Skip to content
Projects

manzi689/cloud-ai-orchestra

View on GitHub

manzi689/cloud-ai-orchestra has 68 GitHub stars and 0 forks. GitHot measured 0 stars in 24 hours and 0 in 7 days. Last checked 2026-11-16 23:28 UTC. Last push 2026-11-07.

How these counts are measured

What you can do

  • 🎛️ Responsive UI — The Control Deck adapts to phones, tablets, ultrawide monitors, and everything in between.
  • 🌐 Multilingual Support — Interface and documentation available in a growing set of languages, with right-to-left layouts fully supported.
  • 🕰️ 24/7 Customer Support — Round-the-clock coverage from humans who actually use the product.
  • 🧩 Composable Plugin System — Write a plugin once, use it across every portage route.

Measured star growth

in 24h
0
in 7 days
0
in 15 days
Update delayed
in 1 month
Update delayed

Net star changes over overlapping periods, not daily totals. Missing measurements are left blank. Stars measure attention, not software quality.

About

OpenCloud AI Orchestrator 2026: Unified Management for Scalable Cloud-Native AI Workloads

Self-hosted

README.md

Original

preview Download

☁️ CloudMesh Portage

The configurable airlock between your ideas and the cloud-native AI cosmos.

CloudMesh Portage is a next-generation orchestration plane for cloud-native AI services. Where its conceptual predecessor focused on unifying deployments, CloudMesh Portage focuses on the journey — the portage between environments, runtimes, providers, and teams. Think of it as a dry dock for AI workloads: load, inspect, reroute, and launch — all from a single, elegant console.

Download


🧭 Table of Contents

Download


🌌 Why CloudMesh Portage?

Most orchestration platforms treat deployment as a one-way trip. You pick a provider, you pick a runtime, you push, and you hope. CloudMesh Portage treats deployment as a portage — a deliberate, reversible, observable movement of workloads between shores.

Whether you're migrating a model from a notebook to a managed inference endpoint, shifting a pipeline from one region to another, or splitting a monolith into microservices, CloudMesh Portage gives you a helm, a map, and a crew.

Key questions we answer differently:

  • Where does my workload actually live right now? — A living topology map, not a stale diagram.
  • What changes if I reroute this? — Predictive diffing before any change is committed.
  • Who touched what, and when? — A permanent, cryptographically anchored log.
  • Can it speak my team's language? — Yes. Literally. See multilingual support below.

Download


⚓ The Metaphor: Dry Dock for Intelligence

Picture a harbor. Ships come in, get repaired, refitted, and sent back out. CloudMesh Portage is that harbor, but for AI workloads:

  • Dry Dock — A sandboxed staging zone where workloads are assembled and validated.
  • Tide Charts — Real-time metrics that tell you when to launch and when to wait.
  • Harbor Master — The policy engine that keeps order.
  • Lighthouse — The alerting layer that never blinks.
  • Tugboats — Lightweight agents that nudge workloads between nodes.
  • Manifests — Declarative descriptions of intent, not imperative scripts.

The metaphor isn't decoration. It shapes the UX, the API naming, and the mental model your team will adopt on day one.

Download


🛠️ Core Capabilities

CloudMesh Portage is built around eight pillars, each designed to remove a specific kind of friction from cloud-native AI operations.

1. Unified Service Fabric

Bring together managed inference endpoints, self-hosted runtimes, edge nodes, and batch jobs under one control plane. Every workload gets a canonical identity, regardless of where it runs.

2. Declarative Portage Manifests

Describe the what and the why, not the how. Manifests are versioned, diffable, and reviewable in pull requests.

3. Predictive Rerouting

Before you move a workload, CloudMesh Portage simulates the move. Latency, cost, and availability implications are surfaced in a single pane.

4. Zero-Downtime Handoffs

Drain, warm, cutover, verify. The handoff protocol is opinionated and battle-tested.

5. Policy-as-Code Harbor Master

Define guardrails — region restrictions, model allowlists, budget ceilings — and enforce them at the point of change, not after.

6. Cross-Provider Neutrality

No provider is privileged. CloudMesh Portage treats every runtime as a first-class citizen, and every migration as a supported operation.

7. Observable by Default

Every action emits structured events. Every event is queryable. Every query can become an alert.

8. Extensible by Design

A plugin system lets teams encode their own portage routes, validators, and notifiers.

Download


✨ Feature Highlights

A non-exhaustive tour of what you get out of the crate:

  • 🎛️ Responsive UI — The Control Deck adapts to phones, tablets, ultrawide monitors, and everything in between.
  • 🌐 Multilingual Support — Interface and documentation available in a growing set of languages, with right-to-left layouts fully supported.
  • 🕰️ 24/7 Customer Support — Round-the-clock coverage from humans who actually use the product.
  • 🧩 Composable Plugin System — Write a plugin once, use it across every portage route.
  • 📜 Immutable Audit Log — Append-only, hash-chained, exportable.
  • 🔁 Bidirectional Sync — Changes flow in and out; nothing is trapped.
  • 🧪 Sandbox Dry Dock — Rehearse changes without touching production.
  • 🗺️ Topology Map — Live, interactive, and filterable.
  • 🧠 Intent-Aware Diffing — Diffs show meaning, not just YAML noise.
  • 🔐 Secret-Agnostic Design — Bring your own secret manager; we never store what we don't need.
  • 📊 Tide Charts Dashboards — Customizable, shareable, and embeddable.
  • 🚨 Lighthouse Alerts — Route notifications to wherever your team already lives.
  • 🧾 Manifest Linting — Catch policy and syntax issues before they ship.
  • 🔄 Rollback in One Gesture — Every change has a corresponding undo.
  • 🧱 Bring Your Own Compute — Use managed, self-hosted, or hybrid footprints.
  • 🪄 Guided Onboarding — A tour that actually teaches, not just annotates.
  • 📦 Exportable Topologies — Take your configuration with you, always.
  • 🧭 Route Templates — Reusable patterns for common migration scenarios.
  • 🛰️ Edge-Aware Scheduling — Push inference closer to where data is born.
  • 🧮 Cost Horizon — Forecast spend across providers before you commit.

Download


🏗️ Architecture Overview

CloudMesh Portage is organized into five cooperating layers. Each layer can be deployed independently, and each exposes a documented interface.

Layer 1 — Manifest Layer. Declarative descriptions of workloads, routes, and policies. Stored as plain text, versioned in your existing repositories.

Layer 2 — Reconciliation Layer. The engine that compares desired state to actual state and computes the minimal set of actions.

Layer 3 — Portage Layer. The agents that execute handoffs between environments, runtimes, and providers.

Layer 4 — Observation Layer. Metrics, logs, traces, and events, normalized into a single schema.

Layer 5 — Experience Layer. The Control Deck UI, the CLI, the API, and every plugin.

The layers communicate through well-defined contracts. You can replace any layer with your own implementation as long as the contract holds.

Download


🚚 Portage Routes (Workflows)

A portage route is a named, reusable workflow that moves a workload from one state to another. Routes are the heart of CloudMesh Portage.

Common routes shipped out of the crate:

  • Notebook → Managed Endpoint — Turn a research artifact into a production service.
  • Region A → Region B — Move for latency, compliance, or cost.
  • Provider X → Provider Y — Avoid lock-in without rewriting.
  • Monolith → Microservices — Split a large model server into focused components.
  • Batch → Streaming — Reshape a pipeline for real-time use.
  • Cloud → Edge — Push inference closer to the source.
  • Edge → Cloud — Pull it back when the math changes.
  • Canary → Full Rollout — Promote gradually, verify constantly.
  • Blue → Green — Classic swap, modern tooling.
  • Shadow → Primary — Let the new version learn before it leads.

Every route is inspectable, editable, and forkable. If a route doesn't exist, you can compose one from primitives in minutes.

Download


🎛️ The Control Deck

The Control Deck is the primary user interface. It's organized into six modes:

  1. Harbor View — A live map of everything you run, everywhere.
  2. Dry Dock — Staging and rehearsal space for upcoming changes.
  3. Route Editor — Visual and text-based composition of portage routes.
  4. Tide Charts — Dashboards for metrics, cost, and health.
  5. Lighthouse — Alert configuration and incident timeline.
  6. Harbor Master — Policy authoring and enforcement status.

Each mode is keyboard-navigable, screen-reader-friendly, and consistent across devices.

Download


📱 Responsive UI & Accessibility

The Control Deck is built mobile-first. On a phone, it's a capable companion app for on-call engineers. On a tablet, it's a review surface for pull requests. On a desktop, it's a full operations console.

Accessibility commitments:

  • WCAG 2.2 AA conformance as a floor, not a ceiling.
  • Full keyboard navigation with visible focus states.
  • Screen-reader-optimized semantics throughout.
  • Color-blind-safe palettes and configurable contrast.
  • Motion-reduction preferences respected everywhere.

We treat accessibility as a feature, not a checkbox. If something feels hard to use, that's a bug, and we want to hear about it.

Download


🌍 Multilingual Support

CloudMesh Portage speaks many languages — and treats them as peers, not translations.

  • Interface localization with community-maintained strings.
  • Right-to-left layout support for Arabic, Hebrew, and Persian.
  • Locale-aware number, date, and currency formatting.
  • Documentation available in multiple languages, with a translation pipeline that keeps versions in sync.
  • A glossary system so that domain terms stay consistent across languages.

If your team works across borders, CloudMesh Portage will meet them where they are.

Download


🛎️ Support That Never Sleeps

The sun never sets on a cloud-native deployment, and it never sets on our support either.

  • 24/7 Customer Support across chat, email, and community forums.
  • Tiered response targets based on severity, clearly documented.
  • Named engineers for enterprise engagements.
  • Post-incident reviews available on request.
  • Documentation that respects your time — searchable, versioned, and honest about limitations.

Support isn't a cost center here. It's a design input.

Download


🔐 Security Posture

Security is a property of the system, not a feature bolted on.

  • Least-privilege defaults for every component.
  • Scoped credentials, rotated on a schedule you control.
  • Signed manifests and verifiable build provenance.
  • Network policies expressed declaratively alongside workloads.
  • Regular third-party assessments, summaries published openly.
  • Coordinated disclosure process with clear timelines.

We design as if the perimeter is already breached, because in 2026, it often is.

Download


📡 Observability & Telemetry

Every layer emits structured events. Every event is queryable. Every query can become an alert.

  • Metrics — Latency, throughput, error rates, saturation.
  • Logs — Structured, correlated, and searchable.
  • Traces — End-to-end, across providers and runtimes.
  • Events — The narrative of your system, in order.
  • Cost — Attributed to teams, services, and routes.

Bring your own backend, or use the built-in Tide Charts. Either way, you'll never wonder what happened.

Download


🧩 Integration Ecosystem

CloudMesh Portage plays well with others.

  • CI/CD systems — trigger portage routes from your existing pipelines.
  • Secret managers — bring your own, we'll never store what we don't need.
  • Notification channels — Slack, Teams, email, webhooks, and more.
  • Identity providers — SSO via standards you already trust.
  • Storage backends — object stores, databases, and versioned repositories.
  • Observability stacks — OpenTelemetry-compatible by default.

If it has an API, we can probably speak to it. If it doesn't, we can probably write a plugin.

Download


🗺️ Roadmap 2026

A snapshot of what's on the horizon:

  • Q1 2026 — Route Templates Marketplace (community-contributed, reviewed for quality).
  • Q2 2026 — Cost Horizon GA, with per-route forecasting.
  • Q3 2026 — Edge-Aware Scheduling with hardware-aware placement.
  • Q4 2026 — Harbor Master policy simulation mode.

The roadmap is a conversation, not a contract. Join the discussion in the Community section.

Download


🤝 Contributing

We welcome contributions of all shapes: code, documentation, translations, route templates, and bug reports.

  • Read the contributing guide before opening a pull request.
  • Keep changes focused; large PRs are hard to review and easy to stall.
  • Write tests for new behavior. If you can't, explain why in the PR.
  • Be kind in reviews. Assume good intent. Disagree without being disagreeable.
  • Sign off on commits where required.

The project follows a lightweight governance model described in the repository.

Download


💬 Community

The best part of CloudMesh Portage is the people who use it.

  • Discussion forums for questions and ideas.
  • Monthly community calls, recorded and transcribed.
  • A translation guild for multilingual contributors.
  • A route template library curated by the community.
  • Office hours for newcomers.

Whether you're running one workload or ten thousand, there's a place for you here.

Download


📄 License

This project is licensed under the MIT License. See the LICENSE file for the full text.

Copyright (c) 2026 CloudMesh Portage contributors.

Download


⚠️ Disclaimer

CloudMesh Portage is provided "as is," without warranty of any kind, express or implied, including but not limited to the warranties of merchantability, fitness for a particular purpose, and non-infringement. In no event shall the authors or copyright holders be liable for any claim, damages, or other liability, whether in an action of contract, tort, or otherwise, arising from, out of, or in connection with the software or the use of the software.

CloudMesh Portage is not affiliated with, endorsed by, or sponsored by any cloud provider, AI vendor, or infrastructure company mentioned in this document. All trademarks are the property of their respective owners.

Operational decisions — including but not limited to deployment, routing, scaling, and cost management — remain the sole responsibility of the operator. Always test changes in a non-production environment before applying them to live systems.

The 2026 roadmap is aspirational and subject to change without notice.

Download