neuron-explorer
Steering-first neuron explorer for Qwen3.5-4B: push every MLP neuron, keep every rollout, browse the results
翻译准备中
The complete open-source platform for on-call management, incident response, and status pages. Self-hosted, extensible, and built for DevOps & SRE teams.
翻译准备中
取自仓库 README
正在加载完整 README。 以下为已保存的摘录。
Transparent, self-hosted incident operations.
From alert ingestion and on-call routing through response, customer communication, and learning.
Target Release Window: Late this month / First week of next month
Over 260+ PRs have been merged since our last release (v1.4.0). While we initially planned a minor v1.5, the massive improvements across the board—a completely redesigned modern UI, rock-solid availability, scalable architecture, and deep engine upgrades—mean our next milestone will be OpsKnight 2.0!查看完整 README ↗Website · Documentation · Quick start · Watch product tour (3:34) · What's new in 2.0 · Sponsor
See how OpsKnight connects alert ingestion, incident response, on-call, escalation, ChatOps, customer communication, analytics and post-incident learning.
Authenticated inbound integrations with service-bound routing, provider-specific urgency mapping, deduplication and recovery handling.
Unified incident workspace with ownership handoffs, real-time activity timelines, response health SLA timers, action items, and 5-Whys postmortems.
Multi-layer rotation schedules with DST-safe handoffs, temporary overrides, and escalation policies targeting users, schedules, and teams.
Multi-channel paging across Web Push, email, SMS, WhatsApp, and interactive Twilio voice with durable retries and delivery evidence.
One configurable public or private status page with service health, incidents, maintenance, subscriber notifications, uptime history and publishable postmortems.
Actionable reliability metrics including MTTA, MTTR, SLA compliance tracking, configurable dashboards, NOC/TV mode, and browser print/PDF export.
Prerequisites: Git, Docker with Docker Compose, and openssl.
git clone https://github.com/opsknight-labs/OpsKnight.git
cd OpsKnight
cat << 'EOF' > .env
OPSKNIGHT_IMAGE=ghcr.io/opsknight-labs/opsknight:2.0.0
POSTGRES_USER=opsknight
POSTGRES_DB=opsknight_db
POSTGRES_PORT=5432
APP_PORT=3000
NEXTAUTH_URL=http://localhost:3000
NEXT_PUBLIC_APP_URL=http://localhost:3000
EOF
printf 'POSTGRES_PASSWORD=%s\n' "$(openssl rand -hex 32)" >> .env
printf 'NEXTAUTH_SECRET=%s\n' "$(openssl rand -base64 32)" >> .env
printf 'API_KEY_SECRET=%s\n' "$(openssl rand -base64 32)" >> .env
printf 'ENCRYPTION_KEY=%s\n' "$(openssl rand -hex 32)" >> .env
docker compose --env-file .env -f deploy/compose/docker-compose.yml pull
docker compose --env-file .env -f deploy/compose/docker-compose.yml up -d --wait
docker compose --env-file .env -f deploy/compose/docker-compose.yml exec -T opsknight-app \
node scripts/create-bootstrap-code.mjs
Open http://localhost:3000/setup, enter the short-lived bootstrap code, and create the first administrator.
[!IMPORTANT] Before exposing OpsKnight, configure its public URL and keep
NEXTAUTH_SECRET,API_KEY_SECRETand the encryption key outside source control. They must stay stable across restarts, upgrades and restores — losing the encryption key makes stored provider credentials unreadable. Production startup rejects placeholder and reused secrets.
Production installation → · Upgrading from 1.x →
Command Center gives responders real-time operational pulse across active incidents, open alerts, team workload, and approaching SLA breaches.
Responders get ownership, notes, watchers, action items, SLA timers and a live incident timeline in one workspace.
Manage rotation layers, live coverage, DST-safe handoffs, temporary overrides and multi-tier escalation.
Publish service health, incidents, maintenance, uptime history and post-incident reviews from the same incident workflow.
Install OpsKnight from the browser to get a mobile-first workspace for incidents, on-call, escalation policies, services, teams, status, analytics and postmortems. Push registration is per device, the app follows the system light or dark theme, and offline actions stay authorization-bound when they replay. Set up the mobile PWA →
🧱 Split production runtime Web, Scheduler, General / Critical / Bulk Workers and Status Projector scale and fail independently. The integrated runtime stays supported.
📣 Notification control plane Durable delivery intents, retries, provider capacity and per-attempt evidence for email, Web Push, SMS, WhatsApp and voice.
📞 Twilio voice paging Triggered incidents can call responders, who acknowledge from the keypad.
💬 Microsoft Teams ChatOps Adaptive Cards, interactive incident actions, war rooms and meetings — alongside reworked Slack war rooms and Jira sync.
🛡️ Identity and governance SCIM 2.0 Users and Groups, OIDC claim-to-role mapping, an Auditor role, session registry and privacy/DSAR workflows.
🐳 Docker Swarm and deployment planning Integrated and split stacks for Compose, Swarm, Helm and Kustomize, with workload profiles and repeatable load/correctness testing for deployment planning.
Monitoring, cloud, uptime, CI/CD, ChatOps, ticketing, notification and identity providers each have a documented contract for authentication, lifecycle actions and recovery. Browse the integration catalog →
| Topology | Best for | Guide | |
|---|---|---|---|
| Integrated or split, single host | Evaluation and small teams | Docker Compose → | |
| Integrated or split across nodes | Multi-node Docker estates | Docker Swarm → | |
| Schema-validated chart | Production Kubernetes | Helm → | |
| Bases and overlays | GitOps with Argo CD or Flux | Kustomize → |
PostgreSQL 14 or later is required. For production, pin the tested multi-architecture image digest, size the database connection budget and complete the topology's acceptance checklist. Choose a topology → · Plan capacity →
Integrated mode runs Web and background work in one process. Split mode gives Web, Scheduler, General Worker, Critical Worker, Bulk Worker and Status Projector explicit ownership so they scale and fail independently. Both use PostgreSQL for durable state and work coordination; only Web receives ingress. Read the architecture guide →
| OpsKnight | Typical per-seat SaaS | |
|---|---|---|
| Where it runs | Your infrastructure | Vendor cloud |
| Incident data | Stays in your PostgreSQL | Stored by the vendor |
| Pricing | No seat meter, no software fee | Per-user plans |
| Source | Open under AGPL-3.0-only | Closed |
| Operations | Source-visible runtime, health, metrics and delivery evidence | Vendor-operated runtime |
OpsKnight is an independent project and is not affiliated with PagerDuty, Opsgenie or other vendors.
Encrypted provider credentials with key rotation, independent session and API-key signing secrets, fail-closed inbound verification, role-based authorization, session revocation, OIDC, SCIM, audit evidence and CI security scanning. Compliance tooling helps operators implement and evidence controls; it does not itself confer certification.
Security policy · Production hardening · Report a vulnerability privately
The versioned 2.0 documentation is the source of truth for product behavior, configuration, deployment, integrations and operations.
| Get started | Guides | Operate | API reference | Troubleshooting |
|---|---|---|---|---|
| Install and first incident | Day-to-day workflows | Deploy, scale, upgrade | REST API and keys | Diagnose and recover |
Bug reports, focused feature requests, documentation fixes and code contributions are welcome.
OpsKnight is independently maintained. Sponsorship funds open-source development, release infrastructure, security maintenance and documentation.
OpsKnight 2.0 is distributed under AGPL-3.0-only. OpsKnight 1.4.0 and earlier remain available under the licenses they were published with; see LICENSE-TRANSITION.md.