Sevro — Open-source low-code backend

Modules are the unit of organization in Sevro. A module contains entities, workflows, and triggers. You define modules as YAML files and apply them with the CLI.

Module structure

name: my_module
entities:
  - ref: entity_name
    label: Entity Name
    description: Optional description
    attributes: [...]
    methods: [...]
    triggers: [...]
workflows:
  - name: workflow_name
    nodes: [...]
    connections: {...}

Entities

An entity is a type of thing in your system. Each entity has a ref (machine name), label (display name), and optional description.

entities:
  - ref: contact
    label: Contact
    description: A person in your CRM
    attributes: [...]
    methods: [...]

Attributes

Attributes define the data shape of an entity.

Field reference

Field	Type	Required	Description
`ref`	string	Yes	Machine name (snake_case)
`label`	string	Yes	Display name
`description`	string	No	Human-readable description
`type`	string	Yes	One of: `string`, `number`, `boolean`, `date`, `datetime`, `enum`, `reference`
`is_required`	boolean	No	Reject saves when field is missing (default: false)
`is_unique`	boolean	No	Enforce uniqueness across all records of this entity (default: false)
`default_value`	string	No	Default value applied on create if field is absent
`allowed_values`	[]string	No	For `enum` type — the valid values
`target_entity_ref`	string	No	For `reference` type — the entity this points to
`on_delete`	string	No	For `reference` type — `nullify`, `restrict`, or `cascade` (default: `nullify`)
`computed_expression`	string	No	CEL expression — value is computed, read-only
`validation_expression`	string	No	CEL expression — must return true or save is rejected

Attribute types

Type	Stored as	Example
`string`	string	`"hello"`
`number`	int64 or float64	`42`, `3.14`
`boolean`	bool	`true`
`date`	string (RFC 3339 date)	`"2024-01-15"`
`datetime`	string (RFC 3339 datetime)	`"2024-01-15T10:30:00Z"`
`enum`	string (must be in `allowed_values`)	`"draft"`
`reference`	string (record ID)	`"uuid-of-target"`

Examples

attributes:
  - ref: email
    label: Email
    type: string
    is_required: true
    is_unique: true
    validation_expression: "record.email.contains('@')"

  - ref: status
    label: Status
    type: enum
    allowed_values: ["draft", "published", "archived"]
    default_value: "draft"

  - ref: created_at
    label: Created At
    type: datetime
    computed_expression: "now()"

  - ref: author_id
    label: Author
    type: reference
    target_entity_ref: user
    on_delete: nullify

Methods

Methods define how data changes. Every mutation goes through a method — there are no direct create/update endpoints.

Field reference

Field	Type	Required	Description
`ref`	string	Yes	Machine name
`label`	string	Yes	Display name
`is_constructor`	boolean	No	If true, this method creates new records
`conditions`	[]string	No	CEL expressions — all must be true to proceed
`parameters`	[]object	No	Input parameters the caller must provide
`actions`	[]object	Yes	Field assignments to apply

Parameters

Field	Type	Required	Description
`ref`	string	Yes	Parameter name
`label`	string	Yes	Display name
`type`	string	Yes	One of: `string`, `number`, `boolean`, `date`, `datetime`, `enum`
`is_required`	boolean	No	Caller must provide this value
`allowed_values`	[]string	No	For `enum` type

Actions

Field	Type	Description
`target`	string	The attribute ref to set
`expression`	string	CEL expression — the value to assign

Examples

methods:
  # Constructor — creates a new invoice
  - ref: create_invoice
    label: Create Invoice
    is_constructor: true
    parameters:
      - ref: customer_name
        label: Customer Name
        type: string
        is_required: true
      - ref: amount
        label: Amount
        type: number
        is_required: true
    actions:
      - target: customer_name
        expression: "params.customer_name"
      - target: amount
        expression: "params.amount"
      - target: status
        expression: '"draft"'

  # Method — marks an invoice as paid
  - ref: mark_paid
    label: Mark Paid
    conditions:
      - 'record.status == "draft"'
    actions:
      - target: status
        expression: '"paid"'
      - target: paid_at
        expression: "now()"

  # Method — cancels an invoice
  - ref: cancel
    label: Cancel
    conditions:
      - 'record.status != "paid"'
    actions:
      - target: status
        expression: '"cancelled"'

CEL context

In conditions and actions, you have access to:

record — the current record's fields (e.g., record.status, record.amount)
params — the method parameters (e.g., params.customer_name)
now() — current timestamp
uuid() — generate a UUID

See the CEL reference for the full expression language.

Triggers

Triggers fire workflows automatically when events happen.

triggers:
  - event_type: on_create
    workflow_ref: notify_new_contact
  - event_type: on_update
    workflow_ref: sync_contact
    field_watcher: "email"  # only fires when the email field changes
  - event_type: on_delete
    workflow_ref: cleanup_contact_data

Cron triggers

triggers:
  - event_type: on_create  # event_type is required but ignored for cron
    workflow_ref: daily_report
    cron_expression: "0 9 * * *"  # every day at 9am

Workflows

Workflows are DAGs (directed acyclic graphs) of nodes connected by edges. They orchestrate multi-step processes that span entities, modules, and external services.

Node types

Type	Description	Key parameters
`invoke_method`	Call a method on a record	`module`, `entity`, `method`, `record_ref`, `arguments`
`invoke_connection`	Call an external API	`connection`, `action`, `arguments`
`evaluate`	Evaluate CEL expressions	`expressions` (map of name → CEL)
`branch`	Conditional routing	`condition` (CEL, must return bool)
`sleep`	Pause for a duration	`duration` (e.g., `"30s"`, `"5m"`)

Error handling

Each node declares an on_error policy:

on_error:
  abort: true        # stop the workflow (default)
  # OR
  skip: true         # log error, continue
  # OR
  retry:
    max_attempts: 3
    backoff: exponential  # constant | linear | exponential
    initial_delay: 1s
    max_delay: 30s

Example

workflows:
  - name: order_fulfillment
    description: "Fulfill an order end-to-end"
    input:
      type: object
      properties:
        order_id:
          type: string
    nodes:
      - id: validate
        name: "Validate Order"
        type: invoke_method
        parameters:
          module: commerce
          entity: Order
          method: validate
          record_ref: "$.input.order_id"
        on_error:
          abort: true
      - id: check_stock
        name: "Check Stock"
        type: invoke_method
        parameters:
          module: warehouse
          entity: Inventory
          method: check_availability
        on_error:
          abort: true
      - id: in_stock
        name: "In Stock?"
        type: branch
        parameters:
          condition: "$nodes.check_stock.output.available == true"
        on_error:
          abort: true
      - id: reserve
        name: "Reserve Inventory"
        type: invoke_method
        parameters:
          module: warehouse
          entity: Inventory
          method: reserve
        on_error:
          abort: true
      - id: notify_oos
        name: "Notify Out of Stock"
        type: invoke_method
        parameters:
          module: commerce
          entity: Order
          method: mark_backordered
          record_ref: "$.input.order_id"
        on_error:
          skip: true
    connections:
      validate:
        main:
          - node: check_stock
      check_stock:
        main:
          - node: in_stock
      in_stock:
        true:
          - node: reserve
        false:
          - node: notify_oos

Data flow

Nodes reference data from earlier nodes using CEL with path syntax:

$.input.<field> — workflow input
$nodes.<node_id>.output.<field> — output of a previous node
$nodes.<node_id>.error — error message if the node failed

Applying modules

# Create or update a module from a YAML file
sevro module apply -f module.yaml

# Export an existing module to YAML
sevro module export <module-id> -o module.yaml

# List modules
sevro module list

# Delete a module
sevro module delete <module-id>

sevro module apply is idempotent — if a module with the same name exists, it updates it; otherwise it creates a new one.

Import/export

Modules can be exported and imported via the admin UI or CLI:

CLI: sevro module export <id> -o file.yaml → sevro module apply -f file.yaml
UI: Module list → Export dropdown (JSON or YAML) → Import button

The YAML format uses snake_case field names matching the proto definitions. Exported YAML strips server-generated fields (IDs, timestamps) so it can be re-applied cleanly.

Module Authoring

Module structure

Entities

Attributes

Field reference

Attribute types

Examples

Methods

Field reference

Parameters

Actions

Examples

CEL context

Triggers

Cron triggers

Workflows

Node types

Error handling

Example

Data flow

Applying modules

Import/export

On this page