The strongly.manifest.yaml File

The manifest file provides declarative configuration for building and deploying your application. It must be placed in your project root alongside the Dockerfile.

Basic Structure

# strongly.manifest.yaml (place in project root)
version: "1.0"
type: nodejs  # react, nodejs, flask, rshiny, mcp_server, static, fullstack, custom
name: my-app
description: Brief description of your application
author: Your Name or Organization
homepage: https://github.com/yourorg/my-app
license: MIT
icon: "https://example.com/icon.png"
tags:
  - api
  - backend

ports:
  - port: 3000
    name: http
    expose: true
    target_port: 3000  # Optional: if different from port

env:
  - name: NODE_ENV
    value: "production"
    required: true

  - name: DATABASE_URL
    value: ""  # User must provide
    required: true
    secret: true  # Stored as Kubernetes Secret
    description: PostgreSQL connection string
    inject_as: env  # How to inject: 'env' (default), 'file', 'config'

runtime:
  command: ["node", "server.js"]
  working_dir: /app
  health_check_path: /health
  readiness_check_path: /ready  # Defaults to health_check_path if not set
  startup_timeout: 60  # Startup timeout in seconds

health_check:
  path: /health
  initial_delay: 15
  period: 30

resources:
  cpu_request: "100m"
  memory_request: "256Mi"
  cpu_limit: "500m"
  memory_limit: "512Mi"
  ephemeral_storage: "1Gi"  # Optional: ephemeral storage limit
  gpu: 1  # Optional: number of GPUs required

Application Types

The type field determines default configuration and runtime behavior:

Type	Default Port	Use Case	Special Features
`react`	3000	React/Vue SPA with client-side routing	Static build serving, SPA fallback
`nodejs`	3000	Node.js APIs, Express, NestJS	Standard Node.js runtime
`flask`	5000	Python Flask/FastAPI applications	Python WSGI/ASGI support
`rshiny`	3838	R Shiny interactive dashboards	R Shiny server runtime
`static`	80	Static HTML/CSS/JS sites	Nginx static file serving
`fullstack`	3000, 8000	Monorepo with frontend + backend	Multiple port exposure
`mcp_server`	8080	MCP protocol tool servers	MCP-specific configurations
`custom`	8080	Any other application type	No special defaults

Configuration Sections

Ports

Define which ports your application listens on:

ports:
  - port: 3000
    name: http
    expose: true  # Expose externally via ingress
    target_port: 3000  # Optional: target port if different from port
  - port: 8080
    name: metrics
    expose: false  # Internal only

Fields:

port: Port number
name: Friendly name for the port (default: "http")
expose: Whether to expose externally (default: true)
target_port: Target port if different from port (optional)

Declared Port vs Container Port

App containers run as a non-root user and always listen on the platform's fixed internal port. The port you declare in the manifest becomes the Kubernetes Service (external) port, which is mapped down to the internal port via targetPort. The platform sets the PORT environment variable to the internal port, so your application should bind to PORT rather than hard-coding a port number.

Environment Variables

Define runtime configuration with proper metadata:

env:
  # Regular environment variable
  - name: API_URL
    value: "https://api.example.com"
    required: false
    description: Backend API endpoint
    inject_as: env  # How to inject: 'env' (default), 'file', 'config'

  # Secret (passwords, API keys, tokens)
  - name: API_KEY
    value: ""  # Must be provided by user
    required: true
    secret: true  # Encrypted in Kubernetes
    description: Third-party API key

  # Build-time variable
  - name: REACT_APP_VERSION
    value: "1.0.0"
    buildtime: true  # Passed as Docker build arg
    description: App version

Fields:

name: Variable name (uppercase with underscores)
value: Default value (can be empty for required vars)
required: Whether variable must have a value (default: false)
secret: Store as Kubernetes Secret (default: false)
buildtime: Pass to Docker build as ARG (default: false)
description: User-friendly description
inject_as: How to inject the variable - env (default), file, or config

Secrets Management

Variables marked as secret: true are stored in Kubernetes Secrets and encrypted at rest. Use this for passwords, API keys, and tokens.

React/Static App Configuration

For React and static applications, configure nginx serving:

type: react
static:
  root: /app/build
  index: index.html
  spa: true  # Fallback to index.html for client-side routing
  config_file: /app/build/config.js
  config_placeholder: __APP_CONFIG__

build:
  args:
    NODE_ENV: production
  output_dir: /app/build

Static Fields:

root: Directory containing built static files
index: Index file name (default: index.html)
spa: Enable SPA routing fallback (default: true for react)
config_file: Runtime config injection file path
config_placeholder: Placeholder to replace with config

Health Checks

Your app must expose a health endpoint for Kubernetes liveness and readiness probes:

health_check:
  path: /health
  port: 3000
  initial_delay: 10  # Seconds before first check
  period: 30  # Check every 30 seconds
  timeout: 3
  failure_threshold: 3

Implementation Example (Express.js):

app.get('/health', (req, res) => {
  res.status(200).json({ status: 'ok' });
});

Implementation Example (Flask):

@app.route('/health')
def health():
    return {'status': 'ok'}, 200

Required Endpoint

Health checks are mandatory. If your app doesn't respond to health checks, Kubernetes will continuously restart it.

Runtime Configuration

Customize the container runtime:

runtime:
  command: ["node", "dist/server.js"]
  working_dir: /app
  health_check_path: /health
  readiness_check_path: /ready  # Defaults to health_check_path
  startup_timeout: 60  # Startup timeout in seconds (default: 60)
  user: appuser  # User to run as (if not set in Dockerfile)

Fields:

command: Override container CMD (string or array format)
working_dir: Working directory (default: /app)
health_check_path: Health check endpoint path (default: /health)
readiness_check_path: Readiness check path (defaults to health_check_path)
startup_timeout: Startup timeout in seconds (default: 60)
user: User to run as (optional)

Resource Requirements

Set CPU and memory limits to ensure stable performance:

resources:
  cpu_request: "100m"    # 0.1 CPU cores (default)
  memory_request: "256Mi"  # 256 MiB (default)
  cpu_limit: "500m"      # 0.5 CPU cores (default)
  memory_limit: "512Mi"  # 512 MiB (default)
  ephemeral_storage: "1Gi"  # Optional: ephemeral storage limit
  gpu: 1                  # Optional: number of GPUs required

Default Resource Values:

Field	Default
`cpu_request`	`100m`
`memory_request`	`256Mi`
`cpu_limit`	`500m`
`memory_limit`	`512Mi`
`ephemeral_storage`	not set
`gpu`	not set

Recommended Resources by App Type:

App Type	CPU Request	Memory Request	CPU Limit	Memory Limit
React/Static	50m	128Mi	200m	256Mi
Node.js API	100m	256Mi	500m	512Mi
Flask API	200m	512Mi	1000m	1Gi
Heavy Processing	500m	1Gi	2000m	2Gi

CPU Units

1000m = 1 CPU core
500m = 0.5 CPU cores
100m = 0.1 CPU cores

Proxy Configuration

Configure ingress proxy behavior:

proxy:
  websocket: true  # Enable WebSocket support
  timeout: 30  # Request timeout in seconds (default: 30)
  preserve_host: false  # Preserve original host header
  rewrite_paths:  # Path rewriting rules
    "/api": "/api/v1"
  headers:  # Additional headers to add
    X-App-Version: "1.0.0"

Fields:

websocket: Enable WebSocket proxying (default: false)
timeout: Proxy timeout in seconds (default: 30)
preserve_host: Preserve original host header (default: false)
rewrite_paths: Path rewriting rules (optional)
headers: Additional headers to add (optional)

MCP Configuration

For MCP server applications (type: mcp_server):

mcp:
  protocol_version: "1.0"
  transport: http  # 'http', 'stdio', 'sse'
  capabilities:
    - tools
    - prompts
    - resources
  authentication_required: true
  tools:
    - name: example_tool
      description: An example tool
      parameters:
        type: object
        properties:
          input:
            type: string
        required: [input]
  prompts:
    - name: analyze_data
      description: Analyzes data
      arguments:
        - name: data_source
          required: true
  resources:
    - uri: "mcp://data/*"
      name: "Data Resource"
      description: "Access to data resources"

Complete Examples

Node.js API

version: "1.0"
type: nodejs
name: user-api
description: REST API for user management
author: Acme Corp
tags: [nodejs, api, backend]

ports:
  - port: 3000
    name: http
    expose: true

env:
  - name: NODE_ENV
    value: "production"
    required: true

  - name: DATABASE_URL
    value: ""
    required: true
    secret: true
    description: PostgreSQL connection string

  - name: JWT_SECRET
    value: ""
    required: true
    secret: true
    description: Secret for JWT signing

runtime:
  command: ["node", "dist/server.js"]
  working_dir: /app
  health_check_path: /health

health_check:
  path: /health
  initial_delay: 15
  period: 30

proxy:
  websocket: true
  timeout: 30

resources:
  cpu_request: "100m"
  memory_request: "256Mi"
  cpu_limit: "500m"
  memory_limit: "512Mi"

React SPA

version: "1.0"
type: react
name: admin-dashboard
description: Admin dashboard SPA
author: Acme Corp
tags: [react, frontend]

ports:
  - port: 3000
    name: http
    expose: true

static:
  root: /app/build
  index: index.html
  spa: true
  config_file: /app/build/config.js
  config_placeholder: __APP_CONFIG__

build:
  args:
    NODE_ENV: production
    REACT_APP_VERSION: "1.0.0"
  output_dir: /app/build

env:
  - name: REACT_APP_API_URL
    value: "https://api.acme.com"
    buildtime: true
    description: Backend API endpoint

health_check:
  path: /
  initial_delay: 5
  period: 30

resources:
  cpu_request: "50m"
  memory_request: "128Mi"
  cpu_limit: "200m"
  memory_limit: "256Mi"

Flask API

version: "1.0"
type: flask
name: ml-api
description: Machine learning inference API
author: Data Team
tags: [python, flask, ml]

ports:
  - port: 5000
    name: http
    expose: true

env:
  - name: FLASK_ENV
    value: "production"
    required: true

  - name: MODEL_PATH
    value: "/app/models/model.pkl"
    description: Path to ML model file

  - name: REDIS_URL
    value: ""
    required: true
    secret: true
    description: Redis cache connection string

runtime:
  command: ["gunicorn", "-w", "4", "-b", "0.0.0.0:5000", "app:app"]
  working_dir: /app
  health_check_path: /health

health_check:
  path: /health
  port: 5000
  initial_delay: 20
  period: 30
  timeout: 5

resources:
  cpu_request: "200m"
  memory_request: "512Mi"
  cpu_limit: "1000m"
  memory_limit: "1Gi"

R Shiny Dashboard

version: "1.0"
type: rshiny
name: analytics-dashboard
description: Interactive R Shiny analytics dashboard
author: Data Team
tags: [rshiny, analytics, dashboard]

ports:
  - port: 3838
    name: http
    expose: true

env:
  - name: SHINY_LOG_LEVEL
    value: "INFO"
    description: Shiny log level

runtime:
  working_dir: /app
  health_check_path: /health

health_check:
  path: /
  initial_delay: 15
  period: 30

resources:
  cpu_request: "200m"
  memory_request: "512Mi"
  cpu_limit: "1000m"
  memory_limit: "1Gi"

MCP Server

version: "1.0"
type: mcp_server
name: my-mcp-server
description: MCP server providing custom tools
author: Your Organization
homepage: https://github.com/yourorg/mcp-server
license: MIT
tags: [mcp, tools]

ports:
  - port: 8080
    name: http
    expose: true

env:
  - name: LOG_LEVEL
    value: "info"
    description: Logging level

runtime:
  command: ["python", "server.py"]
  working_dir: /app
  health_check_path: /health
  startup_timeout: 60

health_check:
  path: /health
  initial_delay: 10
  period: 30

resources:
  cpu_request: "100m"
  memory_request: "256Mi"
  cpu_limit: "500m"
  memory_limit: "512Mi"

mcp:
  protocol_version: "1.0"
  transport: http
  capabilities: [tools, prompts, resources]
  authentication_required: true
  tools:
    - name: example_tool
      description: An example tool that processes data
      parameters:
        type: object
        properties:
          input:
            type: string
            description: Input data to process
        required: [input]

Validation

The platform validates your manifest during deployment. Common errors:

Missing required field: type is the only required field (version defaults to "1.0")
Invalid type: Must be one of the 8 supported types (react, nodejs, static, fullstack, flask, rshiny, mcp_server, custom)
Wrong field types: Fields must match their expected types (e.g., port must be an integer, env must be a list)
Invalid resource format: Resource strings must use Kubernetes format (e.g., "500m", "1Gi"); invalid formats are rejected when the deployment is created

Basic Structure​

Application Types​

Configuration Sections​

Ports​

Environment Variables​

React/Static App Configuration​

Health Checks​

Runtime Configuration​

Resource Requirements​

Proxy Configuration​

MCP Configuration​

Complete Examples​

Node.js API​

React SPA​

Flask API​

R Shiny Dashboard​

MCP Server​

Validation​

Next Steps​

Basic Structure

Application Types

Configuration Sections

Ports

Environment Variables

React/Static App Configuration

Health Checks

Runtime Configuration

Resource Requirements

Proxy Configuration

MCP Configuration

Complete Examples

Node.js API

React SPA

Flask API

R Shiny Dashboard

MCP Server

Validation

Next Steps