AI Processors

AI processors define the prompts that NetOrca uses to automate service operations. Each processor is linked to a service, an LLM model, and an action type — together they determine when the processor runs and what it does.

Processors use the /external/{context}/ai_processors/ endpoint.

Accessed via client.ai_processors.

Action types

`action_type`	When it runs
`config`	Generates service configuration from a consumer's declaration
`verify`	Verifies a generated configuration before it is applied
`execution`	Executes an action on a deployed service item
`change_instance_validator`	Validates a change instance before it is applied
`optimiser`	Optimises a service configuration

Model properties

Property	Type	Description
`id`	`int`	Processor ID
`name`	`str`	Processor name
`action_type`	`str`	One of the action types above
`prompt`	`str`	The LLM prompt text
`service_id`	`int`	Associated service ID
`llm_model_id`	`int`	Associated LLM model ID
`active`	`bool`	Whether the processor is active
`response_schema`	`dict`	JSON Schema for structured LLM output
`is_validator`	`bool`	`True` if `action_type == "change_instance_validator"`
`created`	`datetime`	Creation timestamp
`modified`	`datetime`	Last modified timestamp

`extra_data` properties

These are accessed directly on the model and reflect the extra_data settings object returned by the API.

Property	Type	Default	Description
`extra_data`	`dict`	`{}`	Raw extra_data dict
`include_change_instance`	`bool`	`False`	Include the last change instance in the payload
`enable_pack_context`	`bool`	`False`	Enable vector document (pack_context) in the payload
`include_previous_declaration`	`bool`	`False`	Include the previous declaration for the service item
`include_previous_pipeline_data`	`list`	`[]`	Action types to include from previous pipeline in retrigger mode (`config`, `verify`, `execution`)
`include_service_config`	`bool`	`False`	Include the latest service config JSON in the payload
`enable_generative_ui`	`bool`	`False`	Enable generative UI catalog for this processor
`generative_ui_schema`	`dict`	`None`	Per-processor generative UI schema
`schedule_enabled`	`bool`	`False`	Whether scheduled execution is enabled
`schedule_crontab`	`str`	`None`	Cron expression for scheduled execution (5-part format, e.g. `'0 /6 * *'`)
`allow_auto_approval`	`bool`	`False`	Allow automatic approval of change instances
`allow_auto_rejection`	`bool`	`False`	Allow automatic rejection of change instances
`send_service_info`	`bool`	`False`	Include service info in the payload
`send_existing_service_items`	`bool`	`False`	Include existing service items in the payload

Methods

`list(**filters)`

Returns a lazy generator of all AI processors. Pagination is handled automatically.

Filters:

Filter	Type	Description
`service_id`	`int`	Filter by service
`llm_model_id`	`int`	Filter by LLM model
`action_type`	`str`	Filter by action type
`active`	`bool`	Filter by active status
`ordering`	`str`	Sort field — prefix with `-` for descending

# All processors
for processor in client.ai_processors.list():
    print(processor.name, processor.action_type, processor.active)

# Config processors only
for processor in client.ai_processors.list(action_type="config"):
    print(processor.name, processor.service_id)

# Active processors for a service
for processor in client.ai_processors.list(service_id=45, active=True):
    print(processor.name)

`filter(**filters)`

Same as list() but returns all results as a list. Accepts the same filters.

# All execution processors
processors = client.ai_processors.filter(action_type="execution")
print(f"{len(processors)} execution processors")

# Active config processors for a service
processors = client.ai_processors.filter(service_id=45, action_type="config", active=True)

`get(id)`

Retrieves a single processor by ID.

processor = client.ai_processors.get(1)

print(processor.id)
print(processor.name)
print(processor.action_type)
print(processor.prompt)
print(processor.active)
print(processor.service_id)
print(processor.created)
print(processor.modified)

`create(data)`

Creates a new AI processor and links it to a service. The processor is inactive by default unless you pass "active": True.

Field	Type	Required	Description
`name`	`str`	yes	Unique processor name
`service`	`int`	yes	Service ID to link to
`llm_model`	`int`	yes	LLM model ID to use
`action_type`	`str`	yes	Action type (see table above)
`prompt`	`str`	yes	The prompt text
`response_schema`	`dict`	no	JSON Schema for structured LLM output
`extra_data`	`dict`	no	Extra settings (see `extra_data` properties table)
`active`	`bool`	no	Enable immediately (default: `False`)

# Config processor
processor = client.ai_processors.create({
    "name": "database-service_config",
    "service": 45,
    "llm_model": 1,
    "action_type": "config",
    "prompt": "Given the following consumer request, generate a valid database configuration...",
    "active": True
})
print(processor.id)

# Change instance validator
processor = client.ai_processors.create({
    "name": "database-service_validator",
    "service": 45,
    "llm_model": 1,
    "action_type": "change_instance_validator",
    "prompt": "Validate whether the proposed change is safe to apply...",
    "extra_data": {
        "include_change_instance": True,
        "include_previous_declaration": True,
        "allow_auto_approval": True,
        "allow_auto_rejection": False
    },
    "active": True
})
print(processor.is_validator)  # True

# Processor with a structured response schema
processor = client.ai_processors.create({
    "name": "database-service_config",
    "service": 45,
    "llm_model": 1,
    "action_type": "config",
    "prompt": "Generate a database configuration from the consumer request.",
    "response_schema": {
        "type": "object",
        "properties": {
            "host":     {"type": "string"},
            "port":     {"type": "integer"},
            "database": {"type": "string"}
        },
        "required": ["host", "port", "database"]
    },
    "active": True
})

`update(id, data, partial=True)`

Updates a processor. Every update creates a history entry (see get_history()). Defaults to PATCH.

Parameter	Type	Description
`id`	`int`	Processor ID
`data`	`dict`	Fields to update
`partial`	`bool`	`True` (default) = PATCH, `False` = PUT

# Update the prompt text
client.ai_processors.update(1, {
    "prompt": "Updated prompt with improved instructions..."
})

# Deactivate a processor
client.ai_processors.update(1, {"active": False})

# Reactivate it
client.ai_processors.update(1, {"active": True})

`delete(id)`

Permanently deletes a processor.

client.ai_processors.delete(1)

`get_history(processor_id)`

Returns the full version history of a processor — every prompt update, activation change, and other modification. Useful for auditing what prompt was active at any point in time.

Parameter	Type	Description
`processor_id`	`int`	Processor ID

History entry change types: + (added), ~ (modified), - (removed).

history = client.ai_processors.get_history(processor_id=1)
for entry in history:
    print(entry)

`by_service(service_id)`

Returns all processors linked to a specific service as a list. Shortcut for filter(service_id=service_id).

processors = client.ai_processors.by_service(service_id=45)
for p in processors:
    print(p.name, p.action_type)

`get_active()`

Returns all processors where active=True as a list. Shortcut for filter(active=True).

active = client.ai_processors.get_active()
print(f"{len(active)} processors currently active")

AI Processors

Action types

Model properties

extra_data properties

Methods

list(**filters)

filter(**filters)

get(id)

create(data)

update(id, data, partial=True)

delete(id)

get_history(processor_id)

by_service(service_id)

get_active()