Compare commits

...
Author SHA1 Message Date
KilianandCopilot 22a3864b1e Unify workflow ConfigMap configuration for Hub 0.154.0 (#184)
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
2026-10-06 11:39:21 +02:00
KilianandCopilot 38806cae0e Add additive workflow ConfigMaps for Hub 0.153.0 (#181)
* Add additive workflow ConfigMaps and bump Hub chart to 0.150.0

Preserve bundled Start metadata and shared classifications when adding custom contracts/catalogs. Reject conflicting mount paths and retain the 0.149.0 replacement option for compatibility.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* Bump Hub chart to 0.151.0

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* Remove workflow-specific CI checks and scripts

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* Bump Hub chart to 0.153.0

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* Support externally managed classification catalogs

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

---------

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
2026-10-06 09:19:01 +02:00
Cédric Verstraeten b353678d5a Merge pull request #183 from kerberos-io/feature/mongodb-app-name-per-service
feature/mongodb-app-name-per-service
2026-10-05 21:43:07 +02:00
Cédric VerstraetenandCopilot ec62edd5af Set a per-service MongoDB appName in the hub chart
Every hub workload shares mongodb.uri through the mongodb-config ConfigMap,
so all of them connect with the same appName and Atlas cannot attribute
queries to a service. Each workload now gets an explicit MONGODB_URI env
entry, rendered from mongodb.uri with appName set to the service name; an
explicit env entry takes precedence over envFrom, so no application change
is needed.

The behaviour is controlled by mongodb.appNamePerService (default true) and
renders nothing when mongodb.uri is empty. Chart version 0.152.0.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
2026-10-05 19:30:36 +00:00
Cédric Verstraeten 26957e239f Merge pull request #182 from kerberos-io/feature/add-disable-heartbeat-media-proxy
feature/add-disable-heartbeat-media-proxy
2026-10-05 17:16:59 +02:00
Cédric Verstraeten 03210c53c0 Add Hub media proxy and feature controls
Expose media proxy settings, a device polling disable switch, and an analysis filter toggle for both frontends. Document the options and bump the chart to 0.150.0.
2026-10-05 17:12:59 +02:00
KilianandCopilot 2111c0f9d6 Add workflow condition contracts and shared catalog packaging (#180)
* Add workflow condition contracts and shared catalog packaging

Define versioned contract and catalog schemas, Start and custom-stage examples, and opt-in bundled/external ConfigMap packaging. Extract classification defaults without changing the rendered default manifests or existing overrides.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* Bump Hub chart to 0.149.0

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* Omit optional classification catalog override by default

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

---------

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
2026-10-05 16:40:48 +02:00
KilianandCopilot eef273bf3c Support explicit IDs for deployment-defined workflows (#179)
Render and validate optional stable workflow IDs, retain legacy name-derived identities, pin documented examples, and exercise identity consistency across API and engine manifests.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
2026-09-29 16:15:02 +02:00
Cédric Verstraeten dc6c79df2e Merge pull request #178 from kerberos-io/feature/extend-workflow-designer
feature/extend-workflow-designer
2026-09-28 17:17:44 +02:00
35 changed files with 931 additions and 189 deletions
@@ -1,31 +0,0 @@
name: Workflows queue consistency
# Fails the build if the analysis producer, the workflows engine and the stage
# workers would render onto different WORKFLOWS_QUEUE names — the silent
# producer/consumer queue-name drift that leaves runs piling up with no
# consumer. Pure `helm template` render check, no cluster required.
on:
workflow_dispatch:
pull_request:
paths:
- 'charts/hub/**'
- 'scripts/check-workflows-queue-consistency.sh'
- '.github/workflows/workflows-queue-consistency.yaml'
permissions:
contents: read
jobs:
check:
runs-on: ubuntu-24.04
steps:
- name: Checkout repository
uses: actions/checkout@v4
- name: Set up Helm
uses: azure/setup-helm@v4
with:
version: v3.16.2
- name: Check WORKFLOWS_QUEUE consistency
run: ./scripts/check-workflows-queue-consistency.sh charts/hub
+1 -1
View File
@@ -16,7 +16,7 @@ type: application
# This is the chart version. This version number should be incremented each time you make changes
# to the chart and its templates, including the app version.
# Versions are expected to follow Semantic Versioning (https://semver.org/)
version: 0.148.0
version: 0.154.0
# This is the version number of the application being deployed. This version number should be
# incremented each time you make changes to the application. Versions are not expected to
+285 -2
View File
@@ -91,6 +91,60 @@ Waves are not a substitute for message compatibility, readiness checks, or
stopping old RabbitMQ consumers. This example does not add health probes or
change the Deployment rolling-update strategy.
### Deployment-global workflow identity
`kerberoshub.workflows.definitions` is a map keyed by workflow name. Enabled
definitions are passed as the same `WORKFLOW_DEFINITIONS` JSON to Hub API and the
workflow engine. These deployment-global, ops-managed workflows are read-only
through the API; manage them in Helm values. There are no per-project on/off
overrides.
Supply an explicit, stable `id` for new definitions:
```yaml
kerberoshub:
workflows:
enabled: true
definitions:
tracking-workflow:
id: "ec7e72bc4cedd32018e1f92a"
enabled: true
stages:
- operation: objecttracking
```
The worker and its queue must also be configured under
`kerberoshub.services.objecttracking`. As before, omitted triggers default to a
single automatic trigger and disabled definitions are not sent to either service.
Identity is separate from mutable names and content versions. Keep an explicit
ID unchanged when renaming or revising a workflow; changing it creates an
independent identity, not a new version of the existing workflow. The chart
does not generate random IDs or introduce a version field.
For compatibility, an omitted `id` remains omitted from the rendered JSON. The
backend derives the effective ID as the first 24 hexadecimal characters of
SHA-256 over the **exact map key**, including case and whitespace. Before
renaming an existing ID-less definition, pin its current derived ID explicitly
to preserve references and history. For example:
```sh
printf '%s' 'tracking-workflow' | sha256sum | cut -c1-24
# ec7e72bc4cedd32018e1f92a
```
The chart ships an empty definitions map. Its commented `tracking-workflow` and
`vlm-workflow` examples pin their current name-derived IDs for compatibility.
Explicit IDs must be quoted, non-zero, 24-character hexadecimal strings (Mongo
ObjectIDs); uppercase is accepted and normalized to lowercase in JSON. Empty,
null, all-zero, malformed, and non-string IDs fail Helm rendering with the
definition name and validation requirement. Duplicate effective IDs also fail,
naming both definitions and the colliding ID, whether explicit/explicit or
explicit/name-derived. Validation includes disabled definitions and runs even
when the engine is disabled, so identities remain reserved. Omit the `id` key
entirely to use the legacy fallback; invalid explicit IDs never fall back.
### Parameters
Below all configuration options and parameters are listed.
@@ -131,6 +185,7 @@ kerberospipeline:
| `mongodb.authenticationMechanism` | MongoDB authentication mechanism (for example `SCRAM-SHA-256`). | `"SCRAM-SHA-256"` |
| `mongodb.username` | MongoDB user account, we are using in the hub installation `'root'`. | `"yourusername"` |
| `mongodb.password` | MongoDB user password, by default `'yourmongodbpassword'` | `"yourpassword"` |
| `mongodb.appNamePerService` | When `mongodb.uri` is set, override its `appName` per workload with the service name (for example `hub-api`, `hub-pipeline-monitor`), so the MongoDB Atlas query profiler shows which service issued each operation. | `true` |
| `mongodb.retryWrites` | Enable or disable MongoDB retryable writes. | `"true"` |
| `mongodb.flavor` | Backend engine flavor: `"mongodb"` (native MongoDB / Atlas) or `"documentdb"` (AWS DocumentDB). The `documentdb` flavor disables features DocumentDB does not support (geospatial queries/indexes, complex `$lookup` pipelines). When set to `documentdb`, also set `mongodb.retryWrites: "false"`. | `"mongodb"` |
| `mongodb.tls.enabled` | Enable TLS for MongoDB connections. When `mongodb.uri` is set, the chart appends missing `tls=true` and `tlsCAFile` query parameters. | `false` |
@@ -212,6 +267,11 @@ kerberospipeline:
| `kerberoshub.api.jwtSecret` | A secret that is for generating JWT tokens. | `"this-is-a-secret-please-change-to-random-string"` |
| `kerberoshub.api.schema` | The protocol to serve the Kerberos Hub API, `'http'` or `'https'`. | `"https"` |
| `kerberoshub.api.url` | The Kerberos Hub API ingress to access the API. | `"api.yourdomain.com"` |
| `kerberoshub.api.disableAgentHeartbeat` | Temporarily stop serving the polled `GET /devices` list (hub-api returns `204` without database reads). Devices still load once through `/session/bootstrap`; online/offline status stops refreshing. | `false` |
| `kerberoshub.api.mediaProxy.enabled` | Serve media through hub-api (`/media/proxy`) instead of handing out Vault/storage URLs, hiding the storage endpoint from clients. | `false` |
| `kerberoshub.api.mediaProxy.publicUrl` | Public hub-api base URL used in proxied media URLs. Defaults to `{schema}://{url}`. | `""` |
| `kerberoshub.api.mediaProxy.key` | Key used to encrypt proxied media URLs (`openssl rand -hex 32`). Defaults to a key derived from `jwtSecret`. | `""` |
| `kerberoshub.api.mediaProxy.ttl` | Lifetime of proxied media URLs (Go duration, max `168h`). | `"24h"` |
| `kerberoshub.api.resources.requests.memory` | Memory request for `kerberoshub.api`. | `"100Mi"` |
| `kerberoshub.api.resources.requests.cpu` | CPU request for `kerberoshub.api`. | `"250m"` |
| `kerberoshub.api.resources.limits.memory` | Memory limit for `kerberoshub.api`. | `"100Mi"` |
@@ -337,6 +397,7 @@ kerberospipeline:
| `kerberoshub.frontend.features.media.filter.devices.enabled` | Enable or disable `kerberoshub.frontend.features.media.filter.devices`. | `"true"` |
| `kerberoshub.frontend.features.media.filter.objectDetection.enabled` | Enable or disable `kerberoshub.frontend.features.media.filter.objectDetection`. | `"true"` |
| `kerberoshub.frontend.features.media.filter.star.enabled` | Enable or disable `kerberoshub.frontend.features.media.filter.star`. | `"true"` |
| `kerberoshub.frontend.features.media.filter.analysis.enabled` | Enable or disable `kerberoshub.frontend.features.media.filter.analysis`. | `"true"` |
| `kerberoshub.frontend.features.media.filter.region.enabled` | Enable or disable `kerberoshub.frontend.features.media.filter.region`. | `"true"` |
| `kerberoshub.frontend.features.media.filter.sort.enabled` | Enable or disable `kerberoshub.frontend.features.media.filter.sort`. | `"true"` |
| `kerberoshub.frontend.features.media.filter.category.enabled` | Enable or disable `kerberoshub.frontend.features.media.filter.category`. | `"true"` |
@@ -599,8 +660,10 @@ As mentioned during the Post installation step, you'll import some `.nosql` file
Within the Kerberos Hub front-end you'll see the option to filter through classifications. This filtered is stored in the `settings` collection. By changing the entries of the `classifications` object, you can add, edit or remove items from the filters.
New deployments should define the shared classification list through
`classificationCatalog`. Each entry contains the stable classifier output
The default shared classification list lives in
[`configuration/catalogs/classifications.json`](configuration/catalogs/classifications.json),
independently of workflow contracts. Set `classificationCatalog` to a list to
replace those defaults; unset or `null` uses the bundled file. Each entry contains the stable classifier output
`key`, the user-facing `label`, and an `icon` key. Hub API exposes this catalog
to alerts and filters. When the mounted catalog is unavailable, compatible Hub
API versions fall back to the legacy `settings` document and then the built-in
@@ -615,6 +678,226 @@ classificationCatalog:
icon: vehicle
```
To manage the catalog outside Helm, create a ConfigMap in the release namespace
with a `classifications.json` key containing the same JSON array, then set:
```yaml
classificationCatalogEnabled: true
classificationCatalogExistingConfigMap: hub-classification-catalog
```
This takes precedence over `classificationCatalog` and bundled defaults. Helm
does not create or checksum the external ConfigMap. The API mount and bundled
Start metadata both reference it; in bundled mode, do not add it again to
`workflowConfiguration.configMaps`. With `includeDefaults: false`, map any required
catalog explicitly through `configMaps`; the API's catalog selection remains
independent.
Disabling `classificationCatalogEnabled` disables the API catalog mount regardless
of this setting. The external ConfigMap and key must exist before the API pod can
start. Manage updates and any required rollout/reload through deployment tooling;
Helm cannot inspect or validate its contents. The catalog schema remains available
for authoring validation.
### Workflow condition contracts (preparatory)
This chart defines the initial v1 metadata format and opt-in file packaging.
**Hub API and the frontend do not yet consume these contracts.** Enabling the
packaging only mounts files; it neither enables an editor feature nor registers
workers or changes execution. No new loader environment variable is set.
The earlier device-bound workflow configurations were development fixtures, not
a recommended deployment model. Charts describe capabilities, not selected
devices, project IDs, concrete workflow edges, or user-authored conditions.
Device filtering is optional. Saved selections belong in individual workflows.
The configuration is split into independent shared catalogs and their consumers:
```text
configuration/
catalogs/
classifications.json
workflow-contracts/
start.yaml
schemas/
catalog.schema.json
workflow-contract.schema.json
```
- [Start contract](configuration/workflow-contracts/start.yaml): current device,
site, group and initial-classification presets, plus schedule/condition controls.
- [Custom pose example](examples/workflow-contracts/pose.yaml): hypothetical
worker output, demonstrating a file catalog, inline choices and a numeric field.
This example does not deploy a worker or claim a universal pose-model format.
- [Contract schema](schemas/workflow-contract.schema.json): format, source-name
documentation and YAML editor completion via the `$schema` comment.
- [Catalog schema](schemas/catalog.schema.json): shared JSON lists of
`{key, label, icon?}`, compatible with the current classification file format.
Each contract defines a `stage` type, a `schemaVersion`, a `contractVersion` and
its own `fields`. A field declares `id`, `label`, `path`, scalar `type` and
supported `operators`; `optional`, numeric bounds and `choices` are optional.
The `type` describes each wildcard-selected candidate, not the containing array.
IDs must be unique within a contract. Increment `contractVersion` when changing
its meaning/output shape; the future runtime must bind workflows to compatible
versions rather than silently reinterpreting saved conditions.
#### Choice origins
Use exactly one of these, or omit `choices` for typed user input:
```yaml
choices:
source: project.devices
```
```yaml
choices:
file: ../catalogs/classifications.json
```
```yaml
choices:
values: [standing, sitting, lying]
```
The supported system-resource vocabulary for v1 is:
| `choices.source` | User-facing labels | Stored comparison values |
| --- | --- | --- |
| `project.devices` | Authorized device names | Stable device keys |
| `project.sites` | Authorized site names | Stable site IDs |
| `project.groups` | Authorized group names | Stable group IDs |
These are direct built-in API resolver identifiers, not database paths or
deployer-selected authorization scopes. They still require implementation in the
future loader. Resolvers must reuse existing project/resource authorization and
legacy-owner compatibility; metadata cannot grant access.
File references resolve relative to the mounted contract, within the shared
configuration root. V1 uses `../catalogs/<name>.json`; no URLs, arbitrary file
access or recursive catalog references. Catalog keys are stable strings, with
labels for display; inline lists may contain strings, numbers or booleans matching
the field type. Catalogs are reusable by filters, alerts and other consumers,
not owned by workflows. A library may ship a catalog without implementing an API.
#### Bundled and external packaging
`workflowConfiguration.enabled` defaults to `false`. With it enabled and
`includeDefaults` omitted or `true`, the chart mounts the bundled Start contract and the
**effective** classification list under `/etc/kerberos/configuration`, using
the directory layout above. The projected volume reuses the existing
classification ConfigMap, including `classificationCatalog` overrides or
`classificationCatalogExistingConfigMap`, rather
than storing another copy in the workflow ConfigMap.
The default classification ConfigMap name, mount path, `CLASSIFICATION_CATALOG_FILE`
and API endpoint remain unchanged. `classificationCatalog: []` remains an
explicit empty list, not a request for chart defaults; the current API falls
back to settings/built-in values for an empty/invalid catalog. The preparatory
bundle does not reproduce that runtime fallback and must not be treated as a
resolved API catalog. The future loader must share the existing classification
loader. Bundled packaging requires `classificationCatalogEnabled: true`; an
external bundle can be packaged independently.
Custom deployers can maintain files in their own repository, create ConfigMaps
in the release namespace, and use the unchanged chart. `configMaps` adds
their files alongside the bundled Start contract and production classification
catalog. Catalogs can be supplied separately from contracts and mounted into
other consumers without copying their contents. For example, from this
chart directory (replace the namespace):
```sh
kubectl -n YOUR_NAMESPACE create configmap custom-workflow-contracts \
--from-file=pose.yaml=examples/workflow-contracts/pose.yaml
kubectl -n YOUR_NAMESPACE create configmap shared-pose-catalog \
--from-file=pose-keypoints.json=examples/catalogs/pose-keypoints.json
```
Then use [the example values overlay](examples/workflow-configuration-values.yaml).
Each `configMaps` entry has a `name` and `items` mapping ConfigMap keys to
files in `workflow-contracts/` and `catalogs/`. A single read-only projected
volume combines these with the built-in sources, without `subPath`. Paths must
be unique across all sources, including `workflow-contracts/start.yaml`
and `catalogs/classifications.json` when defaults are included. Collisions fail rendering instead of
silently overriding data. Use `classificationCatalog` or
`classificationCatalogExistingConfigMap` to override classification choices rather
than shadowing the shared file. ConfigMap size limits apply.
To manage the complete bundle yourself, set `includeDefaults: false` and supply
a nonempty `configMaps` list. No Start ConfigMap is generated and no default files
are projected. The standalone classification API remains independently configured:
```yaml
workflowConfiguration:
enabled: true
includeDefaults: false
configMaps:
- name: hub-start-contract
items:
- key: start.yaml
path: workflow-contracts/start.yaml
- name: hub-classification-catalog
items:
- key: classifications.json
path: catalogs/classifications.json
```
The older `existingConfigMap`, top-level `items`, and `extraConfigMaps` options
retain their behavior: a nonempty `existingConfigMap` replaces the defaults,
while `extraConfigMaps` extends either that replacement or the defaults.
New deployments should use `includeDefaults` and `configMaps`. Do not mix the
interfaces, even with empty legacy fields. Remove legacy keys from all values
overlays when migrating. The chart omits both interfaces from its defaults so
Helm's value merging does not create a mixed configuration.
Migration: replace legacy extras with `configMaps` and keep `includeDefaults: true`
for additive mode. For legacy replacement mode, set `includeDefaults: false`,
move `existingConfigMap` and its `items` into the first `configMaps` entry, and
append any legacy extras. This produces the same mounts without a special base
ConfigMap. Duplicate destinations remain errors, never order-dependent overrides.
The chart checks item paths and required options but cannot inspect an existing
ConfigMap at render time. Validate its files against the supplied schemas.
Bundled changes update the pod checksum. For externally managed ConfigMaps,
manage rollout/reload in your deployment tooling; the chart cannot checksum
their contents. Missing referenced ConfigMaps or keys prevent Kubernetes from
mounting the volume: create them before syncing Hub (for example using an earlier
Argo CD sync wave). This is distinct from invalid contract/catalog contents,
which the future application loader should report per section without stopping
Hub. Content diagnostics should be advisory in CI, while rendering/mount
configuration errors and failing tests remain blocking. Unavailable sections
must carry explicit errors, never become unrestricted choices or conditions.
Helm validates `values.schema.json`, **not** arbitrary bundled YAML/JSON files.
The standalone schemas support authoring validation. The future loader must
also reject duplicate field/catalog keys, unknown sources, missing/invalid
files, incompatible bounds, unsupported contract versions and unsafe paths.
Validation failures must be surfaced, not silently replaced with empty choices.
#### Runtime semantics to preserve when wiring consumers
Start paths refer to sanitized initial context, never credentials. Other stage
paths are relative to that node instance's result. Conditions may use Start and
all available upstream node outputs, not just the immediately preceding stage;
future or unrelated parallel results are unavailable. Node instances remain
distinct. Conditional branches and missing outputs require explicit availability
handling, even for a field not marked `optional`.
Wildcard positive comparisons match any candidate. Independent wildcard
conditions can match different objects; use the existing same-element `anyMatch`
semantics when predicates must match one detection/keypoint.
The existing Start controls retain their behavior: alternatives within a scope
category are OR; populated device/site/group/classification categories, schedule
and advanced conditions are AND. `conditionMode` only combines the advanced
conditions. Empty selections/schedules are unrestricted. Multiple automatic
Start edges are OR for creating one run; only matched entry branches dispatch.
Manual runs are created before continuation conditions are evaluated, and their
automatic trigger settings remain dormant. Schedule uses recording time and an
IANA timezone, not an invented output field. The advanced editor retains logical
groups and existence checks; neither scheduling nor logical groups are scalar
output fields. Contracts do not redefine the saved workflow wire format.
### Indexing
Following indexes should be executed on the MongoDB database (Kerberos) to improve future performance. Within Kerberos Hub several queries are executed, following indexes will improve the loading times. If not applied you might experience application timeouts or reduced performance when storing lots of data.
@@ -0,0 +1,11 @@
[
{"key": "animal", "label": "Animal", "icon": "animal"},
{"key": "pedestrian", "label": "Pedestrian", "icon": "pedestrian"},
{"key": "cyclist", "label": "Cyclist", "icon": "cyclist"},
{"key": "motorbike", "label": "Motorbike", "icon": "motorbike"},
{"key": "lorry", "label": "Lorry", "icon": "lorry"},
{"key": "car", "label": "Car", "icon": "car"},
{"key": "handbag", "label": "Handbag", "icon": "handbag"},
{"key": "suitcase", "label": "Suitcase", "icon": "suitcase"},
{"key": "cell phone", "label": "Cell phone", "icon": "cell phone"}
]
@@ -0,0 +1,44 @@
# yaml-language-server: $schema=../../schemas/workflow-contract.schema.json
# Metadata only; no current API loader consumes this contract.
schemaVersion: 1
stage: start
contractVersion: 1
# Paths refer to the safe initial context, not the full event or credentials.
fields:
- id: device
label: Device
path: device.deviceKey
type: string
operators: [eq, in]
choices:
source: project.devices
- id: sites
label: Sites
path: device.siteIds.*
type: string
operators: [in]
choices:
source: project.sites
- id: groups
label: Groups
path: device.groupIds.*
type: string
operators: [in]
choices:
source: project.groups
- id: classifications
label: Initial classifications
path: inputs.classify.details.*.classified
type: string
operators: [in]
optional: true
choices:
file: ../catalogs/classifications.json
# Existing structured controls, not fields produced by Start.
controls:
- id: weeklySchedule
label: Weekly schedule
- id: conditions
label: Advanced conditions
@@ -0,0 +1,5 @@
[
{"key": "nose", "label": "Nose"},
{"key": "left_shoulder", "label": "Left shoulder"},
{"key": "right_shoulder", "label": "Right shoulder"}
]
@@ -0,0 +1,13 @@
# Packaging only; does not activate contract loading or deploy a pose worker.
workflowConfiguration:
enabled: true
includeDefaults: true
configMaps:
- name: custom-workflow-contracts
items:
- key: pose.yaml
path: workflow-contracts/pose.yaml
- name: shared-pose-catalog
items:
- key: pose-keypoints.json
path: catalogs/pose-keypoints.json
@@ -0,0 +1,27 @@
# yaml-language-server: $schema=../../schemas/workflow-contract.schema.json
# Hypothetical custom worker output; not an installed stage or universal pose schema.
schemaVersion: 1
stage: pose
contractVersion: 1
fields:
- id: keypoint
label: Keypoint
path: people.*.keypoints.*.name
type: string
operators: [in]
choices:
file: ../catalogs/pose-keypoints.json
- id: posture
label: Posture
path: people.*.posture
type: string
operators: [in]
choices:
values: [standing, sitting, lying]
- id: confidence
label: Keypoint confidence
path: people.*.keypoints.*.confidence
type: number
operators: [gt, gte, lt, lte]
minimum: 0
maximum: 1
+17
View File
@@ -0,0 +1,17 @@
{
"$schema": "http://json-schema.org/draft-07/schema#",
"title": "Hub shared choice catalog v1",
"description": "Compatible with the existing classification JSON format. Keys must be unique; no references to other catalogs.",
"type": "array",
"minItems": 1,
"items": {
"type": "object",
"additionalProperties": false,
"required": ["key", "label"],
"properties": {
"key": {"type": "string", "minLength": 1},
"label": {"type": "string", "minLength": 1},
"icon": {"type": "string", "minLength": 1}
}
}
}
@@ -0,0 +1,142 @@
{
"$schema": "http://json-schema.org/draft-07/schema#",
"title": "Hub workflow condition contract v1 (preparatory)",
"type": "object",
"additionalProperties": false,
"required": ["schemaVersion", "stage", "contractVersion", "fields"],
"properties": {
"schemaVersion": {"const": 1},
"stage": {
"type": "string",
"pattern": "^[a-z][a-z0-9-]*$",
"description": "Stage type, not a node instance. start describes initial context."
},
"contractVersion": {"type": "integer", "minimum": 1},
"fields": {
"type": "array",
"minItems": 1,
"items": {"$ref": "#/definitions/field"}
},
"controls": {
"type": "array",
"items": {
"type": "object",
"additionalProperties": false,
"required": ["id", "label"],
"properties": {
"id": {"enum": ["weeklySchedule", "conditions"]},
"label": {"type": "string", "minLength": 1}
}
}
}
},
"definitions": {
"choices": {
"type": "object",
"additionalProperties": false,
"properties": {
"source": {
"type": "string",
"description": "Built-in authorized project-resource resolver; not a database path.",
"oneOf": [
{"const": "project.devices", "description": "Device keys as values, device names as labels."},
{"const": "project.sites", "description": "Site IDs as values, site names as labels."},
{"const": "project.groups", "description": "Group IDs as values, group names as labels."}
]
},
"file": {
"type": "string",
"pattern": "^\\.\\./catalogs/[A-Za-z0-9_-][A-Za-z0-9._-]*\\.json$",
"description": "Direct shared catalog file relative to the contract; no URLs or nested references."
},
"values": {
"type": "array",
"minItems": 1,
"uniqueItems": true,
"items": {"type": ["string", "number", "boolean"]}
}
},
"oneOf": [
{"required": ["source"]},
{"required": ["file"]},
{"required": ["values"]}
]
},
"field": {
"type": "object",
"additionalProperties": false,
"required": ["id", "label", "path", "type", "operators"],
"properties": {
"id": {
"type": "string",
"pattern": "^[A-Za-z][A-Za-z0-9_-]*$",
"description": "Stable identifier, unique within this contract."
},
"label": {"type": "string", "minLength": 1},
"path": {
"type": "string",
"pattern": "^[A-Za-z_][A-Za-z0-9_-]*(\\.(\\*|[A-Za-z_][A-Za-z0-9_-]*))*$",
"description": "Safe Start-context path, or relative path inside this node instance's output. Wildcards select scalar candidates."
},
"type": {"enum": ["string", "integer", "number", "boolean"]},
"operators": {
"type": "array",
"minItems": 1,
"uniqueItems": true,
"items": {"enum": ["eq", "ne", "in", "exists", "contains", "matches", "gt", "gte", "lt", "lte"]}
},
"optional": {
"type": "boolean",
"default": false,
"description": "Field may be absent even when the node completes; branch availability is determined separately."
},
"minimum": {"type": "number"},
"maximum": {"type": "number"},
"choices": {"$ref": "#/definitions/choices"}
},
"allOf": [
{
"if": {"properties": {"type": {"const": "string"}}},
"then": {
"properties": {
"operators": {"items": {"enum": ["eq", "ne", "in", "exists", "contains", "matches"]}},
"choices": {"properties": {"values": {"items": {"type": "string"}}}}
},
"not": {"anyOf": [{"required": ["minimum"]}, {"required": ["maximum"]}]}
}
},
{
"if": {"properties": {"type": {"const": "boolean"}}},
"then": {
"properties": {
"operators": {"items": {"enum": ["eq", "ne", "in", "exists"]}},
"choices": {"properties": {"values": {"items": {"type": "boolean"}}}}
},
"not": {"anyOf": [{"required": ["minimum"]}, {"required": ["maximum"]}]}
}
},
{
"if": {"properties": {"type": {"enum": ["integer", "number"]}}},
"then": {
"properties": {
"operators": {"items": {"enum": ["eq", "ne", "in", "exists", "gt", "gte", "lt", "lte"]}},
"choices": {"properties": {"values": {"items": {"type": "number"}}}}
}
}
},
{
"if": {"properties": {"type": {"const": "integer"}}},
"then": {"properties": {"choices": {"properties": {"values": {"items": {"type": "integer"}}}}}}
},
{
"if": {"properties": {"type": {"enum": ["integer", "number", "boolean"]}}},
"then": {
"properties": {
"choices": {"not": {"anyOf": [{"required": ["file"]}, {"required": ["source"]}]}}
}
}
}
]
}
}
}
+78
View File
@@ -0,0 +1,78 @@
{{- define "hub.classificationCatalog.configMapName" -}}
{{- default "classification-catalog" .Values.classificationCatalogExistingConfigMap -}}
{{- end -}}
{{/* Keep the legacy endpoint and the preparatory bundle on the same chart data. */}}
{{- define "hub.classificationCatalog" -}}
{{- if kindIs "invalid" .Values.classificationCatalog -}}
{{- .Files.Get "configuration/catalogs/classifications.json" | mustFromJson | toPrettyJson -}}
{{- else if kindIs "slice" .Values.classificationCatalog -}}
{{- .Values.classificationCatalog | toPrettyJson -}}
{{- else -}}
{{- fail "classificationCatalog must be null (bundled defaults) or a list" -}}
{{- end -}}
{{- end -}}
{{/* Normalize both interfaces without changing the caller's values. */}}
{{- define "hub.workflowConfiguration.settings" -}}
{{- $config := .Values.workflowConfiguration | default dict -}}
{{- $unified := or (hasKey $config "includeDefaults") (hasKey $config "configMaps") -}}
{{- $legacy := or (hasKey $config "existingConfigMap") (hasKey $config "items") (hasKey $config "extraConfigMaps") -}}
{{- if and $unified $legacy -}}
{{- fail "workflowConfiguration cannot mix includeDefaults/configMaps with existingConfigMap/items/extraConfigMaps" -}}
{{- end -}}
{{- $includeDefaults := true -}}
{{- $sources := $config.extraConfigMaps | default list -}}
{{- if $unified -}}
{{- $sources = $config.configMaps | default list -}}
{{- if hasKey $config "includeDefaults" -}}
{{- $includeDefaults = $config.includeDefaults -}}
{{- end -}}
{{- else -}}
{{- if $config.existingConfigMap -}}
{{- if and $config.enabled (not $config.items) -}}
{{- fail "workflowConfiguration.items is required with existingConfigMap" -}}
{{- end -}}
{{- $includeDefaults = false -}}
{{- $sources = prepend $sources (dict "name" $config.existingConfigMap "items" ($config.items | default list)) -}}
{{- else if and $config.enabled $config.items -}}
{{- fail "workflowConfiguration.items requires existingConfigMap" -}}
{{- end -}}
{{- end -}}
{{- dict "enabled" ($config.enabled | default false) "includeDefaults" $includeDefaults "configMaps" $sources | toYaml -}}
{{- end -}}
{{- define "hub.workflowConfiguration.validate" -}}
{{- $config := include "hub.workflowConfiguration.settings" . | fromYaml -}}
{{- if $config.enabled -}}
{{- $paths := dict -}}
{{- if $config.includeDefaults -}}
{{- if not .Values.classificationCatalogEnabled -}}
{{- fail "bundled workflowConfiguration requires classificationCatalogEnabled; disable includeDefaults and supply configMaps for a custom bundle" -}}
{{- end -}}
{{- $_ := set $paths "workflow-contracts/start.yaml" true -}}
{{- $_ := set $paths "catalogs/classifications.json" true -}}
{{- else -}}
{{- if not $config.configMaps -}}
{{- fail "workflowConfiguration.configMaps must be nonempty when enabled without defaults" -}}
{{- end -}}
{{- end -}}
{{- range $config.configMaps -}}
{{- if not .items -}}
{{- fail (printf "workflowConfiguration ConfigMap %q requires nonempty items" .name) -}}
{{- end -}}
{{- range .items -}}
{{- if not (regexMatch "^[A-Za-z0-9._-]+$" .key) -}}
{{- fail "workflowConfiguration.items keys must be valid ConfigMap keys" -}}
{{- end -}}
{{- if or (not (regexMatch "^(workflow-contracts|catalogs)/[A-Za-z0-9_-][A-Za-z0-9._-]*\\.(yaml|json)$" .path)) (ne (clean .path) .path) -}}
{{- fail "workflowConfiguration.items paths must be normalized relative files under workflow-contracts/ or catalogs/" -}}
{{- end -}}
{{- if hasKey $paths .path -}}
{{- fail (printf "duplicate workflowConfiguration.items path %q" .path) -}}
{{- end -}}
{{- $_ := set $paths .path true -}}
{{- end -}}
{{- end -}}
{{- end -}}
{{- end -}}
+36
View File
@@ -33,6 +33,42 @@
{{- $uri -}}
{{- end -}}
{{/*
MongoDB URI with appName set to one workload, so the Atlas query profiler and
currentOp attribute each operation to the service that issued it. Renders
nothing when no URI is configured or mongodb.appNamePerService is false.
Usage: include "hub.mongodb.appUri" (dict "root" $ "app" "hub-api")
*/}}
{{- define "hub.mongodb.appUri" -}}
{{- $uri := include "hub.mongodb.uri" .root -}}
{{- if and $uri .root.Values.mongodb.appNamePerService -}}
{{- if regexMatch "(?i)[?&]appName=" $uri -}}
{{- $uri = regexReplaceAll "(?i)([?&])appName=[^&]*" $uri (printf "${1}appName=%s" .app) -}}
{{- else -}}
{{- $separator := "&" -}}
{{- if not (contains "?" $uri) -}}
{{- $separator = "?" -}}
{{- else if or (hasSuffix "?" $uri) (hasSuffix "&" $uri) -}}
{{- $separator = "" -}}
{{- end -}}
{{- $uri = printf "%s%sappName=%s" $uri $separator .app -}}
{{- end -}}
{{- $uri -}}
{{- end -}}
{{- end -}}
{{/*
Container env entry that overrides MONGODB_URI from mongodb-config with the
workload's own appName. An explicit env entry takes precedence over envFrom.
Usage: {{- include "hub.mongodb.appUriEnv" (dict "root" $ "app" "hub-api") | nindent 12 }}
*/}}
{{- define "hub.mongodb.appUriEnv" -}}
{{- with include "hub.mongodb.appUri" . -}}
- name: MONGODB_URI
value: {{ . | quote }}
{{- end -}}
{{- end -}}
{{/* Render the shared MongoDB CA Secret volume. */}}
{{- define "hub.mongodb.tlsVolume" -}}
- name: mongodb-tls
@@ -1,4 +1,4 @@
{{- if .Values.classificationCatalogEnabled }}
{{- if and .Values.classificationCatalogEnabled (not .Values.classificationCatalogExistingConfigMap) }}
apiVersion: v1
kind: ConfigMap
metadata:
@@ -6,5 +6,5 @@ metadata:
namespace: {{ .Release.Namespace }}
data:
classifications.json: |-
{{- toPrettyJson .Values.classificationCatalog | nindent 4 }}
{{- include "hub.classificationCatalog" . | nindent 4 }}
{{- end }}
@@ -0,0 +1,12 @@
{{- include "hub.workflowConfiguration.validate" . -}}
{{- $config := include "hub.workflowConfiguration.settings" . | fromYaml -}}
{{- if and $config.enabled $config.includeDefaults (or (eq .Values.mode "all") (eq .Values.mode "ui")) }}
apiVersion: v1
kind: ConfigMap
metadata:
name: {{ printf "%s-workflow-configuration" .Release.Name | trunc 63 | trimSuffix "-" }}
namespace: {{ .Release.Namespace }}
data:
start.yaml: |-
{{- .Files.Get "configuration/workflow-contracts/start.yaml" | nindent 4 }}
{{- end }}
+63 -4
View File
@@ -1,4 +1,5 @@
{{- if or (eq .Values.mode "all") (eq .Values.mode "ui") -}}
{{- $workflowConfiguration := include "hub.workflowConfiguration.settings" . | fromYaml -}}
{{- if .Values.kerberoshub.api.serviceEnabled }}
apiVersion: v1
kind: Service
@@ -120,9 +121,12 @@ spec:
metadata:
annotations:
checksum/config-mongodb: {{ include (print $.Template.BasePath "/configmap-mongodb.yaml") . | sha256sum }}
{{- if .Values.classificationCatalogEnabled }}
{{- if and .Values.classificationCatalogEnabled (not .Values.classificationCatalogExistingConfigMap) }}
checksum/classification-catalog: {{ include (print $.Template.BasePath "/configmap-classification-catalog.yaml") . | sha256sum }}
{{- end }}
{{- if and $workflowConfiguration.enabled $workflowConfiguration.includeDefaults }}
checksum/workflow-configuration: {{ include (print $.Template.BasePath "/configmap-workflow-configuration.yaml") . | sha256sum }}
{{- end }}
labels:
app: hub-api
spec:
@@ -136,12 +140,40 @@ spec:
{{- $serverTLS := .Values.kerberoshub.api.serverTLS }}
{{- $mongodbTLS := .Values.mongodb.tls }}
{{- $classificationCatalogEnabled := .Values.classificationCatalogEnabled }}
{{- if or $classificationCatalogEnabled .Values.kerberoshub.api.volumes (and $serverTLS.enabled $serverTLS.secretName) (and $mongodbTLS.enabled $mongodbTLS.existingSecret) }}
{{- if or $workflowConfiguration.enabled $classificationCatalogEnabled .Values.kerberoshub.api.volumes (and $serverTLS.enabled $serverTLS.secretName) (and $mongodbTLS.enabled $mongodbTLS.existingSecret) }}
volumes:
{{- if $workflowConfiguration.enabled }}
- name: workflow-configuration
projected:
sources:
{{- if $workflowConfiguration.includeDefaults }}
- configMap:
name: {{ printf "%s-workflow-configuration" .Release.Name | trunc 63 | trimSuffix "-" }}
items:
- key: start.yaml
path: workflow-contracts/start.yaml
- configMap:
name: {{ include "hub.classificationCatalog.configMapName" . }}
items:
- key: classifications.json
path: catalogs/classifications.json
{{- end }}
{{- range $workflowConfiguration.configMaps }}
- configMap:
name: {{ .name | quote }}
items:
{{- toYaml .items | nindent 20 }}
{{- end }}
{{- end }}
{{- if $classificationCatalogEnabled }}
- name: classification-catalog
configMap:
name: classification-catalog
name: {{ include "hub.classificationCatalog.configMapName" . }}
{{- if .Values.classificationCatalogExistingConfigMap }}
items:
- key: classifications.json
path: classifications.json
{{- end }}
{{- end }}
{{- with .Values.kerberoshub.api.volumes }}
{{- toYaml . | nindent 8 }}
@@ -170,8 +202,13 @@ spec:
ports:
- containerPort: 80
name: http
{{- if or $classificationCatalogEnabled .Values.kerberoshub.api.volumeMounts (and $serverTLS.enabled $serverTLS.secretName) (and $mongodbTLS.enabled $mongodbTLS.existingSecret) }}
{{- if or $workflowConfiguration.enabled $classificationCatalogEnabled .Values.kerberoshub.api.volumeMounts (and $serverTLS.enabled $serverTLS.secretName) (and $mongodbTLS.enabled $mongodbTLS.existingSecret) }}
volumeMounts:
{{- if $workflowConfiguration.enabled }}
- name: workflow-configuration
mountPath: /etc/kerberos/configuration
readOnly: true
{{- end }}
{{- if $classificationCatalogEnabled }}
- name: classification-catalog
mountPath: /etc/kerberos/classifications
@@ -194,6 +231,7 @@ spec:
- configMapRef:
name: mongodb-config
env:
{{- include "hub.mongodb.appUriEnv" (dict "root" $ "app" "hub-api") | nindent 12 }}
{{- if $classificationCatalogEnabled }}
- name: CLASSIFICATION_CATALOG_FILE
value: /etc/kerberos/classifications/classifications.json
@@ -275,6 +313,27 @@ spec:
- name: KERBEROS_JWT_SECRET
value: "{{ .Values.kerberoshub.api.jwtSecret }}"
# Temporarily stop serving the polled GET /devices list (204, no DB reads).
- name: DISABLE_AGENT_HEARTBEAT
value: {{ .Values.kerberoshub.api.disableAgentHeartbeat | default false | quote }}
# Media proxy: hide Vault/storage endpoints behind hub-api URLs.
{{- $mediaProxy := .Values.kerberoshub.api.mediaProxy | default dict }}
- name: MEDIA_PROXY_ENABLED
value: {{ $mediaProxy.enabled | default false | quote }}
{{- if $mediaProxy.publicUrl }}
- name: MEDIA_PROXY_PUBLIC_URL
value: {{ $mediaProxy.publicUrl | quote }}
{{- end }}
{{- if $mediaProxy.key }}
- name: MEDIA_PROXY_KEY
value: {{ $mediaProxy.key | quote }}
{{- end }}
{{- if $mediaProxy.ttl }}
- name: MEDIA_PROXY_TTL
value: {{ $mediaProxy.ttl | quote }}
{{- end }}
# SSO (OIDC) setup
- name: SSO_DOMAINS
value: "{{- range .Values.kerberoshub.api.sso }}{{ .domain }};{{- end }}"
@@ -114,6 +114,7 @@ spec:
- configMapRef:
name: mongodb-config
env:
{{- include "hub.mongodb.appUriEnv" (dict "root" $ "app" "hub-audit") | nindent 12 }}
- name: RABBITMQ_HOST
value: {{ .Values.rabbitmq.host | quote }}
- name: RABBITMQ_EXCHANGE
@@ -66,6 +66,7 @@ spec:
ports:
- containerPort: 8080
env:
{{- include "hub.mongodb.appUriEnv" (dict "root" $ "app" "hub-cleanup") | nindent 12 }}
- name: MODE
value: "{{ .Values.kerberoshub.cleanup.mode }}"
- name: LOG_LEVEL
@@ -369,6 +369,8 @@ spec:
value: "{{ .Values.kerberoshub.frontend.features.media.filter.objectDetection.enabled }}"
- name: FEATURE_MEDIA_FILTER_STAR_ENABLED
value: "{{ .Values.kerberoshub.frontend.features.media.filter.star.enabled }}"
- name: FEATURE_MEDIA_FILTER_ANALYSIS_ENABLED
value: "{{ .Values.kerberoshub.frontend.features.media.filter.analysis.enabled }}"
- name: FEATURE_MEDIA_FILTER_REGION_ENABLED
value: "{{ .Values.kerberoshub.frontend.features.media.filter.region.enabled }}"
- name: FEATURE_MEDIA_FILTER_SORT_ENABLED
@@ -461,6 +461,8 @@ spec:
value: "{{ .Values.kerberoshub.frontend.features.media.filter.objectDetection.enabled }}"
- name: FEATURE_MEDIA_FILTER_STAR_ENABLED
value: "{{ .Values.kerberoshub.frontend.features.media.filter.star.enabled }}"
- name: FEATURE_MEDIA_FILTER_ANALYSIS_ENABLED
value: "{{ .Values.kerberoshub.frontend.features.media.filter.analysis.enabled }}"
- name: FEATURE_MEDIA_FILTER_REGION_ENABLED
value: "{{ .Values.kerberoshub.frontend.features.media.filter.region.enabled }}"
- name: FEATURE_MEDIA_FILTER_SORT_ENABLED
@@ -64,6 +64,7 @@ spec:
- configMapRef:
name: mongodb-config
env:
{{- include "hub.mongodb.appUriEnv" (dict "root" $ "app" "hub-monitor-device") | nindent 12 }}
- name: LOG_LEVEL
value: "{{ .Values.kerberoshub.monitordevice.logLevel }}"
- name: READ_ONLY
@@ -57,6 +57,7 @@ spec:
- configMapRef:
name: mongodb-config
env:
{{- include "hub.mongodb.appUriEnv" (dict "root" $ "app" "hub-reactivatesubscriptions") | nindent 12 }}
- name: LOG_LEVEL
value: "{{ .Values.kerberoshub.reactivate.logLevel }}"
- name: READ_ONLY
@@ -1,3 +1,28 @@
{{/*
Reserve every effective identity, including disabled definitions, so enabling a
workflow later cannot silently alias another workflow's runs or references.
*/}}
{{- define "kerberoshub.workflows.validateDefinitions" -}}
{{- $identities := dict -}}
{{- range $name, $wf := .Values.kerberoshub.workflows.definitions -}}
{{- $id := sha256sum $name | trunc 24 -}}
{{- if hasKey $wf "id" -}}
{{- $invalid := printf "kerberoshub.workflows.definitions[%q].id must be a non-zero 24-character hexadecimal string (Mongo ObjectID); omit id for name-derived compatibility" $name -}}
{{- if not (kindIs "string" $wf.id) -}}
{{- fail $invalid -}}
{{- end -}}
{{- if or (not (regexMatch "^[0-9a-fA-F]{24}$" $wf.id)) (eq $wf.id "000000000000000000000000") -}}
{{- fail $invalid -}}
{{- end -}}
{{- $id = lower $wf.id -}}
{{- end -}}
{{- if hasKey $identities $id -}}
{{- fail (printf "kerberoshub.workflows.definitions: duplicate effective workflow id %q for %q and %q (including disabled definitions)" $id (index $identities $id) $name) -}}
{{- end -}}
{{- $_ := set $identities $id $name -}}
{{- end -}}
{{- end -}}
{{/*
Assemble the deployment-global workflow definitions (WORKFLOW_DEFINITIONS) as a
JSON array from every *enabled* workflow under kerberoshub.workflows.definitions.
@@ -7,9 +32,9 @@ opens its own run and dispatches only its own stages. Organisation-scoped
database workflows are discovered separately at runtime.
Each enabled definition contributes one workflow object:
name the map key (the workflow's human-readable name; also its identity —
the engine derives a stable id from it when the definition carries
no explicit id).
id optional explicit stable identity, normalized to lowercase. Omitted
for legacy definitions; their effective id is SHA256(name)[:24].
name the map key (mutable display name when an explicit id is supplied).
enabled always true here (a disabled definition is skipped entirely).
source "config" — provenance marking a Helm-defined, deployment-global,
ops-managed workflow (read-only in the API, no owning organisation).
@@ -28,6 +53,7 @@ Each enabled definition contributes one workflow object:
needsMode conditional stages: "any" (default) | "all".
*/}}
{{- define "kerberoshub.workflows.workflowDefinitions" -}}
{{- include "kerberoshub.workflows.validateDefinitions" . -}}
{{- $defs := list -}}
{{- $services := .Values.kerberoshub.services | default dict -}}
{{- range $name, $wf := .Values.kerberoshub.workflows.definitions -}}
@@ -43,6 +69,7 @@ Each enabled definition contributes one workflow object:
{{- $stages = append $stages $entry -}}
{{- end -}}
{{- $def := dict "name" $name "enabled" true "source" "config" "triggers" (default (list (dict "type" "automatic")) $wf.triggers) "stages" $stages -}}
{{- if hasKey $wf "id" }}{{- $_ := set $def "id" (lower $wf.id) -}}{{- end }}
{{- $defs = append $defs $def -}}
{{- end -}}
{{- end -}}
@@ -1,3 +1,4 @@
{{- include "kerberoshub.workflows.validateDefinitions" . -}}
{{- if and (or (eq .Values.mode "all") (eq .Values.mode "pipeline")) .Values.kerberoshub.workflows.enabled -}}
apiVersion: apps/v1
kind: Deployment
@@ -64,6 +65,7 @@ spec:
- configMapRef:
name: mongodb-config
env:
{{- include "hub.mongodb.appUriEnv" (dict "root" $ "app" "hub-workflows") | nindent 8 }}
- name: LOG_LEVEL
value: "{{ .Values.kerberoshub.services.workflows.logLevel }}"
- name: QUEUE_SYSTEM
@@ -68,6 +68,7 @@ spec:
- configMapRef:
name: mongodb-config
env:
{{- include "hub.mongodb.appUriEnv" (dict "root" $ "app" "hub-pipeline-analysis") | nindent 8 }}
- name: LOG_LEVEL
value: "{{ .Values.kerberospipeline.analysis.logLevel }}"
- name: CLOUD_PROVIDER
@@ -70,6 +70,7 @@ spec:
- configMapRef:
name: mongodb-config
env:
{{- include "hub.mongodb.appUriEnv" (dict "root" $ "app" "hub-pipeline-event") | nindent 8 }}
- name: LOG_LEVEL
value: "{{ .Values.kerberospipeline.event.logLevel }}"
- name: READ_ONLY
@@ -64,6 +64,7 @@ spec:
- configMapRef:
name: mongodb-config
env:
{{- include "hub.mongodb.appUriEnv" (dict "root" $ "app" "hub-pipeline-export") | nindent 8 }}
- name: LOG_LEVEL
value: "{{ .Values.kerberospipeline.export.logLevel }}"
{{- if .Values.kerberospipeline.export.playerAssetsPath }}
@@ -68,6 +68,7 @@ spec:
- configMapRef:
name: mongodb-config
env:
{{- include "hub.mongodb.appUriEnv" (dict "root" $ "app" "hub-pipeline-monitor") | nindent 8 }}
# Application
- name: LOG_LEVEL
value: "{{ .Values.kerberospipeline.monitor.logLevel }}"
@@ -64,6 +64,7 @@ spec:
- configMapRef:
name: mongodb-config
env:
{{- include "hub.mongodb.appUriEnv" (dict "root" $ "app" "hub-pipeline-notification-test") | nindent 8 }}
- name: QUEUE_SYSTEM
value: "{{ .Values.queueProvider }}"
@@ -68,6 +68,7 @@ spec:
- configMapRef:
name: mongodb-config
env:
{{- include "hub.mongodb.appUriEnv" (dict "root" $ "app" "hub-pipeline-notification") | nindent 8 }}
- name: LOG_LEVEL
value: "{{ .Values.kerberospipeline.notify.logLevel }}"
- name: NOTIFICATION_EXPIRY_MINUTES
@@ -68,6 +68,7 @@ spec:
- configMapRef:
name: mongodb-config
env:
{{- include "hub.mongodb.appUriEnv" (dict "root" $ "app" "hub-pipeline-sequence") | nindent 8 }}
- name: CLOUD_PROVIDER
value: "{{ .Values.cloudProvider }}"
- name: QUEUE_SYSTEM
@@ -68,6 +68,7 @@ spec:
- configMapRef:
name: mongodb-config
env:
{{- include "hub.mongodb.appUriEnv" (dict "root" $ "app" "hub-pipeline-throttler") | nindent 8 }}
- name: LOG_LEVEL
value: "{{ .Values.kerberospipeline.throttler.logLevel }}"
- name: CLOUD_PROVIDER
@@ -62,6 +62,7 @@ spec:
- configMapRef:
name: mongodb-config
env:
{{- include "hub.mongodb.appUriEnv" (dict "root" $ "app" "vault-forwarder") | nindent 10 }}
- name: ITERATION_SPEED
value: "5"
- name: BUFFER_TIME
+76
View File
@@ -0,0 +1,76 @@
{
"$schema": "http://json-schema.org/draft-07/schema#",
"type": "object",
"properties": {
"classificationCatalogExistingConfigMap": {
"type": "string",
"maxLength": 253,
"pattern": "^$|^[a-z0-9]([-a-z0-9]*[a-z0-9])?(\\.[a-z0-9]([-a-z0-9]*[a-z0-9])?)*$"
},
"classificationCatalog": {"type": ["array", "null"]},
"workflowConfiguration": {
"type": "object",
"additionalProperties": false,
"required": ["enabled"],
"not": {
"allOf": [
{"anyOf": [{"required": ["includeDefaults"]}, {"required": ["configMaps"]}]},
{"anyOf": [{"required": ["existingConfigMap"]}, {"required": ["items"]}, {"required": ["extraConfigMaps"]}]}
]
},
"properties": {
"enabled": {"type": "boolean"},
"includeDefaults": {
"type": "boolean",
"description": "Include bundled Start metadata and the effective classification catalog; true when omitted."
},
"configMaps": {
"type": "array",
"items": {"$ref": "#/definitions/configMapSource"}
},
"extraConfigMaps": {
"type": "array",
"description": "Legacy additive sources; use configMaps for new deployments.",
"items": {"$ref": "#/definitions/configMapSource"}
},
"existingConfigMap": {
"type": "string",
"maxLength": 253,
"pattern": "^$|^[a-z0-9]([-a-z0-9]*[a-z0-9])?(\\.[a-z0-9]([-a-z0-9]*[a-z0-9])?)*$"
},
"items": {
"type": "array",
"items": {"$ref": "#/definitions/configMapItem"}
}
}
}
},
"definitions": {
"configMapSource": {
"type": "object",
"additionalProperties": false,
"required": ["name", "items"],
"properties": {
"name": {
"type": "string",
"maxLength": 253,
"pattern": "^[a-z0-9]([-a-z0-9]*[a-z0-9])?(\\.[a-z0-9]([-a-z0-9]*[a-z0-9])?)*$"
},
"items": {
"type": "array",
"minItems": 1,
"items": {"$ref": "#/definitions/configMapItem"}
}
}
},
"configMapItem": {
"type": "object",
"additionalProperties": false,
"required": ["key", "path"],
"properties": {
"key": {"type": "string", "pattern": "^[A-Za-z0-9._-]+$"},
"path": {"type": "string", "pattern": "^(workflow-contracts|catalogs)/[A-Za-z0-9_-][A-Za-z0-9._-]*\\.(yaml|json)$"}
}
}
}
}
+70 -49
View File
@@ -46,6 +46,10 @@ mongodb:
# MongoDB URI (for example for a SaaS service like MongoDB Atlas)
# If uri is set, the below properties are not used (host, adminDatabase, username, password)
#uri: "mongodb+srv://xx:xx@kerberos-hub.xxx.mongodb.net/?retryWrites=true&w=majority&appName=xxx"
# When uri is set, give each workload its own appName (hub-api,
# hub-pipeline-monitor, ...) instead of the one in uri, so the Atlas query
# profiler and currentOp show which service issued an operation.
appNamePerService: true
# Self-hosted mongodb, you could also use the uri property above.
host: mongodb.mongodb
@@ -74,34 +78,25 @@ mongodb:
# downstream processing. Classifier output keys must match these values.
# Disable this to use the legacy classification catalog stored in MongoDB.
classificationCatalogEnabled: true
classificationCatalog:
- key: animal
label: Animal
icon: animal
- key: pedestrian
label: Pedestrian
icon: pedestrian
- key: cyclist
label: Cyclist
icon: cyclist
- key: motorbike
label: Motorbike
icon: motorbike
- key: lorry
label: Lorry
icon: lorry
- key: car
label: Car
icon: car
- key: handbag
label: Handbag
icon: handbag
- key: suitcase
label: Suitcase
icon: suitcase
- key: cell phone
label: Cell phone
icon: cell phone
# Optional existing ConfigMap in the release namespace with classifications.json.
# When set, do not create a catalog ConfigMap; this takes precedence over the list.
classificationCatalogExistingConfigMap: ""
# classificationCatalog is optional and omitted by default, which uses
# configuration/catalogs/classifications.json. Set it to a list to override.
# Explicit null also uses defaults; [] retains the API fallback behavior.
# Preparatory configuration packaging only: the API does not yet load workflow
# contracts. Enabling this mounts files but does not activate a new editor/runtime.
workflowConfiguration:
enabled: false
# Omitted includeDefaults means true; false requires a nonempty configMaps list.
# includeDefaults: true
# configMaps: []
# Each ConfigMap supplies name and items [{key, path}] in the release namespace.
# Paths under workflow-contracts/ or catalogs/ must be unique across all sources.
# Legacy existingConfigMap/items/extraConfigMaps remain supported, but cannot
# be mixed with includeDefaults/configMaps. Both interfaces are omitted here
# so Helm's default merging does not introduce a false mixed configuration.
###################################################
# MQTT configuration (bi-directional communication)
###################################################
@@ -295,6 +290,19 @@ kerberoshub:
jwtSecret: "this-is-a-secret-please-change-to-random-string" # change to a random value, this is for generating JWT tokens.
schema: "https"
url: "api.yourdomain.com"
# Temporarily stop serving the polled GET /devices list: hub-api answers 204
# without querying MongoDB. The frontend still loads devices once through
# /session/bootstrap, but device online/offline status stops refreshing.
disableAgentHeartbeat: false
# Media proxy: when enabled, hub-api wraps every media URL it receives from
# Kerberos Vault into an opaque hub-api URL ({schema}://{url}/media/proxy?t=...)
# and streams the object itself, so clients never see Vault or the
# underlying storage endpoint. Increases hub-api egress/bandwidth.
mediaProxy:
enabled: false
publicUrl: "" # Optional override of the public hub-api base URL; defaults to {schema}://{url}.
key: "" # Optional encryption key (openssl rand -hex 32); defaults to a key derived from jwtSecret. Must be identical on all replicas.
ttl: "" # Optional lifetime of proxied URLs (Go duration, e.g. "12h"); defaults to 24h, max 168h.
resources:
requests:
memory: 100Mi
@@ -583,6 +591,8 @@ kerberoshub:
enabled: "true" # Enable or disable object detection filter in media 'true' or 'false'
star:
enabled: "true" # Enable or disable star filter in media 'true' or 'false'
analysis:
enabled: "true" # Enable or disable the "with analysis" filter (hides recordings without markers, description or detections) in media 'true' or 'false'
region:
enabled: "true" # Enable or disable region filter in media 'true' or 'false'
sort:
@@ -791,6 +801,13 @@ kerberoshub:
# more config workflows.
#
# Each definition:
# id optional stable identity: quote a non-zero 24-hex Mongo ObjectID.
# Recommended for new definitions; rendered lowercase. Omit for
# legacy SHA256(exact map key)[:24] identity. Pin that old ID before
# renaming an existing workflow; changing id creates a new identity.
# Invalid IDs and duplicate effective IDs fail Helm rendering,
# including disabled definitions. Name/version are not identity
# when id is explicit. No per-project on/off overrides are provided.
# enabled include this workflow (soft-delete toggle).
# source always rendered as "config" (a Helm-defined, ops-managed,
# deployment-global workflow — read-only in the API).
@@ -815,27 +832,31 @@ kerberoshub:
# kerberoshub.services.<operation> (deploy the objecttracking / loitering
# workers below).
definitions: {}
#tracking-workflow:
# enabled: true
# triggers:
# - type: automatic
# stages:
# - operation: objecttracking
# dispatch: always
# - operation: loitering
# dispatch: conditional
# # Fire loitering once objecttracking has resolved — a readiness join
# # (no condition ⇒ gate on the upstream's presence, not a value).
# needs:
# - operation: objecttracking
#vlm-workflow:
# # Describe every recording with the hub-vlm worker (services.vlm below).
# enabled: true
# triggers:
# - type: automatic
# stages:
# - operation: vlm
# dispatch: always
# Replace {} above with the indented examples below to enable them. Their IDs
# are pinned to the existing name-derived identities; do not regenerate them.
# tracking-workflow:
# id: "ec7e72bc4cedd32018e1f92a"
# enabled: true
# triggers:
# - type: automatic
# stages:
# - operation: objecttracking
# dispatch: always
# - operation: loitering
# dispatch: conditional
# # Fire loitering once objecttracking has resolved — a readiness join
# # (no condition ⇒ gate on the upstream's presence, not a value).
# needs:
# - operation: objecttracking
# vlm-workflow:
# id: "9e3666c56854b2befdb1f645"
# # Describe every recording with the hub-vlm worker (services.vlm below).
# enabled: true
# triggers:
# - type: automatic
# stages:
# - operation: vlm
# dispatch: always
# Stages tenants may place on the workflow editor canvas next to the device
# node. Each entry needs a deployed worker (services.<operation>.queue);
# entries without one are hidden. Empty means the editor offers no stages.
@@ -1,97 +0,0 @@
#!/usr/bin/env bash
#
# Render the hub chart and assert that every deployment which carries the
# workflows hand-off queue (the WORKFLOWS_QUEUE env var) resolves to the SAME,
# non-empty value.
#
# Why: the analysis pipeline (pipe-analysis) publishes opened workflow runs to
# WORKFLOWS_QUEUE, the workflows engine (hub-workflows) consumes it, and every
# stage worker (hub-stage) routes its result back to it. All three templates
# read the single key `kerberoshub.services.workflows.queue`. If a future edit
# hardcodes a value, reads the wrong key, or drops the env on one of them, the
# producer and consumer silently drift onto different queue names and messages
# pile up with no consumer. This check fails the build before that can ship.
#
# Usage: scripts/check-workflows-queue-consistency.sh [chart-dir]
# (chart-dir defaults to charts/hub, relative to the repo root)
set -euo pipefail
CHART_DIR="${1:-charts/hub}"
PROBE="drift-probe-queue-name"
# Flags that force all three deployment kinds (analysis, engine and one stage
# worker) to render, so the check actually has something to compare. The chart
# ships NO enabled stage worker by default (custom stages are values-only and
# opt-in), so we synthesise a throwaway stage purely to exercise the generic
# hub-stage path. The name is a neutral fixture ("queuecheck") on purpose: any
# arbitrary stage key must render the same way, so the check must not depend on
# a specific bundled worker.
STAGE="queuecheck"
RENDER_FLAGS=(
--set mode=all
--set kerberoshub.workflows.enabled=true
--set "kerberoshub.workflows.stages.${STAGE}.enabled=true"
--set "kerberoshub.services.${STAGE}.enabled=true"
--set "kerberoshub.services.${STAGE}.repository=example.invalid/queuecheck"
--set "kerberoshub.services.${STAGE}.tag=test"
--set "kerberoshub.services.${STAGE}.queue=queuecheck-fixture-queue"
)
# Read `helm template` output on stdin and print one WORKFLOWS_QUEUE value per
# line. Matches the `- name: WORKFLOWS_QUEUE` env entry and captures the value
# from the following `value:` line, skipping blank/comment lines in between.
extract_workflows_queue() {
awk '
/^[[:space:]]*-[[:space:]]*name:[[:space:]]*WORKFLOWS_QUEUE[[:space:]]*$/ { want=1; next }
want==1 {
if ($0 ~ /^[[:space:]]*#/ || $0 ~ /^[[:space:]]*$/) next
v=$0
sub(/^[[:space:]]*value:[[:space:]]*/, "", v)
sub(/^"/, "", v); sub(/"[[:space:]]*$/, "", v)
sub(/[[:space:]]+$/, "", v)
print v
want=0
}
'
}
assert_all_equal() {
local expected="$1"; shift
local label="$1"; shift
local -a vals=("$@")
if [ "${#vals[@]}" -lt 2 ]; then
echo "FAIL (${label}): expected at least 2 WORKFLOWS_QUEUE values (analysis + engine), found ${#vals[@]}" >&2
return 1
fi
local v
for v in "${vals[@]}"; do
if [ -z "${v}" ]; then
echo "FAIL (${label}): a deployment rendered an empty WORKFLOWS_QUEUE value" >&2
return 1
fi
if [ "${v}" != "${expected}" ]; then
echo "FAIL (${label}): WORKFLOWS_QUEUE drift detected — expected '${expected}' but a deployment rendered '${v}'" >&2
printf ' rendered values: %s\n' "${vals[*]}" >&2
return 1
fi
done
echo "OK (${label}): ${#vals[@]} deployments all use WORKFLOWS_QUEUE='${expected}'"
}
echo "== Rendering ${CHART_DIR} with the chart's default workflows queue =="
default_out="$(helm template hub "${CHART_DIR}" "${RENDER_FLAGS[@]}")"
mapfile -t default_vals < <(printf '%s\n' "${default_out}" | extract_workflows_queue)
default_queue="${default_vals[0]:-}"
assert_all_equal "${default_queue}" "default values" "${default_vals[@]}" || exit 1
echo "== Rendering ${CHART_DIR} with an overridden workflows queue (-> ${PROBE}) =="
probe_out="$(helm template hub "${CHART_DIR}" "${RENDER_FLAGS[@]}" \
--set kerberoshub.services.workflows.queue="${PROBE}")"
mapfile -t probe_vals < <(printf '%s\n' "${probe_out}" | extract_workflows_queue)
assert_all_equal "${PROBE}" "override probe" "${probe_vals[@]}" || exit 1
echo "All WORKFLOWS_QUEUE consistency checks passed."