Skip to content

v1.15.0

Latest
Latest

Choose a tag to compare

View all tags
@github-actions github-actions released this 14 Jun 00:46
· 4 commits to main since this release
v1.15.0
a4a4ffb
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

KubeBolt 1.15.0 — Enterprise foundation + Kobi grows up

1.15 lays the foundation for KubeBolt SaaS / Enterprise while staying a
clean, behavior-compatible upgrade for OSS. The headline is W1 — the
store-interface seams and the now-materialized organization → team → user
hierarchy — alongside a big round of Kobi maturation (a read-only MCP
server, conversation history, dry-run safety, and an in-vivo reliability
pass), first-class Cilium NetworkPolicy support, and a ground-up Overview
/ Capacity dashboard redesign.

Backend ships as KubeBolt 1.15.0. The agent is unchanged this cycle — it
stays at kubebolt-agent 1.1.0 (no agent image or chart bump). See the
compatibility matrix for supported pairings.

No breaking changes. Clean helm upgrade from 1.14.x. OSS behavior is
unchanged except that the org → team → user hierarchy is now materialized
(every user becomes a real member of the auto-seeded default team on first
boot — idempotent, additive) and the single-org/single-team boundary is now
enforced server-side. New BoltDB buckets are created automatically. The
bundled VictoriaMetrics moves to the -scratch image (see Security).

Enterprise / SaaS foundation (W1)

The plumbing a multi-tenant SaaS / Enterprise build needs, landed as no-op seams
in single-tenant OSS so the boundary is real before the EE features exist.

  • Multi-tenancy substrate — tenant-scoped runtime: tenant context on every
    request, per-cluster runtime keys, connection-pool eviction, and tenant-scoped
    WebSocket fan-out. Inert in OSS (one implicit default tenant); the layer EE
    builds on. (#77)
  • API service tokens — kbs_ service tokens (default Autopilot scopes,
    reach the read-only MCP endpoint out of the box) and kbk_ API keys
    (fail-closed, no default scopes). The auth surface for headless automation and
    the MCP server. (#78)
  • W1 store interfaces + identity model — the org → team → user hierarchy is
    now materialized in OSS (every user is a real member of the auto-seeded
    default team; single-org / single-team enforced server-side), backed by the
    store-interface seams (User / Tenant / Team / Cluster / Token / RefreshToken
    stores) and a no-op UsageStore metering seam. Administration is
    reorganized into domain hubs — Access, Agents & Ingest, AI (Kobi), System
    — plus standalone API Tokens. (#90)

Kobi grows up

  • Read-only MCP server — drive Kobi's investigation tools (overview,
    resources, YAML, describe, logs, events, insights, topology, metrics) from any
    MCP host (Claude Code, Cursor, a CI step, another agent) over Streamable
    HTTP     or stdio. 17 read tools, read-only enforced server-side (no
    mutating verb is reachable even by name), Secret redaction inherited. BYO host
    LLM. (#83, #87)
  • Conversation history — Kobi conversations persist per user and resume
    across refresh / re-login. (#81)
  • Usage analytics — per-user and per-trigger cost breakdown plus a
    reliability strip in the admin usage view. (#84)
  • Action safety — every propose_* action now renders a dry-run preview
    (dryRun=All server-side) in the proposal card before you confirm; a
    configurable action-progress timeout with auto-root-cause when an
    executed action stalls (e.g. a scale blocked by a ResourceQuota); and a
    Stop button to cancel a streaming turn with latency hints. (#85, #86, #88,
    #89)
  • Reliability round (in-vivo) — the multi-step loop's round budget is now
    configurable (KUBEBOLT_AI_MAX_ROUNDS, default 20) with a graceful close
    turn + a Continue control instead of a dead error on exhaustion;
    propose_debug_pod takes a non-interactive command whose output Kobi reads
    back from logs; and new reasoning rules make Kobi walk the dependency graph
    (Service endpoints → pod readiness → logs) before blaming the network, and
    recognize selectorless / external dependencies instead of guessing a
    workload is missing. (#91)
  • Prompt / UX quick-wins — per-action RBAC vs governance error clarity,
    real-time fallback-model transparency, an explicit hallucination guard,
    tool-error handling guidance, and persisted per-message timestamps. (#82)

Cilium L3–L7 policies

CiliumNetworkPolicy and CiliumClusterwideNetworkPolicy (cilium.io/v2)
are now first-class managed resources — the L3-L7 layer a standard
NetworkPolicy can't express and that KubeBolt previously only saw as Hubble
metrics. List + detail with a rendered rule breakdown (peers / ports / L7
http·dns·kafka), a matched-pods tab, YAML view/edit/delete, + New
starter templates, and Kobi can now read a Cilium policy to verify whether
an L7 deny actually blocks a flow. Present only on Cilium clusters (graceful
empty everywhere else). (#92)

Overview & Capacity dashboard redesign

  • Overview KPIs become ring gauges + legend rows so the dashboard reads as
    one family; the previously-invisible Degraded bucket (Pending /
    CrashLoopBackOff / partially-ready) is surfaced via a new degraded
    pseudo-status, with filtered deep links (?status=, ?severity=) so a KPI
    links to a list that matches its count. Usage cards collapse into a graded
    Used donut + bullet bar; Workload Health gains a Needs attention section;
    namespace tiles segment ready / not-ready. (#97)
  • Capacity trends gain cluster-wide request / limit / capacity reference
    overlays (capacity ships hidden behind a pill), with toggle state persisted.
    (#97)
  • The top limited-access banner is now dismissible (persisted per
    cluster + access shape, re-appears on an RBAC change). (#100)

Fixes & security

  • Bundled VictoriaMetrics + vmagent → v1.145.0-scratch. The metrics images
    move to VictoriaMetrics' application-binary-only scratch variant — no OS
    packages, no OpenSSL — clearing the Go-stdlib and OpenSSL (CVE-2026-42504,
    CVE-2026-45447) findings on both the release Trivy gate and the public
    Artifact Hub security scan, with no suppression. Drop-in (same entrypoint +
    flags), verified end to end.
  • Configurable informer cache-sync timeout (KUBEBOLT_CACHE_SYNC_TIMEOUT_SECONDS,
    default 45s; runtime-overridable in Settings → General) for large clusters
    that flake on first connect. (#79)
  • endpoints resolves by both keys — the EndpointSlice name and the
    fronting Service name — closing a 404 that read downstream as "the endpoint
    was deleted" and steered RCA toward recreating it. (#96)

Upgrade

helm upgrade kubebolt oci://ghcr.io/clm-cloud-solutions/kubebolt/helm/kubebolt --version 1.15.0

Preserve your overrides by capturing them and re-applying with -f (do not
use --reuse-values — it carries the old chart's bundled-image defaults
forward, keeping the previous VictoriaMetrics image):

helm get values kubebolt -n kubebolt-system -o yaml > values.yaml
helm upgrade kubebolt oci://ghcr.io/clm-cloud-solutions/kubebolt/helm/kubebolt \
  -n kubebolt-system --version 1.15.0 -f values.yaml

The agent is unchanged — it stays at kubebolt-agent 1.1.0. New BoltDB
buckets auto-create and the org → team → user hierarchy materializes
idempotently on first boot. No configuration changes required.

Install

Helm (recommended for Kubernetes)

helm install kubebolt oci://ghcr.io/clm-cloud-solutions/kubebolt/helm/kubebolt
kubectl port-forward svc/kubebolt 3000:80

Single Binary (download below)

# macOS Apple Silicon
curl -LO https://github.com/clm-cloud-solutions/kubebolt/releases/download/v1.15.0/kubebolt-darwin-arm64
chmod +x kubebolt-darwin-arm64 && mv kubebolt-darwin-arm64 /usr/local/bin/kubebolt

# Linux amd64
curl -LO https://github.com/clm-cloud-solutions/kubebolt/releases/download/v1.15.0/kubebolt-linux-amd64
chmod +x kubebolt-linux-amd64 && sudo mv kubebolt-linux-amd64 /usr/local/bin/kubebolt

kubebolt --kubeconfig ~/.kube/config

All platform binaries are attached below. Verify with sha256sum -c CHECKSUMS.txt.

Docker (single container)

docker run -p 3000:3000 -v ~/.kube:/root/.kube:ro \
  ghcr.io/clm-cloud-solutions/kubebolt:1.15.0

Homebrew (macOS, Linux)

brew install clm-cloud-solutions/tap/kubebolt

kubectl krew plugin

kubectl krew index add clm https://github.com/clm-cloud-solutions/krew-index.git
kubectl krew install clm/kubebolt
kubectl kubebolt

Docker Compose

git clone https://github.com/clm-cloud-solutions/kubebolt.git
cd kubebolt/deploy && docker compose up -d

Container Images

  • Single container (binary + frontend): ghcr.io/clm-cloud-solutions/kubebolt:1.15.0
  • API only: ghcr.io/clm-cloud-solutions/kubebolt/api:1.15.0
  • Web only: ghcr.io/clm-cloud-solutions/kubebolt/web:1.15.0

Changelog

Features

  • feat(web): make the limited-access banner dismissible ()
  • feat(web): request/limit/capacity overlays on Capacity trends ()
  • feat(api+web): degraded pseudo-status and filtered deep links ()
  • feat(web): redesign Overview dashboard with gauges and legend rows ()
  • feat(web): Cilium policies UI + grouped/complete + New picker ()
  • feat(cluster): CiliumNetworkPolicy + CiliumClusterwideNetworkPolicy as first-class resources ()
  • feat(web): step-limit notice + Continue control, Max-steps setting ()
  • feat(copilot): configurable tool-call rounds + graceful close, debug_pod command ()
  • feat(copilot): Kobi reasoning rules + max-rounds close directive ()
  • feat(web): surface the org → team → user hierarchy in the UI ()
  • feat(auth): W1 — operationalize the org→team→user hierarchy in OSS ()
  • feat(api): W1 — wire the UsageStore metering seam end to end ()
  • feat(usage): W1 — UsageStore metering seam (no-op in OSS) ()
  • feat(auth): W1 — RefreshTokenStore + APITokenStorer interfaces (token seams) ()
  • feat(cluster): W1 — ClusterStore interface (cluster persistence seam) ()
  • feat(auth): W1 — UserStore interface (User seam) ()
  • feat(auth): W1 — TenantStore interface (Organization seam) ()
  • feat(auth): W1 — TeamStore + Team/Membership identity model ()
  • feat(copilot): automatic dry-run preview on Kobi action proposals ()
  • feat(copilot): cancel in-flight turn + streaming latency feedback ()
  • feat(copilot): configurable action-progress timeout + auto-root-cause on stall ()
  • feat(copilot): Kobi Usage breakdown — per-user / per-trigger + reliability strip ()
  • feat(mcp): read-only Kobi MCP server (HTTP + stdio transports) ()
  • feat(copilot): autofocus input on new conversation + timestamps in export ()
  • feat(copilot): persist per-message timestamps for the conversation timeline ()
  • feat(copilot): cross-reference action audit records to their conversation ()
  • feat(copilot): prominent fallback badge + name the user's role on a 403 ()
  • feat(copilot): prompt — tool errors are cluster signal + no hallucinated resources ()
  • feat(copilot): Kobi conversation history + resume UI ()
  • feat(copilot): persist Kobi conversations — per-user history + resume (backend) ()
  • feat(ws): scope broadcasts by (tenant,cluster) — A.4 OSS seam ()
  • feat(cluster): A.3d — parked-runtime pool eviction (idle + cap) ()
  • feat(cluster): A.3c — pool-based instant switch (park + promote) ()
  • feat(cluster): lazy-spin pooled runtimes with single-flight (W2 A.3b) ()
  • feat(cluster): resolve getters through a per-(tenant,cluster) runtime (W2 A.3a) ()
  • feat(cluster): thread request ctx into Connector/Collector/Engine (W2 A.2) ()
  • feat(cluster): thread (tenant,cluster) RuntimeKey through the request (W2 A.1) ()
  • feat(auth): REST API tokens — service (kbs_) + customer keys (kbk_) ()
  • feat(auth): tenant-context seam for multi-tenant (W0) ()

Fixes

  • fix(api): resolve by both slice name and service name ()
  • fix(api): create-from-manifest registries were out of sync with the GVR map ()
  • fix(mcp): review fixes — token scope, Content-Type, panic recovery, packaging ()
  • fix(copilot): stop resumed/historical proposals from re-investigating a stall ()
  • fix(copilot): stick with the fallback provider for the rest of the session ()
  • fix(copilot): fall over to the secondary provider on a 404 ()
  • fix(copilot): scope Kobi conversations per cluster, not just per user ()
  • fix(copilot): record token usage for every LLM call — no untracked spend ()
  • fix(copilot): isolate Kobi conversations per user across login/logout ()
  • fix(security): bump VictoriaMetrics + vmagent to v1.145.0-rc0 (CVE-2026-42504) ()
  • fix(cluster): configurable informer cache-sync timeout (default 45s) ()
  • fix(insights): gate parked-runtime notifications to active cluster (temporary) ()
  • fix(cluster): gate WS broadcasts of parked pool runtimes ()
  • fix(cluster): stale overview after switch + cache slow ServerVersion ()

Documentation

  • docs(auth): document why hashToken uses SHA-256, not bcrypt ()
  • docs(incident-sim): Kobi incident-simulation lab — scenarios + 3-tier cascade app ()
  • docs(mcp): make the service-token requirement explicit + record live verification ()
  • docs(mcp): add live-server verification plan to the Kobi MCP guide ()
  • docs: design for read-only multi-tenant Kobi MCP server ()

Other

  • Merge pull request #102 from clm-cloud-solutions/release/1.15.0-ga ()
  • Merge remote-tracking branch 'origin/main' into release/1.15.0-ga ()
  • Merge pull request #101 from clm-cloud-solutions/chore/1.15.0-release ()
  • chore: 1.15.0 release bump (GA) ()
  • Merge pull request #100 from clm-cloud-solutions/feat/dismissible-limited-access-banner ()
  • Merge pull request #99 from clm-cloud-solutions/chore/1.15.0-rc.3-release ()
  • chore: 1.15.0-rc.3 release bump ()
  • Merge pull request #98 from clm-cloud-solutions/chore/1.15.0-rc.2-release ()
  • chore: 1.15.0-rc.2 release bump ()
  • Merge pull request #97 from clm-cloud-solutions/feat/dashboard-metrics-improvements ()
  • Merge pull request #96 from clm-cloud-solutions/feat/endpoints-lookup-both-keys ()
  • Merge pull request #95 from clm-cloud-solutions/chore/1.15.0-rc.1-release ()
  • chore: 1.15.0-rc.1 release bump ()
  • Merge pull request #94 from clm-cloud-solutions/feat/kobi-incident-sim-lab ()
  • Merge pull request #92 from clm-cloud-solutions/feat/cilium-network-policies ()
  • Merge pull request #91 from clm-cloud-solutions/feat/kobi-reliability-invivo ()
  • Merge pull request #90 from clm-cloud-solutions/feat/w1-store-interfaces ()
  • refactor(web): group Administration into domain hubs ()
  • refactor(auth): extract IngestTokenStore from inlined Tenant.IngestTokens ()
  • Merge pull request #87 from clm-cloud-solutions/claude/kobi-mcp-readonly ()
  • Merge pull request #89 from clm-cloud-solutions/feat/kobi-dryrun-preview ()
  • Merge pull request #88 from clm-cloud-solutions/fix/kobi-stall-refire ()
  • Merge remote-tracking branch 'origin/develop' into fix/kobi-stall-refire ()
  • Merge pull request #86 from clm-cloud-solutions/feat/kobi-streaming-cancel ()
  • Merge pull request #85 from clm-cloud-solutions/feat/kobi-action-progress-timeout ()
  • Merge pull request #84 from clm-cloud-solutions/feat/kobi-usage-breakdown ()
  • Merge pull request #83 from clm-cloud-solutions/claude/kobi-mcp-feasibility-vPSZS ()
  • Merge pull request #82 from clm-cloud-solutions/feat/kobi-quick-wins ()
  • test(copilot): update ActionProposalCard mocks for audit object + useAuth ()
  • Merge pull request #81 from clm-cloud-solutions/feat/kobi-conversation-history ()
  • Merge pull request #79 from clm-cloud-solutions/fix/cache-sync-timeout ()
  • Merge pull request #80 from clm-cloud-solutions/fix/trivy-vm-cve ()
  • Merge pull request #78 from clm-cloud-solutions/feat/api-service-tokens ()
  • Merge remote-tracking branch 'origin/develop' into feat/api-service-tokens ()
  • Merge pull request #77 from clm-cloud-solutions/feat/multitenant ()
  • Merge origin/develop into feat/api-service-tokens ()
  • style(copilot): bump chat type ramp one step for readability ()
  • Merge remote-tracking branch 'origin/develop' into feat/multitenant ()
Assets 21
Loading