Skip to content

Add optional authenticated Alertmanager webhook receiver - #78

Open
JeffPengCoder wants to merge 1 commit into
NVIDIA:mainfrom
JeffPengCoder:feat/alertmanager-webhook
Open

JeffPengCoder wants to merge 1 commit into
NVIDIA:mainfrom
JeffPengCoder:feat/alertmanager-webhook

Conversation

@JeffPengCoder

@JeffPengCoder JeffPengCoder commented Oct 29, 2026 •

Copy link
Copy Markdown
Contributor

Description

Mosaic currently pulls active alerts from Alertmanager. Add an optional push receiver so operators can route standard Alertmanager v4 notifications into the existing Mosaic alert list: firing notifications upsert alertmanager:<fingerprint>, repeated deliveries update the same entry, and resolved notifications remove it. Closes #77.

The default-off alertmanagerWebhook component uses a dedicated inbound bearer-token Secret and the existing UI machine-token Secret for internal delivery. It provides a ClusterIP Service and optional TLS ingress restricted to the exact webhook path, validates the complete group before forwarding, bounds request size and delivery time, and returns retryable errors for downstream failure or backpressure. The installation builder includes the receiver and its resource footprint. The guide covers credentials, Alertmanager routes and Secret mounts, administrator responsibilities, and firing/resolution smoke tests.

This change delivers alerts. Existing Auto investigate triggers still come from the open Mosaic browser page; it does not add an unattended investigation worker or Teams notification delivery. It also does not configure an external Alertmanager on the operator's behalf. The UI's process-local alert store requires one UI replica, has no durable incident history, and remains subject to the documented retry/reordering/restart limits. Use webhook delivery with Auto pull off.

Validation

  • GitHub checks passed on 70696fbc8064d13eef438a6912482c597b1f8e65: CI / Tests and Documentation / build.
  • Repository suite with ALERTMANAGER_BINARY set: 236 tests, 234 passed, 2 conditional skips, 0 failures. The skips are the existing OpenClaw-runtime native-memory and external VictoriaMetrics integration checks.
  • Focused receiver/chart suite with real Alertmanager 0.28.1: 19/19 passed, including an isolated Alertmanager sending two real firing webhooks and subsequent per-alert resolution through bearer authentication.
  • Standalone inference suite: 4/4 passed; both Helm charts linted successfully; enabled receiver template rendered successfully.
  • Documentation: Sphinx -W --keep-going build passed; all four YAML examples parsed; amtool check-config and route testing passed with both Mosaic and an existing notification receiver preserved; regenerated agent-documentation bundle passed freshness check.
  • Isolated compatibility test against the chart's default private mosaic-ui:348d4635 image, Node v22.23.2:
Case Receiver HTTP status Firing forwarded Resolved forwarded UI entries after delivery
Grouped firing 200 2 0 2
Repeat delivery 200 2 0 2
Mixed status 200 1 1 1
Resolved 200 0 1 0
Repeat resolved 200 0 1 0

The compatibility check used synthetic alerts, isolated UI state, and separate disposable credentials. Its Pod and ConfigMap were removed. Existing Mosaic deployments and live Alertmanager routing were not changed. Full synthetic result records, the exact API-test harness, and complete local test output are attached to this PR. This validates the receiver protocol and the UI API separately; a customer's network, TLS ingress, routing, browser-triggered investigation, and Teams delivery still need deployment-specific validation.

Checklist

@JeffPengCoder

Copy link
Copy Markdown
Contributor Author

Validation evidence for this PR (2026-10-29)

All UI records below come from a disposable isolated container with synthetic alerts and disposable test credentials. They contain no production alert data, registry credentials, or UI tokens. The default chart image was mosaic-ui:348d4635, resolved in the isolated test to sha256:9787b6df0b689e8b92398bbe0fa35491a6943ad4287fca09089c1ae91332b1ca. The test resources were removed afterward.

The test has two complementary parts: the committed optional test starts a real Alertmanager 0.28.1 and sends standard grouped notifications to the adapter with a fake UI API; the isolated private-image check runs the adapter against the real UI API without an Alertmanager in that Pod. Production network/ingress/routing and unattended investigation are outside those checks.

Full isolated UI API results: every case and complete returned event metadata
{
  "success": true,
  "image": "nvcr.io/0948643769302270/afoa-release/mosaic-ui:348d4635",
  "nodeVersion": "v22.23.2",
  "rows": [
    {
      "label": "grouped firing",
      "responseStatus": 200,
      "response": {
        "success": true,
        "firing": 2,
        "resolved": 0
      },
      "alerts": [
        {
          "id": "alertmanager:0123456789abcdef",
          "payload": "MosaicWebhookSmokeTest | Isolated synthetic compatibility test | severity=warning",
          "source": "alertmanager",
          "createdAt": "2026-10-29T08:31:25.830Z",
          "stage": "active",
          "labels": {
            "alertname": "MosaicWebhookSmokeTest",
            "severity": "warning"
          },
          "annotations": {
            "summary": "Isolated synthetic compatibility test"
          },
          "status": "firing",
          "startsAt": "2026-10-29T08:31:25.830Z",
          "endsAt": "0001-01-01T00:00:00Z",
          "fingerprint": "0123456789abcdef",
          "generatorURL": "https://example.com/test"
        },
        {
          "id": "alertmanager:fedcba9876543210",
          "payload": "MosaicWebhookSmokeTest | Isolated synthetic compatibility test | severity=warning",
          "source": "alertmanager",
          "createdAt": "2026-10-29T08:31:25.830Z",
          "stage": "active",
          "labels": {
            "alertname": "MosaicWebhookSmokeTest",
            "severity": "warning"
          },
          "annotations": {
            "summary": "Isolated synthetic compatibility test"
          },
          "status": "firing",
          "startsAt": "2026-10-29T08:31:25.830Z",
          "endsAt": "0001-01-01T00:00:00Z",
          "fingerprint": "fedcba9876543210",
          "generatorURL": "https://example.com/test"
        }
      ]
    },
    {
      "label": "repeat delivery",
      "responseStatus": 200,
      "response": {
        "success": true,
        "firing": 2,
        "resolved": 0
      },
      "alerts": [
        {
          "id": "alertmanager:0123456789abcdef",
          "payload": "MosaicWebhookSmokeTest | Isolated synthetic compatibility test | severity=warning",
          "source": "alertmanager",
          "createdAt": "2026-10-29T08:31:25.830Z",
          "stage": "active",
          "labels": {
            "alertname": "MosaicWebhookSmokeTest",
            "severity": "warning"
          },
          "annotations": {
            "summary": "Isolated synthetic compatibility test"
          },
          "status": "firing",
          "startsAt": "2026-10-29T08:31:25.830Z",
          "endsAt": "0001-01-01T00:00:00Z",
          "fingerprint": "0123456789abcdef",
          "generatorURL": "https://example.com/test"
        },
        {
          "id": "alertmanager:fedcba9876543210",
          "payload": "MosaicWebhookSmokeTest | Isolated synthetic compatibility test | severity=warning",
          "source": "alertmanager",
          "createdAt": "2026-10-29T08:31:25.830Z",
          "stage": "active",
          "labels": {
            "alertname": "MosaicWebhookSmokeTest",
            "severity": "warning"
          },
          "annotations": {
            "summary": "Isolated synthetic compatibility test"
          },
          "status": "firing",
          "startsAt": "2026-10-29T08:31:25.830Z",
          "endsAt": "0001-01-01T00:00:00Z",
          "fingerprint": "fedcba9876543210",
          "generatorURL": "https://example.com/test"
        }
      ]
    },
    {
      "label": "mixed status",
      "responseStatus": 200,
      "response": {
        "success": true,
        "firing": 1,
        "resolved": 1
      },
      "alerts": [
        {
          "id": "alertmanager:0123456789abcdef",
          "payload": "MosaicWebhookSmokeTest | Isolated synthetic compatibility test | severity=warning",
          "source": "alertmanager",
          "createdAt": "2026-10-29T08:31:25.830Z",
          "stage": "active",
          "labels": {
            "alertname": "MosaicWebhookSmokeTest",
            "severity": "warning"
          },
          "annotations": {
            "summary": "Isolated synthetic compatibility test"
          },
          "status": "firing",
          "startsAt": "2026-10-29T08:31:25.830Z",
          "endsAt": "0001-01-01T00:00:00Z",
          "fingerprint": "0123456789abcdef",
          "generatorURL": "https://example.com/test"
        }
      ]
    },
    {
      "label": "resolved",
      "responseStatus": 200,
      "response": {
        "success": true,
        "firing": 0,
        "resolved": 1
      },
      "alerts": []
    },
    {
      "label": "repeat resolved",
      "responseStatus": 200,
      "response": {
        "success": true,
        "firing": 0,
        "resolved": 1
      },
      "alerts": []
    }
  ]
}
Exact isolated UI API test harness

The receiver source was mounted as receiver.mjs beside this harness. Its environment supplied separate disposable MOSAIC_UI_MACHINE_TOKEN and MOSAIC_ALERTMANAGER_WEBHOOK_TOKEN values. The isolated UI listened on port 3000 using the matching UI token. No production data/PVCs or Kubernetes service-account tokens were mounted.

import assert from 'node:assert/strict';
import { webhookServer } from './receiver.mjs';

const uiUrl = 'http://127.0.0.1:3000';
const headers = { authorization: `Bearer ${process.env.MOSAIC_UI_MACHINE_TOKEN}`, 'content-type': 'application/json' };
const token = process.env.MOSAIC_ALERTMANAGER_WEBHOOK_TOKEN;
const sleep = ms => new Promise(resolve => setTimeout(resolve, ms));
let ready = false;
for (let n = 0; n < 100; n++) {
  try {
    const response = await fetch(uiUrl + '/api/alerts', { headers });
    if (response.ok && (await response.json()).success) { ready = true; break; }
  } catch {}
  await sleep(500);
}
assert.equal(ready, true, 'isolated UI ingestion API must become available');
const server = webhookServer({ token, uiToken: process.env.MOSAIC_UI_MACHINE_TOKEN, uiUrl });
await new Promise(resolve => server.listen(3004, '127.0.0.1', resolve));
const startsAt = new Date(Date.now() - 1000).toISOString();
const alert = (fingerprint, status = 'firing') => ({ fingerprint, status, labels: { alertname: 'MosaicWebhookSmokeTest', severity: 'warning' },
  annotations: { summary: 'Isolated synthetic compatibility test' }, startsAt,
  endsAt: status === 'resolved' ? new Date().toISOString() : '0001-01-01T00:00:00Z', generatorURL: 'https://example.com/test' });
const rows = [];
async function send(label, alerts, expectedCount) {
  const response = await fetch('http://127.0.0.1:3004/api/alertmanager/webhook', {
    method: 'POST', headers: { authorization: `Bearer ${token}`, 'content-type': 'application/json' },
    body: JSON.stringify({ version: '4', status: 'firing', truncatedAlerts: 0, alerts }),
  });
  const body = await response.json();
  assert.equal(response.status, 200, JSON.stringify(body));
  const listResponse = await fetch(uiUrl + '/api/alerts', { headers });
  const list = await listResponse.json();
  assert.equal(list.success, true);
  assert.equal(list.alerts.length, expectedCount);
  rows.push({ label, responseStatus: response.status, response: body, alerts: list.alerts });
}
try {
  await send('grouped firing', [alert('0123456789abcdef'), alert('fedcba9876543210')], 2);
  await send('repeat delivery', [alert('0123456789abcdef'), alert('fedcba9876543210')], 2);
  await send('mixed status', [alert('0123456789abcdef'), alert('fedcba9876543210', 'resolved')], 1);
  await send('resolved', [alert('0123456789abcdef', 'resolved')], 0);
  await send('repeat resolved', [alert('0123456789abcdef', 'resolved')], 0);
  console.log(JSON.stringify({ success: true, image: process.env.TEST_UI_IMAGE, nodeVersion: process.version, rows }, null, 2));
} finally {
  server.closeAllConnections();
  await new Promise(resolve => server.close(resolve));
}
Complete repository test output: 236 tests, 234 passed, 2 conditional skips, 0 failed

Run with Node 24, the chart test dependencies prepared as in CI (the private research dependency is a render-only placeholder), and ALERTMANAGER_BINARY pointing at the official pinned Alertmanager 0.28.1 binary:

node --test utils/*.test.mjs helm/mosaic-stack/tests/*.test.mjs tests/docs/*.test.mjs
✔ webhook resources are absent by default (105.786084ms)
✔ enabling the webhook packages executable source and independent Secret references (88.987458ms)
✔ optional ingress uses TLS and exposes only the exact webhook route (78.332292ms)
✔ validates UI availability, single-process storage, credentials, ingress, and ports (320.874333ms)
✔ maps full per-alert metadata and honors resolved status inside a firing group (0.902292ms)
✔ bounds the display text while preserving complete annotations and labels (0.176ms)
✔ fails closed on missing/short/equal tokens and invalid UI destinations (0.260875ms)
✔ rejects unauthorized requests and exposes only an unauthenticated health check (31.316875ms)
✔ rejects non-JSON requests and malformed JSON without forwarding (7.010959ms)
✔ validates the entire group before delivery, including identifiers and scalar maps (20.581166ms)
✔ rejects truncated notifications rather than acknowledging unseen alerts (2.116083ms)
✔ limits both Content-Length and chunked bodies to 1 MiB (8.882917ms)
✔ retries upsert the same fingerprint without resetting its investigation stage (5.653667ms)
✔ mixed grouped notifications upsert firing alerts and remove resolved alerts idempotently (19.747459ms)
✔ returns 502 on partial downstream failure and safely accepts a complete retry (8.6065ms)
✔ does not acknowledge login HTML, redirects, suppression, or success:false (6.215625ms)
✔ bounds total downstream delivery time and releases the busy state after a timeout (45.414417ms)
✔ returns retryable backpressure while processing another grouped notification (7.982666ms)
✔ real Alertmanager delivers grouped firing and resolved notifications with bearer authentication (2111.093ms)
✔ recognizes cluster-monitor automation (0.601917ms)
✔ recognizes alert automation (0.048792ms)
✔ does not trust unrelated session names (0.096209ms)
✔ defaults conversations to view and never enables automation edits (0.099958ms)
✔ allows mutations only when editing is enabled and the conversation permits it (0.094875ms)
✔ uses verified Slack identity access without changing non-Slack sessions (0.4095ms)
✔ fails closed when a Slack Edit run has no identity approval broker (0.054542ms)
✔ projects and registers access state in every approval-owning plugin (0.432833ms)
✔ separates sandbox execution from external tool permissions (0.233625ms)
✔ applies explicit per-server MCP tool access without trusting unknown servers (0.131375ms)
✔ View permits dashboard authoring while infrastructure changes still require Edit (0.054708ms)
✔ runs BCM tools locally and sends command execution over SSH (0.412417ms)
✔ does not permit BCM configuration to disable TLS verification (0.675041ms)
✔ keeps readonly and admin CMSH on separate tools and identities (0.152291ms)
✔ requires approval for BCM mutation tools (0.083958ms)
✔ uses per-session Auto to skip BCM approvals without weakening automation (0.060167ms)
✔ retains the authenticated admin backend without readonly fallback (0.067375ms)
✔ startup-loads the Kubernetes approval hook (2.026958ms)
✔ startup-loads one shared diagnostic policy (195.078709ms)
✔ startup removes managed plugins before applying the enabled module set (75.375125ms)
✔ startup-loads the remote SSH approval hook (0.24475ms)
✔ startup-loads the automation read-only guard (193.408042ms)
✔ preserves memory maintenance with embedding-free indexed recall (61.981042ms)
﹣ native memory indexes Markdown and preserves appended notes across restarts (0.059791ms) # SKIP
✔ generates one internal Hardware Agent credential for the service and OpenClaw (59.581375ms)
✔ preserves the secured Hardware Agent image entrypoint (70.021209ms)
✔ wires Hardware Agent BCM lookup through the internal SSH adapter (59.904959ms)
✔ applies global scheduling controls to the Hardware Agent (67.416917ms)
✔ requires the BCM SSH adapter for Hardware Agent installs (66.622958ms)
✔ renders only official OpenClaw and OpenShell runtime images (50.709084ms)
✔ pins every default runtime image (53.079458ms)
✔ renders namespace labeling with the configured kubectl image (62.407542ms)
✔ pins the UI to an immutable commit tag (59.437458ms)
✔ protects the UI with a generated login (53.816459ms)
✔ authenticates internal UI callbacks with one machine credential (181.839167ms)
✔ uses an existing UI auth secret for browser and machine credentials (120.251667ms)
✔ seeds BCM manuals without deployment-time network access (190.620916ms)
✔ uses NMC observability services by default (50.685375ms)
✔ configures the Alertmanager endpoint (117.786208ms)
✔ shares the managed Research Agent corpus with the UI (61.916125ms)
✔ shares the managed Research Agent credential with OpenClaw (60.232416ms)
✔ generates a bearer credential with a chart-managed Research Agent Secret (60.273625ms)
✔ uses one no-reasoning request contract for external endpoints (60.751625ms)
✔ configures chart-managed vLLM for non-thinking generation (176.412542ms)
✔ defaults to a single Super inference pod with one GPU (59.142584ms)
✔ the Ultra profile uses four GPUs in a single inference pod (60.348ms)
✔ rejects removed distributed inference and mismatched GPU allocation (98.988667ms)
✔ renders the OpenShell backend with mTLS under XDG_CONFIG_HOME (51.194334ms)
✔ installs the verified OpenShell package through OpenClaw (59.411083ms)
✔ does not render Slack runtime behavior when disabled (50.137584ms)
✔ renders native Slack Socket Mode with isolated execution (53.228083ms)
✔ restricts Slack channel mentions to approved users (50.700625ms)
✔ configures Slack group DMs, replies, and streaming (51.288458ms)
✔ rejects incomplete native Slack configuration (64.437625ms)
✔ does not render web search when disabled (116.992875ms)
✔ enables key-free DuckDuckGo web search by default (174.957ms)
✔ renders Brave web search when selected (116.934584ms)
✔ web search adds only its own OpenClaw config (176.744167ms)
✔ rejects incomplete web search configuration (131.021208ms)
✔ provides a Slack app manifest for approved DMs and mentions (0.221625ms)
✔ keeps Slack messaging access broad while granting Edit to explicit identities (48.8235ms)
✔ omits Slack identity and exec approval configuration when Slack is disabled (50.163583ms)
✔ validates Slack edit identities without weakening the messaging allowlist (65.193875ms)
✔ patches Slack Enterprise Grid workspace events without weakening app checks (44.213459ms)
✔ exposes only blueprint-owned skills to OpenClaw agents (60.005333ms)
✔ renders named external Kubernetes clusters from Secrets (52.33175ms)
✔ renders read-only Kubernetes RBAC including metrics without Secrets or pod execution (58.68275ms)
✔ renders workload editor and pod exec RBAC for OpenClaw when edit mode is enabled (59.660291ms)
✔ binds an installer-supplied editor role without rendering a duplicate role (59.808708ms)
✔ seeds edit mode and approval settings into Kubernetes and BCM plugins (175.611833ms)
✔ tells the assistant about session access modes only when editing is enabled (101.868417ms)
✔ renders an authenticated BCM admin path only for BCM edit mode (53.717833ms)
✔ keeps the managed BCM adapter read-only when edit mode is disabled (52.280875ms)
✔ exposes edit and HITL state to the UI without changing their defaults (180.346584ms)
✔ validates edit mode dependencies and backend configuration (383.793041ms)
✔ renders constrained remote SSH only in the OpenClaw pod (119.591208ms)
✔ omits remote SSH credentials and plugin files by default (116.5165ms)
✔ omits every terminal workload resource when the module is disabled (51.642458ms)
✔ renders the terminal service from the UI image with isolated read-only access (111.542875ms)
✔ rejects terminal deployments without their required modules (99.790375ms)
✔ omits the cluster monitor when its module is disabled (50.1795ms)
✔ renders a credential-isolated persistent cluster monitor from configured targets (113.371417ms)
✔ validates cluster monitor dependencies and configuration (231.64875ms)
✔ installs a checksum-pinned upstream kubectl binary (58.521458ms)
✔ packages every Kubernetes plugin module into the OpenClaw seed (119.466375ms)
✔ creates the base workspace when every optional skill module is disabled (49.981167ms)
✔ initializes a fresh OpenClaw workspace before seeding it (58.894959ms)
✔ preserves the Slurm evidence apiVersion after its SPDX header (58.739667ms)
✔ shares a generated credential between OpenClaw and the Slurm evidence collector (51.568416ms)
✔ rejects an unregistered default Kubernetes cluster (33.357541ms)
✔ preloads selected skill instructions from their authoritative files (123.55425ms)
✔ packaged product and GPU documentation installs locally with research, UI and BCM disabled (291.028708ms)
✔ workspace and audit storage use the cluster default unless overridden (117.189416ms)
✔ configures optional Loki access without exposing its credential to UI or sandbox (294.537209ms)
✔ external inference can trust a site CA with or without the compatibility proxy (205.083916ms)
✔ configures a tenant-scoped VictoriaMetrics backend and native Grafana datasource (181.136416ms)
✔ deployment reference reflects module settings without exporting credentials (118.871666ms)
✔ allows bounded read-only diagnostic argv (0.669083ms)
✔ rejects mutation flags, arbitrary programs, paths, and shell syntax (0.15875ms)
✔ routes Hardware Agent requests through the hardware backend (0.474208ms)
✔ allows kubectl get pods -A -o wide (1.016417ms)
✔ allows kubectl get deployments -n mosaic (0.063125ms)
✔ allows kubectl describe pod openclaw-0 -n mosaic (0.05025ms)
✔ allows kubectl logs openclaw-0 -n mosaic --tail=100 (0.057167ms)
✔ allows kubectl top pods -A (0.046875ms)
✔ allows kubectl explain deployment.spec (0.032833ms)
✔ allows kubectl api-resources (0.0345ms)
✔ allows kubectl api-versions (0.040166ms)
✔ allows kubectl cluster-info (0.342834ms)
✔ allows kubectl version --client (0.154334ms)
✔ allows kubectl auth can-i get pods (0.0435ms)
✔ allows kubectl auth can-i get secrets (0.029667ms)
✔ rejects kubectl create namespace forbidden (0.139584ms)
✔ rejects kubectl apply -f manifest.yaml (0.025584ms)
✔ rejects kubectl delete pod openclaw-0 (0.023459ms)
✔ rejects kubectl patch deployment openclaw -p {} (0.023084ms)
✔ rejects kubectl exec openclaw-0 -- id (0.018583ms)
✔ rejects kubectl port-forward pod/openclaw-0 8080:80 (0.017042ms)
✔ rejects kubectl get secrets -A (0.023458ms)
✔ rejects kubectl describe secret/api-key (0.021ms)
✔ rejects kubectl config view --raw (0.034ms)
✔ rejects kubectl get pods --server=https://example.invalid (0.03225ms)
✔ rejects kubectl get pods -s https://example.invalid (0.033417ms)
✔ rejects kubectl get pods --kubeconfig /tmp/admin (0.02325ms)
✔ rejects kubectl get pods --cluster admin (0.02175ms)
✔ rejects kubectl get pods --token=secret (0.021167ms)
✔ rejects kubectl get pods --user admin (0.063166ms)
✔ rejects kubectl get pods --as cluster-admin (0.029084ms)
✔ rejects kubectl cluster-info dump (0.032625ms)
✔ rejects kubectl auth reconcile -f rbac.yaml (0.022375ms)
✔ rejects malformed argument arrays (0.05875ms)
✔ exec forwards supplied input only with a client stdin option (1.344542ms)
✔ validates arbitrary exact admin argv and optional stdin (0.537958ms)
✔ keeps mutation out of the read tool and protects admin credentials (0.776459ms)
✔ resolves only registered clusters (0.23575ms)
✔ applies installer-owned API endpoint overrides outside model arguments (0.1495ms)
✔ accepts structured registered cluster settings in the plugin schema (0.297208ms)
✔ blocks kubectl fallback through exec without blocking ordinary shell commands (0.219ms)
✔ registers the exec guard through OpenClaw's trusted tool policy (0.147917ms)
✔ instructs the agent to report a Kubernetes policy denial (0.074792ms)
✔ makes the per-session access mode authoritative (0.142542ms)
✔ separates exec payload options from kubectl connection options (0.21075ms)
✔ returns exit status and redacts the kubeconfig path (540.6525ms)
✔ enforces the combined output limit (9.0475ms)
✔ terminates commands after the configured timeout (21.279625ms)
✔ reports command launch failures (0.607375ms)
✔ passes an apply manifest only over stdin (8.882125ms)
✔ preserves raw command streams and structured execution metadata (0.344875ms)
✔ Loki is opt-in, read-only, and rejects credentials in URLs (0.798209ms)
✔ discovers labels and queries bounded timestamped logs with operator credentials (39.734042ms)
✔ Loki registers without Prometheus and absent endpoints register no query tools (3.567833ms)
✔ numeric LogQL preserves complete labeled samples for instant and range evaluations (7.238834ms)
✔ does not deploy managed MCP servers by default (112.245042ms)
✔ deploys and registers a hardened managed MCP server (156.656666ms)
✔ passes container arguments and secret-backed native MCP headers (194.454375ms)
✔ mounts an optional private CA only in the managed server (128.356917ms)
✔ rejects unsafe or incomplete managed MCP definitions (138.217125ms)
✔ registers installer-defined external MCP without deploying a server or changing the UI (234.420625ms)
✔ rejects unsafe external MCP definitions (178.849916ms)
✔ accepts an in-cluster external MCP service (63.6245ms)
✔ registers Run:ai and external MCP with separate installer policies (66.007416ms)
✔ renews expiring tokens and shares concurrent requests (0.792875ms)
✔ streams MCP responses through the authenticated proxy (24.407375ms)
✔ dashboard_create (prometheus) routes mixed panels to their datasources and rejects failed validation (65.64075ms)
✔ dashboard_create (victoriametrics-metrics-datasource) routes mixed panels to their datasources and rejects failed validation (33.843875ms)
✔ grafana_dashboard_create (prometheus) routes mixed panels to their datasources and rejects failed validation (29.655791ms)
✔ grafana_dashboard_create (victoriametrics-metrics-datasource) routes mixed panels to their datasources and rejects failed validation (26.801917ms)
✔ suppresses a completed replay without persisting arguments (33.809959ms)
✔ suppresses a concurrent replay while the first call is running (10.814083ms)
✔ records failure and rejects reuse with different arguments (39.189541ms)
✔ missing-command results preserve execution evidence and add recovery context (0.718166ms)
✔ other exit statuses and policy denials preserve their existing results (0.07325ms)
✔ resolves configured hosts and preserves exact argv through POSIX quoting (0.195084ms)
✔ accepts explicit shell argv while rejecting a raw command parameter (0.121583ms)
✔ rejects unknown hosts and all connection override fields (0.157834ms)
✔ rejects malformed argv and nested SSH clients (0.099333ms)
✔ builds only fixed hardened SSH connection options (0.717417ms)
✔ redacts all credential paths from process output (0.145292ms)
✔ blocks remote SSH in readonly automation sessions (0.254541ms)
✔ does not deploy Run:ai by default (78.128625ms)
✔ implements Run:ai through the managed MCP module (147.663417ms)
✔ rejects unsafe Run:ai overrides (182.4415ms)
✔ openclaw tool init removes read-only files before rewriting them (86.161417ms)
✔ terminal kubectl init removes the read-only binary before rewriting it (75.677875ms)
✔ Slack approval buttons bind the decision to the requesting Edit user (16.048208ms)
✔ Slack approval broker denies users outside the Edit identity list (0.145292ms)
✔ keeps an approved failed attempt separate from a denied retry (60.5575ms)
✔ authenticates evidence reads and rejects symlink escapes (66.802292ms)
✔ reads only the last MAX_FILE_BYTES of a large file (31.238167ms)
✔ metrics discovery and queries preserve the configured API prefix (root) (17.808583ms)
✔ metrics discovery and queries preserve the configured API prefix (/select/23/prometheus) (6.915083ms)
﹣ VictoriaMetrics serves discovery, MetricsQL instant queries and historical queries (0.072ms) # SKIP
✔ generated selection renders in Helm: {"modules":[]} (75.724833ms)
✔ generated selection renders in Helm: {"modules":["terminal"]} (56.111291ms)
✔ generated selection renders in Helm: {"modules":["slurm"]} (57.45225ms)
✔ generated selection renders in Helm: {"modules":["slurm"],"slurmBackend":"vanilla"} (54.351125ms)
✔ generated selection renders in Helm: {"modules":["alertmanagerWebhook"]} (55.406375ms)
✔ generated selection renders in Helm: {"modules":["kubernetes","prometheus","alertmanager","loki","grafana","bcm","slurm","diagnostics","research","terminal","edit","clusters"]} (68.615542ms)
✔ generated selection renders in Helm: {"inference":"super","modules":["terminal"]} (66.509125ms)
✔ generated selection renders in Helm: {"inference":"ultra","modules":["grafana"]} (70.653875ms)
✔ dependencies are derived without changing requested modules (0.445541ms)
✔ shell metacharacters remain literal arguments and Helm list separators are escaped (9.98125ms)
✔ pinning replaces the development selector (4.829708ms)
✔ Loki connection settings remain optional and use Secret references (13.58175ms)
✔ webhook selection enables a separate receiver with Secret references and no polling URL (8.283875ms)
✔ builder covers every module toggle and observability connection setting (33.369458ms)
✔ each observability service is independently selectable and clears other endpoints (19.567833ms)
✔ Kubernetes quantities preserve decimal and binary units (0.282875ms)
✔ pod scheduling footprint sums app containers and takes init peak before replicas (0.118667ms)
✔ selection deduplicates dependencies and merges shared workload init containers (0.144708ms)
✔ composed footprints match direct Helm renders across inference and module selections (18626.209167ms)
✔ Grafana datasource type preserves the default or configured selection: undefined (8.192667ms)
✔ Grafana datasource type preserves the default or configured selection: victoriametrics-metrics-datasource (6.122292ms)
✔ stdio bridge exposes and invokes tools without a token (153.935041ms)
✔ stdio bridge exposes and invokes tools with a token (148.321667ms)
✔ tracked source files carry the approved NVIDIA SPDX header (3.594416ms)
✔ tracked Markdown outside Helm does not use the cluster manager acronym (1.26875ms)
ℹ tests 236
ℹ suites 0
ℹ pass 234
ℹ fail 0
ℹ cancelled 0
ℹ skipped 2
ℹ todo 0
ℹ duration_ms 19461.042208

@RohanAdwankar

Copy link
Copy Markdown
Member

@coderabbitai review

@coderabbitai

coderabbitai Bot commented Nov 2, 2026 •

Copy link
Copy Markdown
Action not completed

Review rate limited.

Note: CodeRabbit is an incremental review system and does not re-review already reviewed commits. This command is applicable only when automatic reviews are paused.

@coderabbitai

coderabbitai Bot commented Nov 2, 2026

Copy link
Copy Markdown

Warning

Review limit reached

This review ran on the open-source allowance, not this organization's plan, because the pull request author doesn't have an assigned seat. Waiting won't change this — ask an organization admin to assign them a seat, or add seats in Billing if every seat is already assigned, then retry.

Next included review available in 26 minutes.

Check out review usage here.

View limit details

Limit details: You’ve used the included review currently available.

Learn how review limits work.

Review configuration:

Run configuration
  • Source mode: Auto
  • Repository: Unavailable
  • Branch: Unavailable
  • File: No configuration file from this repository applied
  • Commit: Unavailable
  • Layers: Organization UI
  • Review profile: CHILL
  • Plan: Advanced
  • Run ID: 8e46ed3b-c7ee-4e91-80d9-3a60a2e5fddd
Commits

Reviewing files that changed from the base of the PR and between 6b82377 and 70696fb.

Files ignored due to path filters: 1
  • helm/mosaic-stack/files/agent-documentation.tar.gz is excluded by !**/*.gz
Files selected for processing: 12
  • docs/_static/install-builder.mjs
  • docs/alertmanager_webhook.md
  • docs/installation.md
  • docs/installation_builder.md
  • helm/mosaic-stack/files/alertmanager-webhook.mjs
  • helm/mosaic-stack/templates/alertmanager-webhook.yaml
  • helm/mosaic-stack/templates/validation.yaml
  • helm/mosaic-stack/tests/alertmanager-webhook-chart.test.mjs
  • helm/mosaic-stack/tests/alertmanager-webhook.test.mjs
  • helm/mosaic-stack/values.yaml
  • index.md
  • tests/docs/install-builder.test.mjs
  • Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Signed-off-by: Jeff Peng <jepeng@nvidia.com>
@JeffPengCoder
JeffPengCoder force-pushed the feat/alertmanager-webhook branch from 70696fb to c074536 Compare November 19, 2026 06:10

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

[Feature]: Receive Alertmanager webhook notifications in Mosaic

2 participants