system reset @Eternity0207 (#23500)

### Performance-test settings preserved by `mesheryctl perf apply`

- Affects: performance tests configured with `mesheryctl perf apply` and an SMP configuration file or saved profile (Inferred)

Endpoint URLs are also validated when applying settings and creating profiles.

> - [mesheryctl] Consolidate performance test configuration @0xmaster7 (#23442)

### Context-specific Compose file version selection fixed

- Affects: mesheryctl `system start --context` or `system reset` contexts with a version different from the current meshconfig context, including stable contexts with an empty version

Previously, `system start --context X` could download the wrong release's Compose file when X was pinned to a different version from the current meshconfig context. Version validation and Compose-file selection now use the selected or passed context's version; invalid context names fail before downloading, and an empty version for a stable context resolves to `latest`.

> - [mesheryctl] Use the --context context's version in system start @Eternity0207 (#23461)

### Missing Meshery auth tokens saved to the active config

- Affects: Meshery CLI setups using `--config`

When the selected context named a token missing from the supplied file, the token was written to the default config instead, potentially sending later requests to the default server. Now the missing entry is written to the selected file, which remains in use for subsequent requests; without `--config`, the entry is written to the default config.

> - [mesheryctl] Add a missing token entry to the --config file, not the default @Asheesh7298 (#23422)

### [Silent failure] Trailing slashes trimmed from Meshery context endpoints

- Affects: `mesheryctl` contexts with endpoints ending in `/`

Requests from these contexts began with `//api/...`; redirects could resend `DELETE`, `POST`, or `PUT` as `GET` without a body. An `environment delete` command could report success even though nothing was deleted; existing configurations are now normalized when read, while `context create` still stores `--url` as typed.

> - [mesheryctl] Trim trailing slashes from the context endpoint @Asheesh7298 (#23391)

### Connection dialogs now fetch all definitions

- Affects: Meshery UI Connection Wizard or Connection Configure modal with more than 25 available connection definitions

Previously, the query fetched only the default 25 definitions, leaving definitions beyond that limit unavailable in the dialogs. Now the full set is available to populate connection options.

> - [UI] Fetch all connection definitions in Connection Wizard and Configure modals @antcybersec (#23709)

### Workspace Activity dropdown requests all organization workspaces

- Affects: Workspace Activity widgets in organizations with more than 10 workspaces

Previously, the unpaged request defaulted to 10 workspaces, leaving workspace 11 and later unavailable in the dropdown. The dropdown now includes all workspaces the viewer has permission to access.

> - [UI] Fetch all workspaces in Workspace Activity widget @antcybersec (#23676)

### Recursive relationship nesting fixed in `removeDuplicateVersions`

- Affects: UI registry use of `removeDuplicateVersions`

When it aggregates duplicate models with the same name and version, their relationships could become recursively nested during grouping.

> - [UI] Fix recursive relationship grouping and component array initialization in removeDuplicateVersions @madhavzanwar (#23672)

### UI registry relationship deduplication fixed

- Affects: setups using the UI registry helper to group relationships

Duplicate relationships could appear in results because they were not deduplicated by ID; relationships without a kind are now grouped under “Unknown.”

> - [UI] Fix recursive relationship grouping and component array initialization in removeDuplicateVersions @madhavzanwar (#23672)

### Meshery UI Kubernetes node readiness lookup fixed

- Affects: Meshery UI status reporting for Kubernetes nodes and resources with condition arrays

Previously, a node with `Ready=True` followed by another condition, such as `DiskPressure=False`, could appear not ready. It now reports `Ready` based on the `Ready` condition's status; a missing `Ready` condition or one whose status is not `True` still does not report ready.

> - [UI] Fix node readiness condition lookup @0xmaster7 (#23150)

### Filters table hides publication actions when `canPublishFilter` is false

- Affects: Filters table rows when `canPublishFilter` is false

Previously, an Unpublish button appeared on every row in that state, even when the filter was unpublished or the provider did not support publishing. When `canPublishFilter` is true, each row shows the action appropriate to its filter's publication state.

> - [UI] Hide Publish and Unpublish actions when canPublishFilter is false @antcybersec (#23651)

### Empty and non-string input handling fixed in `normalizeURI`

- Affects: Not stated in the release notes

Previously, empty or non-string input could throw or return `/`. It now returns an empty string for those inputs, while non-empty paths keep their existing normalization behavior.

> - [UI] Handle invalid input in normalizeURI @0xmaster7 (#23605)

### Deployment target picker now lists all connections

- Affects: deployment target picker in environments with more than 10 connections

Previously, the picker received only the 10 most recently updated connections, so older Kubernetes connections could be missing and the selected/total counter could be wrong. It now receives every connection, making all Kubernetes connections available for selection while non-Kubernetes connections remain outside the Kubernetes selection count.

> - [UI] List every connection of an environment in the deployment target picker @ChackoPAbraham19 (#23590)

### Organization user page-size forwarding fixed

- Affects: organization user lists with a selected workspace team-member page size

Requests previously sent the selected size as `pagesize` instead of the endpoint’s accepted `pageSize`, so the chosen size could be ignored without an error. The endpoint now applies the selected size; tested options include 10, 20, and 25 users per page.

> - fix(ui): forward organization user page size @0xmaster7 (#23576)

### Kubernetes memory quantity scaling fixed

- Affects: UI paths that normalize Kubernetes memory quantities with `Pi`, `P`, `Ei`, or `E` suffixes (Inferred)

Previously, `Pi` and `P` quantities could be formatted as `1 Bytes`, while `Ei` and `E` values were undercounted by factors of 1024 and 1000. They now use binary scales for `Pi` and `Ei`, and decimal scales for `P` and `E` (for example, `1Ei` formats as `1.15 EB`).

> - [UI] Correct Pi, P, Ei, and E memory quantity conversion @Veditha08 (#23505)

### Workspaces table camelCase field mapping fixed

- Affects: Workspaces table

With camelCase workspace rows, the table previously showed no owner avatar or organization name for ownerless workspaces, owner links pointed to `/user/undefined`, and some column visibility controls did not work. It now displays the owner and organization data, uses the owner field for profile links, and matches column visibility keys to the table’s column names.

> - [UI] Read camelCase owner fields in the Workspaces table @Asheesh7298 (#23453)

### View Info details refresh after visibility and notes changes

- Affects: workspace views viewed or updated in the View Info modal

After a visibility or notes update, the modal could keep showing the old visibility or stale notes, and the list behind it did not refresh after a visibility change. The modal now uses fetched notes when available and preserves edits across a refetch; visibility changes refresh the list without invalidating the paged views list.

> - [UI] Refresh cached views after visibility, notes and delete changes @Asheesh7298 (#23479)

### Deleted view cache entries are invalidated

- Affects: workspace views (Inferred)

A deleted view could still be returned from cache when requested again while its cached result remained, without reflecting the deletion.

> - [UI] Refresh cached views after visibility, notes and delete changes @Asheesh7298 (#23479)

### Structured metadata formatting preserves numeric and boolean values

- Affects: structured notification or data metadata with numeric or boolean values (Inferred)

The generic notification/data formatter previously treated numeric and boolean metadata as empty and discarded it, including `0` and `false` alongside non-empty strings. It now renders those values as strings and passes zero and false to custom property formatters.

> - [UI] Preserve numeric and boolean values in structured metadata @harshitethic (#23449)

### Credentials table date display fixed

- Affects: Credentials table rows with `createdAt` and `updatedAt` fields

The date columns used `created_at` and `updated_at` while the API returned `createdAt` and `updatedAt`, so every row showed the page-load time instead of the credential’s dates. The table now shows the actual creation and update dates in relative-time format, with the full date on hover.

> - [UI] Show real dates in the Credentials table @Asheesh7298 (#23459)

### [Silent failure] Meshery `schemaValidator` validation returns copied results

- Affects: Meshery UI component validation and selectors that read schema-validation results by reference

Previously, validation mutated the existing results object and returned the same reference, so reference-compared selectors could miss updates and leave results stale without an error. Selectors can now detect updates, and an existing failure-message string remains unchanged.

> - fix(ui): handle validation error taps in every design validator state @yi-nuo426 (#23409)

### Docker Extension token requests no longer hang without a websocket client

- Affects: Docker Extension authentication using `/token/store` or `DELETE /token`

Previously, browser login before the extension UI opened `/ws` left `/token/store` hanging, while logout after the websocket disconnected left `DELETE /token` hanging. Token storage and deletion now complete without a connected websocket, and connected clients still receive authentication-state notifications.

> - [Docker Extension] Don't hang token store and logout when no websocket client is connected @CyberRik (#23765)

### Docker Extension proxy concurrency and websocket close races fixed

- Affects: Docker Extension proxy setups (Inferred)

Concurrent access to `token` and `authWsChan` could race, and closing the channel during a send could cause a “send on closed channel” panic. Closing an older websocket no longer disrupts the newest client's authentication updates.

> - [Docker Extension] Don't hang token store and logout when no websocket client is connected @CyberRik (#23765)

### Meshery filter handler nil and missing-value failures fixed

- Affects: Meshery filter handlers

When these handlers encountered a missing user or provider, a nil `SystemID` during event creation, a non-string token in request context, or nil optional dependencies, they could panic or hit nil dereferences. Missing users now receive HTTP 401 on the specified mutating endpoints, missing providers receive HTTP 500 `ErrNilProvider`, invalid token values receive HTTP 500 `ErrFetchToken`, event creation uses `uuid.Nil` when `SystemID` is nil, and nil optional dependencies are guarded.

> - [Server] Guard nil user, nil provider, and nil SystemID in filter handlers @antcybersec (#23759)

### Missing-context panics fixed in Meshery pattern handlers

- Affects: Meshery pattern handlers

Handlers could panic when a user, provider, or `SystemID` was missing. The listed missing-user cases now return 401, missing-provider cases return 500, and event-producing operations use `uuid.Nil` when the system ID is missing; pattern-list requests without context tokens are also handled safely.

> - [Server] Guard nil user, nil provider, and nil SystemID across pattern handlers @antcybersec (#23755)

### Connection handler nil-pointer crashes fixed

- Affects: connection registration, saving, listing, lookup, update, and status-notification handlers

Connection operations could panic when a required user or provider was nil, or when event creation encountered a nil `SystemID`. They now return 401 for a missing user, 500 for a missing provider, or use `uuid.Nil` when system identification is absent.

> - [Server] Guard nil user, nil provider, and nil SystemID in connections handlers @antcybersec (#23752)

### K8sconfig handler nil-pointer failures fixed

- Affects: K8sconfig import, context discovery, Kubernetes ping, and registration handlers

Missing users, providers, or system IDs, and a nil logger, could trigger nil-pointer panics in these handlers. Missing users now receive HTTP 401, missing providers receive HTTP 500 with `ErrNilProvider()`, missing system IDs default to `uuid.Nil` in the specified paths, and logger use is guarded.

> - [Server] Guard nil user, nil provider, and nil SystemID in k8sconfig handlers @antcybersec (#23750)

### Credential handler nil-value failures fixed

- Affects: credential save, update, list, lookup, and delete handlers

Missing users or providers, or unavailable logging, could cause nil dereferences or unexpected failures. Missing users now receive HTTP 401 in `SaveUserCredential`, `UpdateUserCredential`, and `GetUserCredentials`; missing providers return HTTP 500 with `ErrNilProvider()` across the five credential operations, and unavailable logging no longer causes failures.

> - [Server] Guard nil user and nil provider in credentials handlers @antcybersec (#23748)

### Filter handler UUID validation and error responses fixed

- Affects: filter retrieval, file-download, clone, and delete endpoints

Malformed filter path IDs now receive bad-request responses. Missing filters requested for retrieval or cloning return not-found responses; previously, retrieval could return an uninitialized zero-struct alongside a raw error. Provider lookup errors now use `providerStatusOrInternal(err)` for status mapping.

> - [Server] Return 404 for missing filters and validate path UUIDs in filter handlers @antcybersec (#23726)

### Relationship selector model-name mismatches fixed

- Affects: relationship selectors in `models/`

The remaining misspelled or unregistered names caused `TestRelationshipSelectorsNameRegisteredModels` in `server/policies` to fail; for example, `azure-conatiner-service` appeared instead of `azure-container-service`.

> - [Models] Fix remaining unrenamed relationship selector model names @antcybersec (#23720)

### Local provider design lookup responses corrected

- Affects: the Local provider’s `GET /api/pattern/{id}` route

When a requested design does not exist, the route previously returned 400 with an empty design-shaped body; it now returns 404 with a MeshKit error body. Malformed IDs still return 400, but now return a `meshery-server-1432` invalid-UUID error body instead of an empty design.

> - [Server] Return 404 for a design that does not exist @himanishpuri (#23688)

### Workspace API missing-workspace responses corrected for GET and PUT

- Affects: workspace API requests

A request to `GET` or `PUT /api/workspaces/{id}` for a well-formed workspace ID that did not exist returned `502 Bad Gateway`. Such requests now return `404`; other workspace retrieval errors remain unchanged.

> - [Server] Return 404 for a workspace that does not exist @himanishpuri (#23686)

### Detached policy event persistence panics no longer crash the server

- Affects: policy evaluation using `EvaluateRelationshipPolicy`

Before this fix, a panic in the detached `provider.PersistEvent` goroutine during policy evaluation could crash the server process. Such panics are now recovered and logged when a logger is available.

> - [Server] Guard unchecked nil dereferences, unbounded body reads, and detached goroutine panics in EvaluateRelationshipPolicy @Amansingh0807 (#23664)

### Meshery SSE streams end on shutdown

- Affects: the providers stream, adapter events, events subscribe, and controllers status subscribe SSE routes

Previously, active streams could hold shutdown for the full 30-second `httpShutdownTimeout`, leaving handlers running while the provider session and database closed. Shutdown now cancels their request contexts so streams end; if draining times out, it closes remaining connections, though it does not wait for those handlers to return.

> - [Server] End SSE streams when the server starts shutting down @SandstromPL (#23655)

### `WriteCapabilitiesForUser` nil-data panic fixed

- Affects: calls to `UserCapabilitiesPersister.WriteCapabilitiesForUser` with nil capability data

Previously, nil data caused a nil-pointer panic. These calls now return `ErrNilConfigData` instead.

> - [Server] Guard nil capabilities payload and return ErrResultNotFound in UserCapabilitiesPersister @Yasas-Sri (#23644)

### Schema resource lookup returns HTTP 404 for names without a schema

- Affects: calls to `GET /api/schema/resource/{resourceName}`

Requests with an empty, unknown, or otherwise unrecognized resource name returned HTTP 500. They now return HTTP 404, while recognized schemas continue to be returned as JSON.

> - [server] return 404 for unknown resource schemas @rakesh0x (#23635)

### Nil-provider checks added to schedule handlers

- Affects: schedule handlers

When a request reaches one of these handlers with `provider == nil`, it receives `ErrNilProvider()` with HTTP 500 instead of dereferencing the provider.

> - [Server] Guard nil provider and validate schedule ID across schedule handlers @piyxsh31 (#23540)

### AWS and Azure cloud-resource relationship model names corrected

- Affects: AWS and Azure cloud-resource relationship definitions (Inferred)

Some definitions used incorrect model names, including `azure-conatiner-service` and `aws-vpccni-controller`. The corrected mappings associate cloud resources with their appropriate models.

> - fix(server): retain cluster context snapshot for cleanup @0xmaster7 (#23579)

### Seeded relationship selector names corrected

- Affects: relationships using affected selectors in seeded versions of the named models (Inferred)

Some selectors matched only by substring, while others matched no model, so affected relationships could match incorrectly or never apply. One duplicate selector was also removed.

> - [Models] Fix relationship selectors that name models that do not exist @BarbieRajput (#23566)

### Nil-provider dereferences fixed in user handlers

- Affects: requests handled by `GetUserByIDHandler`, `GetUsers`, `UserPrefsHandler`, `ShareDesignHandler`, or `ShareFilterHandler`

When the provider was nil, these requests could trigger a nil-pointer dereference. They now receive an HTTP 500 error instead.

> - [Server] Guard nil provider and nil user dereferences across user handlers @piyxsh31 (#23543)

### Missing-user handling fixed in `UserPrefsHandler` POST

- Affects: POST requests handled by `UserPrefsHandler`

When a POST arrives without user context, it returns HTTP 400.

> - [Server] Guard nil provider and nil user dereferences across user handlers @piyxsh31 (#23543)

### Malformed performance profile IDs rejected before provider access

- Affects: the `GetPerformanceProfileHandler` and `DeletePerformanceProfileHandler` endpoints

Previously, a malformed, non-UUID ID could reach provider internals and produce a generic HTTP 500. These requests are now rejected before provider or GORM access with `ErrInvalidUUIDValueCode`.

> - [Server] Guard invalid UUID status codes, missing 404 responses, and unchecked token extraction in performance profile handlers @Amansingh0807 (#23561)

### HTTP 404 handling for missing profiles in `GetPerformanceProfileHandler`

- Affects: callers of `GetPerformanceProfileHandler`

When the requested profile was absent from a local GORM persister or remote provider, it was reported as an HTTP 500 server failure. It now returns HTTP 404 with a structured MeshKit error.

> - [Server] Guard invalid UUID status codes, missing 404 responses, and unchecked token extraction in performance profile handlers @Amansingh0807 (#23561)

### Performance profile handlers preserve provider HTTP statuses

- Affects: `SavePerformanceProfileHandler`, `GetPerformanceProfilesHandler`, `DeletePerformanceProfileHandler`, and `GetPerformanceProfileHandler`

When a provider error carried an upstream HTTP status, these handlers could return a generic HTTP 500 instead. They now return the provider status; errors without one still use the internal-error fallback.

> - [Server] Guard invalid UUID status codes, missing 404 responses, and unchecked token extraction in performance profile handlers @Amansingh0807 (#23561)

### Performance profile handler token handling fixed

- Affects: performance profile requests using local or remote providers

An absent context token could panic. With a remote provider, a missing or empty token now returns HTTP 401 (`ErrFetchToken`); the local provider accepts an empty token.

> - [Server] Guard invalid UUID status codes, missing 404 responses, and unchecked token extraction in performance profile handlers @Amansingh0807 (#23561)

### Adapter deletion protected from session-sync restoration

- Affects: adapter management with session synchronization

When an adapter was deleted through `DELETE /api/system/adapter/manage?adapter=<loc>`, it was removed only from the local preference slice and could be restored on the next session sync. Deletion now also removes it from `h.config.AdapterTracker`, preventing that restoration.

> - [Server] Remove adapter from AdapterTracker on delete and prevent session sync resurrection @nishant-devv (#23555)

### `MeshOpsHandler` kubeconfig failure response fixed

- Affects: Kubernetes context kubeconfig generation

When kubeconfig generation for a Kubernetes context failed, the client connection hung until a network timeout without an HTTP status or error message. The handler now logs `ErrInvalidKubeConfig` and returns HTTP 400 with a structured error response.

> - [Server] Remove adapter from AdapterTracker on delete and prevent session sync resurrection @nishant-devv (#23555)

### Performance profile deletion preserves provider status

- Affects: performance profile deletion

When deletion failed and the remote provider returned an HTTP status such as 403 or 409, the handler returned 500 instead. It now preserves that status; failures without a provider status still return 500, and not-found handling is unchanged.

> - [Server] Preserve the provider status when deleting a performance profile fails @Shikhyy (#23549)

### Blank load-test name handling fixed

- Affects: load-test requests with a blank name, supplied as a query parameter or in SMP configuration

Previously, either input path caused a nil-pointer panic. These requests now return HTTP 400 with the existing MeshKit blank-name error.

> - [Server] Fix blank load-test name panic @0xmaster7 (#23502)

### Design package export responses fixed

- Affects: design package exports using `?pkg=true`

Previously, uncompressed YAML followed the tar stream, design YAML could be serialized as an escaped scalar, and the MIME type and filename headers did not match the archive. Now the response contains correctly formatted design content and uses a tar content type and `.tar` attachment filename.

> - [Server] Prevent stream corruption and double-serialization in design package export @nishant-devv (#23497)

### Fetch-results token and form handling fixed

- Affects: the five fetch-results endpoints

When a request had a missing or wrong-type token in its context, it could panic; an absent fetch token could also be sent to the provider. Malformed form data returned HTTP 403; now requests without a usable fetch token receive an error response, and malformed forms receive HTTP 400.

> - [Server] Guard unchecked context type assertions, nil provider panics, and 403 form parsing errors in fetch results handlers @Amansingh0807 (#23435)

### Fetch-results handlers now handle provider and result failures safely

- Affects: fetch-results endpoints

Previously, a nil provider could be dereferenced, missing or empty results were not checked in `GetResultHandler` and `FetchSingleSmiResultHandler`, and lookup or conversion failures could yield invalid result data. Now, nil providers return HTTP 500, missing results return HTTP 404 in those handlers, and lookup or conversion failures do not return invalid data.

> - [Server] Guard unchecked context type assertions, nil provider panics, and 403 form parsing errors in fetch results handlers @Amansingh0807 (#23435)

### Background Discovery preserves newer connection statuses

- Affects: connections undergoing background Discovery

If a user disconnected a connection while Discovery or its `NOTFOUND` probe was in flight, it could restore `CONNECTED` and persist a stale resolution. Now the connection remains `DISCONNECTED`, and the stale status write, resolution, and event are refused.

> - [Server] Persist Discovery resolutions only when the stored status still matches @dropps07 (#23294)

### Results and performance-profile handler panics fixed

- Affects: Results and performance-profile endpoints

Requests without token context or with a non-string token could panic, and a nil provider could cause a crash. The handlers now avoid token-extraction panics and return HTTP 500 when the provider is unavailable.

> - [Server] Guard nil provider and unsafe token extraction in results and performance handlers @piyxsh31 (#23420)

### Context retrieval and deletion now reject invalid UUIDs

- Affects: `GetContext` and `DeleteContext` requests

Previously, these handlers did not validate IDs at the handler boundary before provider lookup or state-machine initialization. Malformed strings and nil UUIDs now return HTTP 400 with `models.ErrInvalidUUID` before either step.

> - [Server] Validate context ID and guard nil pointer dereferences in contexts handler @piyxsh31 (#23424)

### Context handler nil-pointer crashes fixed

- Affects: context retrieval and deletion handlers

A nil provider in `GetAllContexts`, `GetContext`, or `DeleteContext` could cause a nil-pointer crash; those requests now return HTTP 500 instead. In `DeleteContext`, missing user or system identifiers, or a nil state-machine tracker, could cause a server panic; these cases no longer cause dereference panics.

> - [Server] Validate context ID and guard nil pointer dereferences in contexts handler @piyxsh31 (#23424)

### `mesheryctl` provider response body leak fixed

- Affects: provider information retrieval with `mesheryctl`

When `GetProviderInfo` passed the successful-response check, its response body was left open whether decoding succeeded or failed. The body is now closed after use.

> - [mesheryctl] Fix HTTP response body and file descriptor leaks in utils @muskan-creates352 (#22645)

### `InstallprereqDocker` closes the docker-compose response body

- Affects: setups using `mesheryctl` Docker prerequisite checks that fetch docker-compose release information

When the check fetched docker-compose release information, the response body was left open after being read. The existing fallback behavior when a version check fails remains unchanged.

> - [mesheryctl] Fix HTTP response body and file descriptor leaks in utils @muskan-creates352 (#22645)

### Nil `TestStartTime` panic fixed in `mesheryctl perf result`

- Affects: performance results with an unpopulated `TestStartTime`

When table conversion or expanded `--view` output handled a result with a nil `TestStartTime`, it could panic while dereferencing the start time. Both now display `None` instead.

> - [mesheryctl] Guard nil TestStartTime, fix dead check, and standardize timestamp in perf result (#23727) @muskan-creates352 (#23729)

### `mesheryctl` Kubernetes log follow output preserves complete lines

- Affects: Kubernetes logs followed with `mesheryctl`

When a log line is split across stream reads, it now appears as a complete line. Content remaining without a trailing newline at EOF is printed with the pod name and a newline; non-follow Kubernetes logs and Docker logs are unchanged.

> - [mesheryctl] Preserve followed Kubernetes log lines @0xmaster7 (#23705)

### Local design import filename handling fixed

- Affects: local design imports on Windows without an explicit import name

Previously, `path.Base` did not recognize Windows path separators, so imports could use the wrong filename and default design name. The import now uses `filepath.Base` to derive both correctly.

> - [mesheryctl] Fix design import filename on Windows @0xmaster7 (#23665)

### EventBroadcaster subscriber-entry leak fixed

- Affects: Meshery server setups using EventBroadcaster subscriptions (Inferred)

When a user's last listener unsubscribed, `Broadcast.clients` kept the subscriber entry indefinitely. Empty entries are now removed, and new subscriptions after eviction—including concurrent subscription changes—are handled without disrupting event notifications.

> - [Server] Prune empty subscriber entries from EventBroadcaster and [mesheryctl] fix portforward errOut ineffassign @Atharva-Mendhulkar (#23516)

### mesheryctl token path resolution fixed

- Affects: meshconfig token `location` set to an absolute or `~/` path

Authenticated commands failed with `Authentication token not found` because these paths were incorrectly joined under `~/.meshery`. They now resolve to the same locations used by login and logout; bare filenames still resolve under `~/.meshery`.

> - [mesheryctl] Resolve token locations for requests the way login does @Asheesh7298 (#23475)

### Docker endpoint check fixed for system start

- Affects: Docker starts using `mesheryctl system start -p docker` with localhost or a remote context endpoint

The check could dial an invalid address and report the endpoint inaccessible after all 30 checks, even when Meshery was running. Now the check can succeed when the selected endpoint’s port responds.

> - [mesheryctl] Wait on a dialable endpoint in system start on Docker @Eternity0207 (#23430)

### Duplicate model aggregation fixed in `removeDuplicateVersions`

- Affects: the UI registry helper `removeDuplicateVersions`

When it combined duplicate models with the same name and version, `allComponents` and `allRelationships` were initialized as `{}` rather than arrays. They are now arrays, so the models’ components and relationships are combined.

> - [UI] Fix recursive relationship grouping and component array initialization in removeDuplicateVersions @madhavzanwar (#23672)

### Filters table timestamp display fixed

- Affects: filter rows in the Filters table

When the table's snake_case column names did not match the camelCase fields returned by `GET /api/filter`, each row showed the current moment instead of its timestamps. Rows with undefined cell values now show their actual `createdAt` and `updatedAt` timestamps.

> - [UI] Show real dates in the Filters table @antcybersec (#23649)

### Broker version and service settings configurable in the `meshery-broker` chart

- Affects: Broker version or service fields configured through the `meshery-operator` Helm chart

Before this change, the chart did not expose Broker version or service exposure settings. Unset values are omitted, so the default Broker rendering remains byte-for-byte unchanged.

> - [Helm] Give MeshSync read-only RBAC, expose Broker service values, clean up CRs on uninstall @yi-nuo426 (#23259)
> - [Helm] Give MeshSync read-only RBAC, expose Broker service values, clean up CRs on uninstall @yi-nuo426 (#23259)

## Enhancements (6)

- Unpaginated Local-provider organization listing (To use: `pagesize=all` on an organization-list request)
- Non-binding label-matched relationships for Chaos Mesh experiments
- Chaos Mesh experiment-to-`Schedule` relationships added
- Remote-cluster reference relationships for Chaos Mesh experiments
- End-to-end Meshery uninstall command
- Accessible names added to dashboard selectors

## Other changes (71)

- REST API layouts show Required/Optional indicators without a content map
- REST API references render missing request-body `content` maps as empty
- Expanded tests for extension install and remove handlers
- Faster binding relationship identification
- Structured response assertions for `EvaluateRelationshipPolicy` tests
- `EvaluateRelationshipPolicy` tests now check response JSON structure and envelope contents
- Test for filter save handling of undecodable provider responses
- Successful filter saves now generate informational events
- Malformed organization IDs rejected by user key lookup
- Missing capability deletion now returns `ErrResultNotFound`
- Meshery server HTTP timeouts updated
- Environment missing-record checks consolidated in `isNotFoundError`
- Local provider pagination errors now return HTTP 400
- Adapter status logs no longer include a trailing newline
- Strict UUID validation for result-handler IDs
- `mesheryctl system context view` lowercases context flag values
- `mesheryctl perf result` empty-result check fixed
- Canonical timestamp formatting in expanded perf results
- Nil `Owner` and `MesheryID` UUIDs display as `None`
- System flag cleanup added to `TestResetCmd`
- Documentation of platform-specific token file ordering
- Argument validation for `mesheryctl environment delete` clarified
- Misspellings fixed in 16 Meshery server and `mesheryctl` messages
- Unexpected positional arguments rejected by `mesheryctl connection create`
- BATS test added for the installer's answered platform prompt
- mesheryctl port-forward `errOut` initialization fix
- BATS test added for the installer ADAPTERS test plan
- Unused direct dependencies removed from provider-ui
- Design configurator accordions no longer auto-scroll on mount
- Connection modals switch to the connection definitions query hook
- `NotificationProps.type` narrowed to `'success' | 'error'`
- MeshSync notification text now uses the secondary theme color
- MeshSync notification backgrounds use theme-aware success and error colors
- MeshSync Retry button adopts Sistent contained error styling
- Registry helper adds nullish-input guards
- UI audit script switches to recursive file scanning
- Performance fetch error notifications show readable messages
- Hex color audit reports filesystem errors and incomplete scans
- Hex color audit skips comments when detecting colors
- Hex color audit targets nested production directories
- More precise `rgb()` and `rgba()` matching in the hex color audit
- Hex color audit skips test files and nested `__tests__` directories
- Registry component details requests scoped by model, kind, and API version
- Wider fields and tighter spacing in Kanvas pattern forms
- MUI import audit uses TypeScript AST parsing
- Server unit-test CI streams output and uploads artifacts
- Installer platform-prompt test added
- BATS test added for installer `ADAPTERS`
- Meeting-minutes changes excluded from docs preview builds
- Contributing guide updated with sidebar navigation guidance
- Kanvas Designer validator handles `TAP_ON_ERROR` in every state
- Bitcask removed from Meshery database documentation
- Broken links in contributing guides fixed
- Documentation breadcrumb accessibility improvements
- Documentation breadcrumb accessibility improvements
- Reference documentation typo fixes
- Daniel Adeogun's community profile added to the meetings documentation
- Docs site lightbox initialization fix
- Docs lightbox initialization ReferenceError fixed
- Meshery provider documentation corrected for URLs, package location, preferences route, and registration example
- Four broken documentation links fixed
- Elisei Zaitsev meeting notes moved to `week-of-11-12-2026`
- Newcomer's Meet heading added to the November 12, 2026 meeting notes
- `.content-table-wrapper` hidden on small screens
- `:focus-visible` outlines added to page-metadata and table-of-contents links
- Code-block wrapper changed from `<div>` to `<span>`
- Visible `:focus-visible` outlines for code-block copy buttons
- Light-mode pagination label color corrected
- Stacks cards arranged in one column on narrow screens
- Pagination links gain accessibility markup
- Visible focus outlines on pagination cards

---

Summaries are produced by AI, so they can be wrong, and they are provided without warranty.
