Skip to content

Deployment System

Manifest Platform uses a ring-based deployment model to move solutions from development through staging into production. Rather than deploying directly to users, changes progress through isolated rings — each with its own configuration, health checks, and approval gates — so problems are caught early and rollbacks are fast.


How Deployments Work

Every solution in Manifest Platform is deployed as a versioned bundle of components: agents, workflows, code blocks, connectors, models, and configuration. When you trigger a deployment, the platform packages the current solution state into an immutable deployment artifact and places it into the target ring.

graph LR
    B["Build"] --> D["Dev Ring"]
    D -->|"Promote"| S["Staging Ring"]
    S -->|"Promote"| P["Production Ring"]

    D ---|"Automated tests"| DH["Health checks"]
    S ---|"Integration tests\n+ Approval gate"| SH["Health checks"]
    P ---|"Canary analysis\n+ Approval gate"| PH["Health checks"]

Deployment Lifecycle

A deployment moves through a defined set of states from creation to completion.

stateDiagram-v2
    [*] --> Queued: Deployment triggered
    Queued --> Building: Artifact packaging
    Building --> Deploying: Build succeeded
    Building --> Failed: Build error
    Deploying --> Validating: Containers healthy
    Deploying --> Failed: Deploy error
    Validating --> Active: Health checks pass
    Validating --> RollingBack: Health checks fail
    RollingBack --> RolledBack: Previous version restored
    Active --> Superseded: New deployment promoted
    Active --> RollingBack: Manual rollback
    Failed --> [*]
    RolledBack --> [*]
    Superseded --> [*]
State Description
Queued Deployment request accepted, waiting for a build slot
Building Solution components are being packaged into a deployment artifact
Deploying Artifact is being rolled out to the target ring's infrastructure
Validating Automated health checks and smoke tests are running
Active Deployment is live and serving traffic in the ring
Failed An error occurred during build, deploy, or validation
Rolling Back The platform is restoring the previous active deployment
Rolled Back The previous deployment has been successfully restored
Superseded A newer deployment has taken over in this ring

Core Concepts

Deployment Artifacts

A deployment artifact is an immutable snapshot that includes:

  • Component code -- Python source for code blocks, agents, and hosted services
  • Configuration -- Environment variables, feature flags, and ring-specific overrides
  • Dependencies -- Pinned requirements.txt or pyproject.toml with locked versions
  • Component manifests -- Pydantic schema definitions, tool registrations, workflow DAGs
  • Metadata -- Git SHA, build timestamp, deploying user, and solution version

Artifacts are content-addressed and stored in the platform's artifact registry. The same artifact can be promoted across rings without being rebuilt.

Deployment Targets

Each deployment targets a specific ring within a workspace. The platform resolves the appropriate infrastructure, configuration, and routing rules for that ring automatically.

One active deployment per ring

Each ring has exactly one active deployment at a time. When a new deployment reaches the Active state, the previous deployment transitions to Superseded.


Deploying from the CLI

The fastest way to deploy is through the flow-sdk CLI:

Coming Soon

The flow-sdk CLI is not yet publicly available.

flow-sdk deploy ./my-solution --ring dev

Coming Soon

The flow-sdk CLI is not yet publicly available.

flow-sdk deploy ./my-solution --ring staging

Coming Soon

The flow-sdk CLI is not yet publicly available.

flow-sdk promote <deployment-id> --target-ring production --notes "Promote from staging"

Automatic dev deployments

When you push code through the platform's Git integration, the dev ring is updated automatically. Configure this behavior in your solution's deployment settings.


Monitoring Deployments

Track deployment progress in the Deployments tab of any solution, or from the CLI:

Coming Soon

The flow-sdk CLI is not yet publicly available.

# List recent deployments
flow-sdk deployments list

# Check a specific deployment's status
flow-sdk deployments status <deployment-id>

The platform emits deployment events to the observability subsystem, so you can set up alerts for failed deployments, slow rollouts, or validation errors.


Next Steps

  • Rings


    Understand the ring-based rollout strategy and how to configure each environment.

    Ring configuration

  • Promotion


    Learn how to promote deployments between rings with approval gates and rollback procedures.

    Promotion workflow