Changelog

Enhanced security monitoring and compliance

Rolled out comprehensive security improvements including network traffic monitoring, automated threat detection with real-time alerts, and extended audit log retention to 365 days.

• ›Network flow logs enabled across all environments
• ›Threat detection with automated alerting for security findings
• ›Cross-region data replication for disaster recovery
• ›Full API audit trail logging for compliance

Infrastructure monitoring and alerting

Added proactive monitoring with alerts for database capacity, background job processing health, and service error rates.

Browse All pagination fix

Fixed broken pagination in the Browse All workloads tab with a cleaner prev/next navigation pattern.

Redesigned user invitations

Merged the separate Invitations settings tab into the Users page for a streamlined admin experience.

Dispatching status for workloads

Introduced a new Dispatching status to provide better visibility when workloads are being sent to the cluster, closing the gap between Queued and Running.

Dashboard UI refresh

Major dashboard overhaul with new stat cards showing trends, period-over-period comparisons, and improved chart visuals.

Cross-team workload visibility

Managers can now see workloads across all nested sub-teams, providing full visibility into their team hierarchy.

Faster dashboard loading

Dashboard now loads progressively with independent data sections, dramatically reducing perceived load time.

Slack notification fixes

Resolved issues with Slack notification delivery and improved message formatting for workload status updates.

Dashboard chart accuracy

Fixed dashboard utilization charts that were occasionally showing gaps in recent data.

Automatic workload status reconciliation

Chamber now automatically detects and corrects workload status mismatches with the underlying cluster, resolving stuck or incorrectly-reported workloads without manual intervention.

Teams and Leaderboard updates

Refreshed the Teams and Leaderboard dashboard tabs with improved visualizations and team performance comparisons.

Workload status notifications

Get notified when your workloads change status. Configurable delivery via Slack and email with per-user preferences.

• ›Real-time notifications for queued, running, completed, and failed states
• ›Bulk notification management (mark-all-read, delete-all)

GCP support for dynamic pools

Dynamic capacity pools now support GCP GPU instance types, expanding multi-cloud GPU management to both AWS and Google Cloud.

On-Demand capacity mode

New On-Demand option for workload submission — run workloads on any available cluster without a pre-allocated reservation.

• ›On-Demand toggle in the workload submission form
• ›Automatic cluster selection based on available GPU types

Status badge visual refresh

Improved workload status badge colors for better readability — Running is green, Completed is dark green, and Pending is muted gray.

AI-powered preemption explanations

When a workload is preempted, Chamber provides a clear, human-readable explanation of why — including which workload took priority and which team submitted it.

Dynamic capacity pools

Capacity pools now support dynamic, auto-scaling GPU allocation — GPUs are provisioned on-demand and released when idle.

Elastic workload auto-requeue

Elastic workloads that are preempted are now automatically requeued instead of being terminated, reducing manual resubmissions.

Notification center overhaul

Redesigned the notification experience with a bell icon in the sidebar, a slide-out panel, and bulk actions like mark-all-read and delete-all.

Error status and failure reasons

Workloads that fail now show a dedicated Error status with a clear failure reason, making it easier to diagnose issues without leaving Chamber.

Smart polling and manual refresh

Workload tables now auto-refresh intelligently and show a last-refreshed indicator. Manual refresh button available for immediate updates.

AI scheduling explanations

Workloads now display AI-generated explanations for their scheduling status, helping users understand why a workload is queued, running, or was preempted.

Customizable email notifications

Configure email notification preferences per user, including which status changes trigger emails and how they are formatted.

Slack integration

Connect Chamber to your Slack workspace for real-time workload status notifications delivered directly to channels or DMs.

Reserved and Elastic queue separation

Reserved and Elastic workloads now use separate scheduling queues, improving fairness and preventing elastic workloads from blocking reserved capacity.

Dashboard fit-and-finish updates

Polished the dashboard with improved KPI card spacing, cleaner chart tooltips, and consistent number formatting across all metrics.

Customer-facing REST APIs

New REST API endpoints for managing workloads, teams, allocations, and templates — enabling programmatic access and third-party integrations.

CLI authentication

Authenticate with Chamber from the command line for headless workload submission and automation workflows.

Redesigned dashboard with overview stats

Completely rebuilt dashboard with overview stats, GPU utilization charts, active workloads timeline, and sortable workloads tables.

• ›Submitter column with user avatars
• ›Demo mode toggle for enterprise-scale previews

Voice input for AI assistant

Use hold-to-talk voice input to interact with the Chamber AI assistant, powered by client-side Whisper transcription.

In-app notification center

Real-time in-app notifications with a bell icon, slide-out panel, toast alerts, and polling for new updates.

GPU leaderboard

New Leaderboard dashboard tab ranking GPU users by utilization, helping teams identify top consumers and optimize resource allocation.

Shareable dashboard URLs

Dashboard state (active tab, time range, filters, cluster selection) is now reflected in the URL, making it easy to share specific views with teammates.

Elastic-only GPU types

GPU types available only through elastic capacity are now visible in the workload submission form, with Reserved disabled when no reservation exists.

Keyboard shortcuts

Added keyboard shortcuts for common actions — Cmd+E to toggle the AI assistant and Cmd+B to open/close the chat drawer.

AI-generated email summaries

Ask the AI assistant to draft and send GPU utilization summaries via email, with a preview and confirmation flow before sending.

AI navigation suggestions

The AI assistant can now suggest in-context navigation to relevant pages based on your questions, with clickable buttons to jump directly there.

Public REST APIs

Launched customer-facing REST APIs for workloads, stats, metrics, and capacity management with full documentation.

Workload templates

Save frequently-used workload configurations as templates for one-click reuse. Templates support all workload fields including GPU type, container image, and environment variables.

Workload cloning

Clone any existing workload with a field-selection modal — choose which fields to copy and which to customize for the new submission.

Advanced scheduling options

New advanced scheduling UI with GPU fraction support, priority controls, and fine-grained scheduling preferences.

Weights & Biases integration

Connect your Weights & Biases account to automatically link experiment tracking with GPU workloads managed by Chamber.

Multi-cluster job scheduling

Chamber now supports scheduling workloads across multiple Kubernetes clusters with unified queue management and resource allocation.

Profile pictures

Upload and edit your profile picture. User avatars now appear in workload tables and team views for quick identification.

Filter workloads by submitter

Filter the workloads table by who submitted each job, with submitter avatars displayed inline.

Command palette

Press Cmd+K to open a command palette for fast navigation across dashboard views, projects, and clusters.

Pinnable sidebar

Pin the navigation sidebar open for quick access, or let it auto-collapse to maximize dashboard space. Page content dynamically adjusts.

Google and GitHub SSO

Sign in with Google or GitHub in addition to email/password, with seamless profile completion for OAuth users.

Redesigned login and signup pages

Refreshed the login and signup experience with a cleaner design, better error messaging, and smoother OAuth flows.

Admin announcements and banners

Platform admins can create and publish announcements that appear as banners across the dashboard for all users in the organization.

Pending invitation visibility

Users can now see pending invitations to other organizations, making it easy to accept invites without a separate email link.

Real-time workload metrics dashboard

Completely redesigned dashboard with live GPU utilization metrics, active workload charts, and organization-level overview stats.

Forgot password and reset flow

Added self-service forgot password and reset password flows for email/password users.

Agent enrollment endpoint

Public API endpoint for cluster agents to enroll with token-based verification, enabling automated cluster onboarding.

Clusters view

New Clusters page showing connected clusters with GPU model metadata, utilization progress bars, and pool-cluster linking.

Settings hub

Comprehensive settings page with change password, organization ID display, and navigation improvements for account management.

Organization tiers (Free and Paid)

Introduced a tiered plan system with free and paid tiers. Premium features are gated with upgrade prompts and video previews.

Freemium signup experience

New streamlined signup flow with email/password registration and Google SSO for free-tier users.

Weekly GPU utilization digest

Receive a weekly email summary of GPU utilization with embedded charts, team performance breakdowns, and direct links to the dashboard.

Redesigned cluster setup

Redesigned the Add Cluster page with a tab-based token flow, interactive tooltips, and GPU plugin FAQ section.

Email preferences

Manage email notification preferences from the Integrations page — control which digest emails you receive and how often.

CI/CD pipeline with multi-stage deployment

Automated deployment pipeline with dev, staging, and production stages, including manual promotion gates and frontend deployment.

YAML workload submission

Submit workloads by pasting Kubernetes YAML manifests. Chamber automatically parses GPU requirements, container images, and resource requests.

• ›Toggle between form-based and YAML submission modes
• ›Automatic GPU resource extraction from manifests

Sortable workload tables

All workload tables now support column-based sorting for name, status, priority, submission time, and GPU count.

GPU metadata in clusters view

Cluster cards now display GPU model information and improved utilization metrics for better resource visibility at a glance.

Workload discovery

Chamber automatically discovers and tracks GPU workloads running across your connected clusters — even workloads submitted outside Chamber.

Real-time GPU metrics

Live GPU utilization metrics in the dashboard with organization-level views, team breakdowns, and per-workload tracking.

• ›Organization-wide utilization overview
• ›Team utilization graphs and comparison tables
• ›Top workloads by GPU utilization

Chamber platform launch

Initial release of the Chamber GPU capacity control plane — enterprise-grade GPU resource management for AI infrastructure at scale.

• ›Project hierarchy for organizational modeling
• ›Capacity pool management with reserved and elastic workload classes
• ›Intelligent scheduler with preemption support
• ›Real-time GPU utilization dashboard
• ›Role-based access control and SSO authentication
• ›Multi-cluster, multi-cloud support