AIOps Platform v0.1.0 — deterministic diagnosis + case memory + aiops CLI

Release Notes — v0.1.0

First public release of aiops-platform: deterministic Kubernetes fault
diagnosis with case memory, surfaced through a zero-setup CLI.
Release tag: v0.1.0 · Date: 2026-08-16 · SemVer 2.0.0 · Keep a Changelog 1.1.0.

What this release provides

• aiops CLI — the star experience:
  • aiops diagnose — deterministic rule-based pod diagnosis using the same
    compiled-in rules as the platform (OOM kill → image pull back-off → crash
    loop back-off → pending). Reads ~/.kube/config (or --kubeconfig) and
    scans a namespace; degrades to built-in demo data when no kubeconfig is
    available, so the tool is usable in 60 seconds with zero setup.
  • aiops cases — searches the historical case library; degrades to a
    built-in rule catalog when no --server is reachable.
  • English output, -o json for scripting, exit codes 0 clean / 1
    findings-or-no-results / 2 error.
• Case memory (RAG) — resolved diagnoses are auto-distilled into a
  searchable case library; two-stage retrieval (structured shortlist + optional
  LLM re-rank) recalls past fixes for recurring failures; historical:N
  evidence is citation-verified.
• Deterministic diagnosis engine — evidence timeline
  (finding → evidence → assessment), multi-signal correlation across
  metrics/logs/events; AI only enhances with verifiable citations.
• ControlledOperation operator — pure client-go controller
  (controlledoperations.aiops.platform CRD) for rollout-restart / scale /
  cronjob-suspend: dry-run → confirm → idempotent → audit.
• Web console — multi-cluster observability, incident response, global
  search, 2D authorization (cluster + namespace).

Install the CLI

# Option A — GitHub Release binary (recommended)
#   Download `aiops-<os>-<arch>` for your platform from the v0.1.0 release
#   assets, then verify the SHA256 checksum published alongside it.

# Option B — from source
git clone https://github.com/guiyi-labs/aiops-platform
cd aiops-platform/backend
go install ./cmd/aiops

Note: the repository is a monorepo (backend/ holds the Go module
k8s-aiops.local/backend), so go install ...@latest from the repo root is
not supported. Use the Release binaries or the source path above.

Quickstart

aiops diagnose                    # no kubeconfig? runs built-in demo fixtures
aiops diagnose --kubeconfig ~/.kube/config --namespace prod
aiops cases --query crashloop -o json

Compatibility

Component	Supported	Verified via
Kubernetes	1.34 – 1.36	kind clusters + real-kind E2E (CI)
PostgreSQL	17 (pgvector 0.8.1)	Compose/Helm deploy + backup/restore drills
Deploy paths	Docker Compose / Helm 3 / Kustomize	CI validation + offline install drills
CLI	Linux/macOS (amd64/arm64)	build + unit tests in CI

Project boundaries (honest, unchanged)

• Kubernetes live resources are queried through the API server, never bulk-copied
  into the database.
• Rule-based diagnosis is the deterministic main path; AI only explains with
  checked citations and never mutates the cluster.
• Production OIDC/MFA and organizational HA/WAL-PITR acceptance are not
  claimed in this release; the platform remains day-2 runtime management for
  existing, reachable clusters.
• The cases --server knowledge API endpoint is not yet part of the platform
  HTTP surface in v0.1.0; the CLI degrades to the built-in catalog until the
  endpoint ships.

Release checklist (coordinated by command center)

1. Confirm HEAD = release candidate commit; all gates green
   (go test ./..., coverage ≥ 70%, lint 0, Playwright e2e).
2. Tag v0.1.0 on main and push tag with explicit user authorization.
3. Build multi-platform CLI binaries (GOOS={linux,darwin} GOARCH={amd64,arm64})
   and publish SHA256 sums as release assets.
4. Publish the README's Release-asset download links.
5. Coordinate same-day launch channels (see docs/star-playbook.md §4).

Full change history: CHANGELOG.md and
docs/changes/.

AIOps Platform v0.3.0-rc.4

Full Changelog: https://github.com/guiyi-labs/aiops-platform/commits/v0.3.0-rc.4