Changelog
This page documents release versions, feature highlights, and fixes for the CollectorCtrl platform.
[Unreleased] — Telemetry Governor: Recurring Pattern Analysis
Not yet in a public release
These changes are merged into the documentation ahead of the next release. Until you upgrade, v0.6.0-beta and earlier still keep errors at 100% and show volume estimates in Telemetry Insights.
Highlights
- Telemetry Insights ranks recurring patterns instead of estimating volume. Counts are the records Governor received. Projected volumes, error estimates, confidence margins and monthly savings are removed.
- One sampling rate for every severity. Warnings, errors and error spans are sampled like everything else, so pattern counts compare fairly across severities. Spans are sampled by trace ID, so a sampled trace stays complete.
- Default tap rate is 10%. Existing taps keep their configured rate.
Telemetry Insights
- Rankings read every record in the selected time range instead of the newest 5,000, with exact received counts.
- Log patterns are grouped by service and severity, and keep URL and file paths (
/api/healthand/api/paymentstay separate) while masking IDs. - Trace routes are one group per service and route, with errors counted inside, so error rate and latency describe the whole route.
- Suggestions say "review", need at least 30 received records, and never target warnings, errors or unknown log severities. Routes with errors are suggested only for tail sampling.
- The rule preview states when it replays only the newest 5,000 records, and no longer reports a monthly projection.
MCP and Copilot
governor_top_patternssuggestions follow the Insights rules;governor_optimize_previewno longer returns a monthly projection.
Upgrade Notes
- Collectors switch to uniform sampling when their supervisor is upgraded and their configuration is recomposed. Until the whole fleet is updated, older hosts still send errors at 100%, and their errors look inflated in Insights.
- S3 and OTLP destinations now receive a sample of errors, not every error.
[v0.6.0-beta] — 2026-10-09 — Latest Public Release — Kubernetes Fleet Visibility (pilot)
Upgrade the server and observer together
Deploy the v0.6.0-beta server and the matching Kubernetes observer image together. See Kubernetes Workload Observation for installation and upgrade steps. Release notes and downloads.
Highlights
- One Fleet Inventory, clearer resource views: Switch between All, Host collectors and Kubernetes workloads. Kubernetes views add cluster and namespace filters; the host view hides Kubernetes-only columns.
- Kubernetes workload details: Observed ConfigMap reference, configuration hash, optional configuration content, pod instances and available source/ownership information. Kubernetes nodes are shown as pod placement, not as managed hosts.
- Pod lifecycle visibility: Pod identities are tracked through replacements. Terminating and completed pods are distinguished, and completed pods are hidden by default.
- Explicit observation freshness: Old or disconnected observations are shown as historical, with freshness based on the reporting interval. Kubernetes readiness is not presented as proof of telemetry delivery.
- Kubernetes-aware Copilot and MCP: Responses distinguish observed workloads from managed hosts, and host-only actions are re-checked when an approved action runs.
- Updated onboarding: Guidance for enrollment Secrets, trusted server connections, namespace-scoped installs and OpenShift.
Management Safeguards
- Observed Kubernetes workloads are excluded from host configuration policies, canaries, Supervisor upgrades, runtime templates and telemetry taps.
- Restart, command, asset deployment and instrumentation operations reject observed workloads. Mixed bulk operations list unsupported targets instead of reporting success.
- Kubernetes deployment and configuration changes stay with your existing deployment system.
Security & Reliability Fixes
- Observer connections are restricted to the administrator-configured server endpoint; cross-namespace authentication Secret references are rejected.
- Enrollment with persistent per-workload credentials, secure reconnect/restart and an explicit legacy migration path. Revoked credentials never fall back to enrollment.
- Removed fixed inventory scan limits that could omit records from UI and MCP results; filtering and counting now happen before pagination.
- Delayed, replayed or far-future observations can no longer replace newer evidence.
- Fixed first-connection identity handling and ConfigMap metadata persistence.
Observer Upgrade Requirements
- Create an Agent Enrollment token and store it in a Secret (key
enrollment-token) referenced byopamp.enrollmentSecret. - Provide persistent credential storage (retained PVC or
credentialStorage.existingClaim) and keep it across upgrades. - Run one observer replica.
- Use a trusted
wss://endpoint inopamp.serverand in every monitor; addopampCAConfigMapfor a private CA. - Shared-secret installs: complete the legacy migration before switching to enrollment.
Pilot Limitations
- Discovery covers Deployments, DaemonSets and StatefulSets with mounted or projected ConfigMaps. Injected sidecars, native OpenTelemetry Operator resources, Secret-backed configuration and OpenShift DeploymentConfig are not yet covered.
- Live validation on Kubernetes, OpenShift, GKE, AKS and EKS is outstanding.
- No collector metrics collection, Git pull-request workflow or emergency direct writes for workloads.
- Configuration upload is opt-in; keep it off for configurations with inline secrets.
[v0.5.5-beta] — 2026-10-01
Highlights
- Sidecar Collectors Visible Across the Fleet: The table, honeycomb and topology views show sidecar status (Running, Starting, Stopped, Error, Unknown) at a glance, with Sidecar issues and Has sidecar quick filters. Fleets without sidecars see no change.
- Fully Isolated Sidecar Start/Restart: Starting or restarting a sidecar no longer makes the main collector look unhealthy. A new Starting state covers the first 20 seconds while the sidecar's health endpoint comes up.
- Operations Toolkit for AI Agents: 32 new MCP tools and 4 runbook prompts. Agents can track their proposals to a result, discover configuration templates, read config history with diffs, and change many collectors through policies or GitOps pull requests.
- ZeroTouch Rebuilt for Large Fleets: One row per application instead of one per process, attention tabs, filter search, and bulk instrument/remove with a preview.
- Tighter App Discovery: Turning discovery off really stops new scans, a node's own
supervisor.yamlsetting is respected, and every toggle is audited.
Fixes
- Notification View Details links open the right page (Git sync alerts, collector events, instrumentation).
- The collector page keeps the Sidecar tab in the URL across refreshes and shared links.
- Hosts without a sidecar show Disabled instead of Stopped.
Upgrade Notes
- Deploy the updated supervisor together with the server.
POST /api/governor/supervisors/configurenow returns per-instance results and non-200 codes when nothing was applied.- Alert rules matching the old text "Collectors are not healthy: …" need updating; the top-level health error now describes the main collector only.
[v0.5.3-beta] — 2026-09-30
Highlights
- Streaming Copilot: Answers and tool steps stream live, with slash commands,
@mentions, syntax-highlighted YAML and a dockable panel. - Fleet Classification Redesigned: Environment and role rules show a live preview of matching collectors before you save. Rules can tag collectors as agent, gateway or sidecar, and policies can target classifications.
- Least-Privilege Access Control: A grant ceiling means you can only grant permissions you hold yourself, across roles, users, API tokens and SSO mappings.
- Tamper-Evident Audit Trail: Hash-chained entries with Verify integrity, server-side search across the full history, shareable filter links and CSV export.
- Notifications: Per-user read state, PagerDuty routing keys and grouped repeat alerts.
- Guided SSO Setup: Presets for Okta, Entra ID, Google Workspace, Keycloak and Auth0, plus a test sign-in that shows the identity, groups and resulting role.
- Independent Main and Sidecar Lifecycle: Maintain one collector without interrupting the other, with a new Restart Both action.
- GitOps with GitHub: Export policies to GitHub, sync automatically, review changes as pull requests, and see drift side by side.
Upgrade Notes ⚠️
- The built-in viewer role no longer includes the user directory or the audit log.
- Git webhooks now require a secret; set the webhook up again from Settings → Git Integration.
- The sidecar needs its own collector executable. Deploy the updated supervisor with the server and UI.
[v0.5.2-beta] — 2026-09-29
Highlights
- Canary Rollouts You Can Trust: Canaries survive server restarts, Abort never releases the new configuration to the whole fleet, and rollback restores the complete policy, including package targets.
- No Silent Config Loss: A save reports success only once it is stored, and the UI shows separately whether it was delivered and confirmed by the collector.
- HTTPS by Default: The UI and API use HTTPS out of the box, and the server refuses to start with a broken certificate instead of falling back to plain HTTP.
- One-Command Backup and Restore:
-backupcaptures the database, keys, certificates and state while the server runs;-restorebrings it back on a new server so collectors reconnect without re-enrolling. - ZeroTouch Hardened: Collectors act only on what they discovered themselves, services are never restarted without consent, and Windows apps are relaunched as the same user.
- Clearer Admin Screens: Redesigned Access Tokens (previously API Tokens) and AI Settings pages.
Upgrade Notes ⚠️
- The UI moves to
https://. Update bookmarks, health checks and reverse proxies; plain HTTP must be requested explicitly. - Docker Compose needs its secrets set.
- Canary rollouts started before the upgrade are held; save them again as a new rollout.
[v0.5.1-beta] — 2026-09-28 — Secure by Default
Highlights
- Per-Agent Enrollment: Every agent gets its own credential from an enrollment token (with expiry and usage limits). Credentials can be revoked, and impersonation attempts raise a notification.
- Encrypted Connections by Default: Each server creates its own certificate authority; install scripts use
wss://and verify the server certificate. - Signed Packages: Collector and sidecar packages must be signed in the UI before agents install them; supervisor updates install only with the release signature.
- Collectors Survive Bad Configurations: The supervisor validates every config before applying it and rolls back to the last known good config on crash loops.
- Configuration History: Every replaced config with who, when and why, side-by-side compare, and one-click restore.
- Safer AI Approvals: AI assistants can't approve their own changes, actions run only once, and an optional four-eyes setting stops self-approval.
Breaking Changes ⚠️
- Agents must enroll. Shared-secret agents are rejected unless the server runs with
OPAMP_LEGACY_MODE=trueduring migration. - Agents that trusted the certificate built into earlier releases must be re-onboarded.
- Install this supervisor by hand once; later supervisor updates roll out from the UI.
- Licenses must be re-issued, and everyone signs in again once.
[v0.5.0-beta] — 2026-09-23
Highlights
- Telemetry Governor Tap: Much lighter, safer and self-healing.
[v0.4.9-beta] — 2026-09-22
Highlights
- Telemetry Governor Dynamic Multi-Destination Routing: Full support for fan-out and selective routing of tapped telemetry streams across multiple Amazon S3 archives and custom OTLP receivers without affecting the primary ingestion pipeline.
- Interactive Switch Controls: Replaced static text badges with modern, accessible sliding toggle switches (
<button role="switch">) featuring explicit Active and Inactive indicators with instant backend cache invalidation. - Copilot Action Approval & Rejection Audit Trail: Every mutating Copilot action proposal (such as collector restarts and tap toggles) now records an immutable, real-time audit record directly within the conversation thread.
- Platform-Standard Modal Dialogs: Eliminated native browser popup prompts (
window.prompt) in favor of keyboard-accessible, themed modal dialogs with Enter/Escape shortcut handling and backdrop dismiss. - Non-Blocking Ingestion Pipeline Guarantees: Telemetry stream tapping uses dedicated asynchronous buffer workers and dead-letter queues (
telemetry_dead_letters), guaranteeing zero backpressure on primary collector ingestion.
New Features & Polish
- Scalable Destination Management:
- Dynamic Search Bar: Real-time destination filtering by destination name, S3 bucket, AWS region, or OTLP endpoint URL.
- Quick Filter Chips: Instant one-click filtering by
All,Active,Inactive,S3, andOTLP. - Routing Summary Badge: Live status pill indicating active routing counts with a green indicator dot.
- Selective Fan-Out & Bulk Routing Toolbar: Multi-select checkboxes for batch destination control with actions: Route to Selected, Stop Selected, Route ONLY These (atomic exclusive routing), Delete Selected, Route All, and Stop All (standby mode).
- Pre-Flight Connectivity Testing: Integrated test action (
⚡ Test) on each destination row verifying S3 credentials/buckets or OTLP HTTP reachability with round-trip latency reporting before activating routes. - Copilot Conversation Audit Records & 3-Tier Sync: Audit cards embedded directly in chat history record action title, unique action ID (
act-...), actor username, outcome, and operator note/rejection reason, with real-time cross-window synchronization (collectorctrl:mcp-action-resolved). - New API Endpoints:
POST /api/governor/destinations/bulk(batch routing) andPOST /api/governor/destinations/test(pre-flight probe).
[v0.4.7-beta] — 2026-09-16
Highlights
- Initial OpenTelemetry Configuration in "Get Started": The onboarding wizard now exposes the initial collector configuration path alongside the binary path, eliminating bootstrap failures on hosts with custom configuration layouts.
- Synchronized Deployment Code Generators: 1-Click curl scripts, manual package steps, Ansible playbooks, Terraform / Cloud-Init, and Windows PowerShell/Silent Setup automatically incorporate custom
--otelcol-config/-OtelConfigarguments and preview them insupervisor.yaml. - Intelligent Platform Defaults: Auto-populates standard paths based on the target OS:
/etc/otelcol-contrib/config.yamlon Linux,C:\otel\config.yamlon Windows, and/usr/local/etc/otelcol-contrib/config.yamlon macOS.
Improvements & Fixes
- Onboarding Endpoint Parameter Forwarding: Fixed non-interactive onboarding wrapper scripts (
/api/onboard/linuxand/api/onboard/mac) properly passing--otelcol-configinto./install.shvia properly quoted array expansion. - PostgreSQL Server Conflict Resolution on Linux: Resolved package manager conflicts on Amazon Linux 2023, RHEL, and Fedora when pre-existing PostgreSQL instances (PostgreSQL 15–18) exist on
@System. The installer now inspectsrpm -qafirst, skipping conflicting package installations and gracefully configuring the existing service. - Resilient TLS Directory Copy: Fixed server installation scripts aborting when
certs/was empty; scripts now verify file presence before copying. - Windows Silent Installer Parameter Extraction: Added
/OTEL_CONFIG="<path>"parameter support toCollectorCtrl-Supervisor-Setup.exe, enabling automated enterprise software management systems to inject custom configuration paths during silent deployments (/VERYSILENT).
[v0.4.6-beta] — 2026-09-15
Highlights
- Single-Approver Model: Removed mandatory two-person approval. Any identity holding the
mcp:approvepermission can now confirm a pending action—including the requester or Copilot session—allowing on-call engineers to resolve incidents rapidly without waiting for a second administrator. - Post-Execution State Verification: Approving lifecycle commands (start, stop, restart for collectors or sidecars) now waits up to ~8 seconds for the agent to report its new state, returning a
post_executionverification verdict (verified: true/false) and surfacing dismissible result banners in Settings → Pending AI Actions. - Self-Healing Log Fetching:
fleet_get_logsautomatically requests live logs when the agent log cache is empty, eliminating the previous two-step requirement to call back withrefresh=true. - Isolated Sidecar Logs: Supervisor and sidecar logs now persist to dedicated columns (
last_sidecar_log), preventing sidecar log fetches from overwriting supervisor logs. Real sidecar logs now display reliably in the Collector Detail view. - Accurate Stopped Collector Status: Intentional process stops now display an explicit Stopped badge in Fleet Overview, report status
"stopped"in MCP tools, and are segregated into a distinctstoppedbucket underneeds_attentioninfleet_health_summaryinstead of masquerading as healthy.
Improvements & Fixes
- Undeliverable Command Handling: Approvals for disconnected agents or failed network dispatches now fail loudly with descriptive errors (e.g., HTTP 502 / 409) and remain in a pending state for retry instead of falsely recording successful execution.
- Permission Safety Rails Preserved: Proposal expiry (15 minutes), blast-radius limits, distinct
mcp:approveprivilege requirements, and dual-gated auto-approval (mcp:auto-approve) remain strictly enforced.
[v0.4.4-beta] — 2026-09-10
Highlights
- 70% Token Reduction & On-Demand Tool Schemas: Copilot now sends full instruction schemas only for core everyday tools, maintaining a compact menu for specialized operations and fetching detailed definitions on demand. This dramatically reduces per-conversation token cost and speeds up responses across all AI providers.
- Root-Cause Analysis Tools: Copilot gains
fleet_health_summaryfor instant fleet-wide health rollups (breakdowns by environment, OS, and version with top offenders) and tail-reading for collector and supervisor logs where errors actually reside. - Resilient Multi-Step Reasoning: Increased reasoning headroom to 10 steps with guaranteed best-effort answers and transparent summaries if iteration limits are reached.
- Fleet Scale & Database-Indexed Lookups: Paginated fleet responses (
fleet_list_agents,supervisors_list,audit_query,fleet_get_samples) with explicit total counts; database-indexed status, OS, and environment filtering for sub-second responses across thousands of collectors.
Improvements & Fixes
- UI Crash Prevention (React Error #310): Fixed conditional React hook registrations that previously caused full-application crash screens in the Collector detail view, Honeycomb view when filtered to zero items, and Admin API documentation tab.
- Copilot Chip Navigation & Accuracy: Copilot links with bare page names now navigate seamlessly in-app without triggering browser reloads; Copilot instructions strictly restrict agent chips to live verified instance IDs.
- Model Tuning Guidance: Added informational hover tooltips to Temperature and Max Output Tokens controls in Admin → BYO-AI Integration, clarifying parameter boundaries and explaining that Max Output Tokens limits reply length rather than conversation spend.
[v0.4.3-beta] — 2026-09-10
Highlights
- Introducing the CollectorCtrl Copilot — an Agentic Control Plane for Your Fleet: A single, persistent AI assistant available on every page (floating button or Ctrl + /). Ask questions in plain English and get answers grounded in your live fleet data through a secure toolbelt — agent inventory, health status, telemetry analytics, audit history, templates, and packages.
- Evidence, Not Vibes: Every answer ships with a collapsible Agent Reasoning trace showing each tool consulted, whether it succeeded, and what it found.
- Inline Approval Cards: The Copilot proposes changes — restarts, config pushes, rollbacks — and presents an approval card right in the chat with blast radius, full diff, and one-click Approve / Reject. Nothing executes without a human decision; every decision is audit-logged, and stale proposals expire automatically.
- Connected UI Journey: Mentioned collectors, templates, policies, and packages render as clickable chips that navigate straight to the component.
- Architecture Diagrams in Chat: Ask for your pipeline topology and get interactive, theme-aware diagrams with a one-click source view — fully self-contained, no external CDN calls.
- Apply to Editor: Suggested YAML lands directly in the collector's config editor with one click, with protection for unsaved work.
- Context-Aware: Page-tailored suggested prompts, "this"-awareness of the open collector including the live config editor buffer, detachable context chips, and persistent per-user chat history (last 50 conversations) that survives refresh.
- Advanced Model Tuning: Admins get Temperature and Max Output Tokens controls, custom model ID support, and Base URL overrides under Admin → BYO-AI Integration — plus built-in guidance for free provider tiers (Gemini, Groq, OpenRouter) and fully private local models via Ollama.
- Security Model: Tools are filtered by the user's RBAC role, the Copilot can never approve its own proposals, PII redaction and AI guardrails carry over, and it works with your configured provider (OpenAI, Anthropic, Gemini, or any OpenAI-compatible endpoint) under Admin → BYO-AI Integration. See the Copilot documentation.
[v0.4.2-beta] — 2026-09-08
Highlights
- Opt-In ZeroTouch Application Discovery (Default: Off): Application discovery is now explicitly opt-in to eliminate unsolicited process scanning on sensitive infrastructure. Manage discovery per host or across the fleet under Settings → Supervisor Management, or configure
discovery.enabled: trueinsupervisor.yaml. - Supervisor Upgrade & Rollback Hardening: Pre-flight verification validates binary architecture, file integrity, and version identity prior to swapping. On Windows, the restart helper automatically restores and launches the previous working binary if an updated supervisor fails to start as a service.
- Security Hardening: Linux installations automatically generate unique self-signed TLS certificates on initial boot; mandatory passphrases are enforced for database backup archives (
.ccbak); and payload ceilings (16 MB for OpAMP, 64 MB for OTLP) protect against memory exhaustion.
Improvements & Fixes
- Package Repository Integrity: Eliminated auto-generated mock/placeholder packages for missing platforms; blocked overwriting package versions actively referenced by fleet policies; and restricted package deletion via
packages:deleteandbuilder:deleteRBAC permissions. - Windows Control Panel Polish: Installed services display clean, stable names (
CollectorCtrl Supervisor) in Programs and Features rather than initial version numbers, and display the official branded icon. - Sanitized Error Rendering: Replaced innerHTML interpolations with safe DOM rendering in global error dialogs to protect against potential client-side script injection.
[v0.4.0-beta] — 2026-09-07
Highlights
- Critical Security Advisory — Unauthenticated Command-Queue API Closed: Endpoints used to queue runtime-injection commands (
/api/commands/queueand/api/commands/status) previously resided on a public allowlist. They now strictly require authentication and explicit ZeroTouch/Fleet control permissions. - Account Lockout & Brute-Force Defense: Accounts automatically lock for 15 minutes after 5 consecutive failed sign-in attempts, displaying remaining lockout countdowns and logging security audit events.
- NIST-Aligned Password Governance: Enforced modern length and complexity standards on password creation/resets, with forced rotation for temporary initial credentials.
- Real-Time Session Invalidation & Account Suspension: Password modifications, role updates, or suspensions immediately terminate active sessions. Admins can suspend user accounts without deleting configuration and audit history.
- Enterprise SSO Enhancements: Dynamic login UI (SSO button hidden unless an IdP is configured), SSO enforcement for non-admin accounts, Just-In-Time (JIT) provisioning toggle, and admin "Test Connection" tool.
Improvements & Fixes
- RBAC Alignment: Settings permissions checklist synchronized with live server catalog; standard operator role restricted from deleting registry items; and canary rollout promotion elevated to
Policies → Control. - Fleet Upgrade Safety: Eliminated phantom upgrade badges on main collectors when sidecar packages are uploaded; fail-fast policy validation rejects assigning sidecars to main collectors or vice versa upon save.
- Windows Installer Special Characters: Fixed token parsing in the Windows installer to handle enrollment tokens containing quotes and special characters.
[v0.3.9-beta] — 2026-09-04
Highlights
- Two-Axis Package Classification: Clear separation between package Purpose (
sidecarvs.main) and Provenance (ocb-buildvs.manual-upload), enforced authoritatively by the server with automated startup migration for legacy packages. - Build Target Selection in OCB: OpenTelemetry Collector Builder supports selecting Build Target (Sidecar or Main Collector), enabling custom distributions to participate in fleet upgrade policies.
- Distribution Versioning: Added custom distribution version field (e.g.
1.4.0) separate from the underlying builder tool version. - Isolated Custom Build Outputs: Builder panel displays only OCB-compiled artifacts; hand-uploaded binaries remain segregated under Package Management → Custom Sidecars.
Improvements & Fixes
- Custom Builder Safety: Eliminated phantom concurrent builds with HTTP 409 conflict responses; automatic CPU architecture tagging; strict module path validation preventing YAML injection; and mandatory receiver/exporter deduplication.
- Sidecar Assignment Validation:
/api/save_sidecarvalidates package existence and purpose classification to prevent misapplying main-fleet binaries as sidecar addons. - Audit Logging: Custom binary builds are now logged under
BUILD_CUSTOM_COLLECTORwith target, OS, and CPU architecture.
[v0.3.8-beta] — 2026-09-03
Highlights
- Automatic Upgrade Rollback & Watchdog: Every collector binary upgrade is continuously health-checked. If a new version stalls, crashes, or fails within 3 minutes, it automatically reverts to the previous working version without manual intervention.
- Strict Sidecar & Main Fleet Separation: Sidecar binaries, configs, and pipelines are completely isolated from standard fleet collectors with dedicated granular permissions (
configure_main,configure_sidecar,control_main,control_sidecar). - CPU Architecture Awareness: Release packages explicitly validate target architectures (
amd64,arm64,universal) with upload-time header inspection to prevent pushing mismatched binaries to hosts. - Windows Agent Deployment Fixes: Streamlined collector and sidecar binary deployment and upgrades on Windows agents, resolving execution errors on package downloads.
- UI & Telemetry Usability: Full pagination in Settings → Audit Logs (15, 25, 50, 100, All); Telemetry Live Stream displays human-readable timestamps, actual collector hostnames, and correlated Trace IDs; responsive Notification Center flyout.
Improvements & Fixes
- Package Management: Prevented identical package names from overwriting rollback binaries; safe cascade cleanup of package references when deleting; reliable sidecar activation on policy assignment.
- Agent & Supervisor: Windows binary extension handling corrected; exponential backoff throttling on package download retries; automatic crash recovery restart for sidecar processes.
- Clean Defaults: Default sidecar configs use no-op exporters instead of writing raw telemetry to disk; custom TLS certificate configurations strictly respected.
[v0.3.7-beta] — 2026-09-01
New Features
- Configuration Drift Inspector (Monaco Diff View): High-fidelity Drift Inspector modal (1120px) featuring a side-by-side Monaco Diff Editor with line numbers, syntax highlighting, and visual change markers; searchable drift timeline; 1-click Locate Agent in Fleet navigation; dedicated Copy Before and Copy After YAML actions.
- Full-Width Executive KPI Strip: Responsive 4-column metric grid displaying Fleet Monitored (with online/offline health badges), Discovered Patterns, Template Coverage (% governed), and 24h Config Drift with an inspect trigger.
- Pattern Sidebar Card Redesign: Widened 340px column with 3-tier card hierarchy (Pattern Identity and status badges, collector count with health dots, and full-width environment color distribution bar).
Improvements & Fixes
- YAML Structural Fingerprinting: Standard 2-space indentation with clean value masking (replacing legacy placeholders with clean empty strings) while preserving empty component mapping structures.
- Automatic Database Upgrade: Database backfill detects legacy placeholders in existing SQLite/PostgreSQL records and automatically updates them.
- Native In-App Dialogs: Replaced browser
alert()popups with styled in-app modals (Historical Pattern Modaland themedErrorModal). - Database Compatibility: Verified 100% dialect-agnostic GORM queries across SQLite and PostgreSQL with auto-migrations and connection pool tuning.
[v0.3.6-beta] — 2026-08-31
New Features
- 10,000-Host High-Scale Architecture: Native GZIP API compression on
/api/agentsand/api/inventory, reducing full-fleet payload sizes from ~20 MB to ~350 KB (a 98% reduction); concurrency guard (isFetchingRef) with a smooth 12-second polling cadence; O(1) fast-path filtering. - Hierarchical Graph Clustering (
SmartscapeTopologyView): Automatically groups edge agents into Cluster/Environment nodes and gateways into HA Gateway Pools when fleets exceed scale thresholds, reducing canvas SVG paths from 1,250,000 to ~20 animated flow links with 60 FPS rendering. - Fleet Heartbeat Engine (
Settings → Database & Maintenance): Configurable ingestion cadence presets (Aggressive 10s, Standard 30s, High Scale 60s, Ultra Scale 120s), dynamic offline detection grace period multiplier, DB write throttling, automated history retention pruning, and live ingestion throughput calculator. - Agent Explorer Modal: Full-screen searchable modal accessible via Explore Agents (N) pill buttons on configuration template cards; clamped running agent tag lists to eliminate scroll bloat.
- High-Scale Load Generator (
cmd/loadgen): Concurrency harness simulating 10,000 concurrent OpAMP agents across 6 realistic OpenTelemetry pipeline archetypes and enterprise environments.
Improvements & Fixes
- Memoized Honeycomb Grid: Health status counters, group slicing, and pagination are memoized in
Honeycomb.tsxfor zero-frame-drop 10k-node rendering. - Design Alignment: Standardized typography, font weights, and descriptions across Settings sub-tabs, Package Management, OCB Builder, and Semantic Registry.
[v0.3.5-beta] — 2026-08-28
New Features
- Modern Borderless Design System: Unified elevation gradients, glass panels (
.glass-panel), modern segmented control tracks, and dynamic fleet topology background grids with animated data-flow pulses. - Telemetry Governor S3 Key Structure Explorer: Interactive 10-tier partition layout with monospace S3 path tags, schema types, field explanations, Athena DDL schema, JSON record models, and AWS lifecycle policies.
- Subtab Routing & Browser Refresh Sync: Query parameter synchronization (
?tab=...) and local storage persistence across Telemetry Governor and Package Management subtabs with full browser back/forward history support. - OpenAPI 3.0 Specification: Published OpenAPI 3.0 specification accessible directly at
GET /api/docs/openapi.json.
Improvements & Fixes
- Global Tap Template Fallback: Monaco editor and backend now provide
DefaultTapYAMLwith pre-configured probabilistic sampling, error filtering, andotlphttpexporter pipelines. - Layout & Contrast Polish: Fixed Get Started page spacing; resolved notification drawer badge wrapping; fixed z-index stacking context for Semantic Registry filter dropdowns.
[v0.3.4-beta] — 2026-08-28
New Features
- ZeroTouch Application SDK Onboarding: Interactive setup modal in the APM Control Plane for 6 major runtimes (Java, Python, Node.js/TS, Go, .NET, Web RUM) with multi-deployment snippets (CLI, Dockerfile, K8s Pod annotations, env matrices) and interactive AI framework tuning.
- Expanded Semantic Registry: Built-in catalog of 70+ standard OpenTelemetry attributes across 10 domains with pre-mapped cross-vendor rules (Datadog, Dynatrace, Coralogix, New Relic, Splunk HEC, Elasticsearch) and 1-click semantic convention seeding.
- Offline-First Local Static Analysis Engine: Automatic fallback to deterministic local parser when external AI providers are offline or rate-limited; instant topology flow generation and security auditing (detecting missing
memory_limiter, missingbatchprocessors, orinsecure: trueendpoints). - Data Wizard Dual Sampling & Drop Filters: Toggle between Tail-Based Sampling (preserving 100% of error traces and high-latency spans) and Hard-Drop Filtering (compiled OTel filterprocessor rules) with AI double-validation.
- Dynamic Time Filtering: Analytics and telemetry time ranges (15m, 1h, 6h, 24h, 7d, 30d, All Time) with automatic time-scale adjustments and persistent pattern durability.
[v0.3.3-beta] — 2026-08-27
New Features
- Autonomous Process Discovery: Supervisor continuously scans and reports running application runtimes (Java, .NET, Node.js, Python) on Windows, Linux, and macOS, detecting standalone processes and registered background Windows Services.
- Graceful 30-Minute Stale Retention: Disconnected nodes enter a
Reconnectingstate with dimmed opacity and dashed telemetry lines for up to 30 minutes instead of abruptly vanishing during network blips or service restarts. - Unified Release Versioning: Unified build scripts, Inno Setup installers, server binaries, and supervisor daemons with official release tags; active Server Version displayed in License Management.
Improvements & Fixes
- Supervisor Version Tracking: Accurate semantic version parsing for pre-release suffixes (
-beta), correctly identifying installed supervisors as Up to date. - Self-Healing Upgrade Status: Outdated upgrade error banners automatically clear once a supervisor verifies connection and runs the target release.
[v0.3.2-beta] — 2026-08-25
New Features
- AI Pattern Optimizer & Variable Masking: Automated discovery of dynamic variables (transaction hashes, session tokens, tenant IDs) across log streams with 1-click rule application, custom regex builders, and multi-provider AI model support (OpenAI, Anthropic, Gemini, Groq, local Ollama).
- Contextual Origin Metadata: Pattern cards display originating service names, source file paths, and host/pod names with multi-source aggregation badges.
- AI OpenTelemetry Edge Drop Filter Generator: 1-click generation of verified OpenTelemetry Transformation Language (OTTL) filter processors (
filter/drop_noisy_patterns) with safe-by-default guardrails (severity_number < 13preserving WARN/ERROR logs) and real-time cloud cost reduction estimates. - Data Wizard Value Chain Guide: Visual 3-step guide for defragmenting patterns, pinpointing noise origins, and deploying edge filters.
[v0.3.1-beta] — 2026-08-24
New Features
- Automated Incident Detection & 1-Click Triage: Real-time operational incident detection for critical OpAMP configuration rejections, collector/supervisor disconnects, and license quota alerts, with direct deep-link routing to affected resources.
- Floating Alert Toasts & Desktop Push: Animated bottom-right toast cards with auto-dismiss timers; native OS desktop push notifications via Web Notification API; pulsating bell indicator for unread alerts.
- Outbound Alert Channels & Webhooks: Multi-platform alert dispatching for Slack, Microsoft Teams, PagerDuty, and generic webhooks with non-blocking background dispatching and delivery testing in Settings → Alert Channels.
Improvements & Fixes
- Port Conflict Isolation: Supervisor sidecars automatically assign offset ports (
14317,14318,8889) withtelemetry.metrics.level: noneto prevent port collisions with primary OpenTelemetry collectors. - WSS / TLS Support: Verified encrypted
wss://agent connectivity with custom CA certificate verification.
[v0.3.0-beta] — 2026-08-24
New Features
- Telemetry Insights Explorer Overhaul: Dynamic fleet filtering (Environment, Cluster, Host, Service); recursive OTLP protobuf log unpacking into clean human-readable text; pattern anomaly pagination and live stream table with severity chips (
All,Errors,Warnings,Info). - Dynamic Fleet Topology Exporters: Automated extraction of configured exporters, receivers, and processors directly from effective collector YAMLs across 20+ destination types (Datadog, Dynatrace, Grafana Cloud, Splunk, New Relic, Elasticsearch, etc.).
- Roles & Permissions (RBAC) Expansion: Granular capabilities added for ZeroTouch auto-instrumentation (
zerotouch) and AI/MCP governance (mcp); explicit system default role visibility (ADMIN,EDITOR,VIEWER). - Settings API Catalog & OpenAPI 3.0: Searchable 50+ endpoint directory with role requirements, HTTP method badges, and OpenAPI 3.0.3 schema sync.
Improvements & Fixes
- Theme-Aware UI Visibility: Resolved contrast and readability issues across the Get Started page, Notification Center, and Telemetry Command Center in Light Theme.
[v0.2.9-beta] — 2026-08-19
New Features
- Next-Gen Pipeline Topology Visualizer: Symmetrical multi-stage DAG flow (
Receivers ➔ Processors ➔ Exporters ➔ Connectors) with compact 280px stage spacing; flow-isolated swimlanes mode for logs, metrics, and traces; interactive signal filter tabs. - Floating Component Config Inspector: Clicking any pipeline component opens an inspector showing active pipeline associations, syntax-highlighted YAML, 1-click clipboard copy, View in Full YAML focus navigation, and Ask AI troubleshooting.
- Real-Time Hover Path Tracer: Highlights transitive upstream and downstream execution routes with animated glowing paths while dimming unrelated components.
Improvements & Fixes
- Monaco Configuration Editor: Standard 2-space indentation (
tabSize: 2,insertSpaces: true), dynamic fleet environment display in policy baselines, and rebranded AI Assistant triggers. - React Runtime Chunking: Unified React runtime bundling in Vite, eliminating dual-instance React dispatcher conflicts (
Minified React error #310).
[v0.2.8-beta] — 2026-08-18
New Features
- APM Instrumentation Control Plane: Zero-code OpenTelemetry APM instrumentation, automatic runtime discovery, and centralized desired-state agent lifecycle governance across the fleet.
- Smart Discovery-Aware Node Targeting: Agent binaries and rollout baselines dispatch exclusively to hosts where that application runtime is discovered, preventing resource waste and ensuring accurate convergence metrics.
- Agent Baselines & Central Catalog: Clean horizontal runtime agent baseline rows, dedicated catalog asset modal for SHA256 checksum inspection and version promotion, and in-app promotion confirmation safeguards.
- Official Runtime Brand Logos: Vector SVG brand logos for Java, .NET, Node.js, Python, and Go across all runtime views.
[v0.2.7-beta] — 2026-08-11
New Features
- Model Context Protocol (MCP) Server: Native MCP integration allowing AI assistants (Claude Desktop, Cursor, and MCP clients) to securely query and manage your OpenTelemetry fleet.
- Phase 1 Read-Only Fleet Access: 12 read-only tools (
fleet_list_agents,fleet_get_samples,governor_volume,governor_cardinality,audit_query,discovery_reports, etc.) authenticated with existing API tokens and RBAC roles. - Phase 2 AI-Proposed Changes with Human Approval: Opt-in propose/confirm workflow (
COLLECTORCTRL_MCP_PHASE2=true) creating 15-minute pending actions requiring two-person human approval in Settings → Pending AI Actions. collectorctrl-mcpCLI Bridge: Stdio bridge adapter for desktop MCP clients connecting to the CollectorCtrl management server.
[v0.2.6-beta] — 2026-08-10
New Features
- Get Started Onboarding Wizard: Dedicated dashboard page that generates ready-to-run install commands for Linux, Windows, macOS, Kubernetes, and Docker — with the server address auto-detected and the latest release package resolved live from GitHub.
- Four Deployment Methods per Platform: 1-click script, manual package steps, Ansible playbook, and Terraform/Cloud-Init snippets.
- Live Agent Check-In Detector: Real-time banner confirming the moment a newly deployed host connects.
- Server Package Mirror: Server mirrors supervisor release packages from GitHub into
/packages/so agent hosts need no internet access. - Kubernetes Operator: Helm-chart-deployed operator (CollectorCtrl-K8s-Operator) for cluster-wide collector management.
- DevSecOps Package URLs: One-click copy of direct release asset links for SAST and anti-virus scanning.
Improvements & Fixes
- Windows Uninstaller Robustness: Idempotent service stop/uninstall with clean exit codes.
- OpAMP Auth Fallback:
OPAMP_LEGACY_MODEdefaults totrue, resolving 401 errors for unauthenticated collectors. - PostgreSQL Linux Installer: Automatic
pg_hba.confconfiguration (scram-sha-256) on RHEL, Amazon Linux, Ubuntu, and Debian. - Pure Go Cross-Compilation: CGO-free release packages for Windows, Linux, and macOS on
amd64andarm64.
[v0.2.5-beta] — 2026-08-07
New Features
- ZeroTouch Instrumentation (Alpha): Discovers running VM applications automatically and facilitates zero-touch OpenTelemetry sidecar instrumentation.
- Interactive Fleet Topology: Visual network connection mapping and pipeline relationship tracking.
- Git Integration (Alpha): Real-time Git synchronization to store, version, and manage collector configurations directly from Git repositories.
Improvements & Fixes
- Public Docker Containers: Multi-architecture container images on GitHub Container Registry (
ghcr.io/collectorctrl/collectorctrl-serverandcollectorctrl-supervisor). - PostgreSQL Script Compatibility: Versioned package resolution (
postgresql15-server,postgresql16-server) on Amazon Linux 2023, Amazon Linux 2, RHEL, CentOS Stream, and Fedora.
[v0.2.4-beta] — 2026-07-15
New Features
- Granular RBAC Engine: Replaced coarse role checks with granular, feature-based Role-Based Access Control across Fleet Orchestration, Telemetry Governor, and Supervisor updates.
- Custom Role Builder: Configure fine-grained permissions per functional area without requiring full administrative privileges.
[v0.2.3] — 2026-07-02
Improvements & Fixes
- Native .tar.gz Extraction: Direct package decompression in supervisor binary extraction logic, eliminating raw binary requirements.
- Cross-Device Binary Swap Fix: Resolved Linux
EXDEVcross-device link errors during supervisor self-upgrades across different mount points.
[v0.2.2-beta] — 2026-06-16
Improvements & Fixes
- Interactive Linux Installer: Guided terminal setup for OpAMP endpoint, collector binary path, and configuration file path with real-time validation.
- Path Validation: Pre-flight checks verifying binary and configuration file existence prior to supervisor startup.
[v0.2.1-beta] — 2026-06-11
New Features
- Telemetry Governor Engine Integration: Standardized deployment workflow, unified credentials, and real-time connectivity feedback for the ClickHouse + OpenTelemetry Collector data plane.
[v0.2.0-beta] — 2026-06-10
New Features
- Cross-Platform Linux Support: Automated
install.shserver deployment and systemd supervisor service (collectorctrl-supervisor.service). - PostgreSQL Database Support: Production-grade PostgreSQL database engine support alongside embedded SQLite.
[v0.1.0-alpha] — 2026-05-05
Core Platform
- Initial Alpha Release: Windows-native management server and native supervisor agent for OpenTelemetry fleets.
- OpAMP WebSocket Gateway: Agent management protocol implementation with centralized dashboard and active configuration reconciliation.
- Config Drift Prevention: Continuous reconciliation engine with supervised restarts and fallback to known-good configurations.
CollectorCtrl follows Semantic Versioning. For detailed commit history and assets, see the GitHub repository releases.
CollectorCtrl