Variables Copier

Chaque question que le template pose au moment de copier copy et copier update, groupée comme copier.yml les groupe. Les réponses sont enregistrées dans le .copier-answers.yml de l’acteur — copier update les réutilise, donc un défaut enregistré une fois ne se redérive jamais aux mises à jour suivantes (l’exception est resource_profile: auto, qui se résout à nouveau au rendu). Passez les réponses de façon non interactive avec --data key=value ; acceptez tous les défauts avec --defaults.

Il n’existe pas de variable actor_id — l’UUID marketplace est attribué par la plateforme à l’enregistrement/publication, pas par le template.

Paramètres du modèle (clés privées — jamais demandées)

Clé	Valeur
`_min_copier_version`	`9.4.0`
`_subdirectory`	`template`
`_templates_suffix`	`.jinja`
`_exclude`	`SCHEMA_README.md`
`_message_after_copy`	`Generated {{ project_name }}. Publishing this actor to GitHub is what puts it in the fleet. From your kaboum checkout: s…`

Copier lui-même doit être ≥ 9.4.0 (_min_copier_version) — épinglé à la plus ancienne version réellement testée par la CI ; les revendications plus anciennes (9.0.0) ne se sont jamais installées sur Python 3.12.

Identité et nommage

Variable	Type	Défaut	Options/Notes
`project_name`	`str`	(none — required)	Lowercase project name (e.g., kalendar, kamail) — must match `^[a-z][a-z0-9_]{1,30}$`
`actor_display_name`	`str`	`{{ project_name \| replace('_', ' ') \| title }}`	Human-readable name (e.g., Kalendar, Kamail)
`actor_description`	`str`	`AI Assistant actor for the Kaoohi platform`	One-line description (e.g., ‘AI Booking Assistant’)
`domain_module`	`str`	(none — required)	Domain subdirectory name (e.g., booking, email) — must match `^[a-z][a-z0-9_]{1,20}$` — reserved: `mesh`, `agent`, `api`, `core`, `kaoohi`, `shared`, `integrations`, `middleware`, `models`, `storage`
`domain_display_name`	`str`	`{{ domain_module \| replace('_', ' ') \| title }}`	Human-readable domain name (e.g., Booking, Email)
`config_class_name`	`str`	`{{ actor_display_name \| replace(' ', '') \| replace('-', '') \| replace('_', '') }}Config`	Config class name (e.g., KalendarConfig)
`agent_class_name`	`str`	`{{ actor_display_name \| replace(' ', '') \| replace('-', '') \| replace('_', '') }}Agent`	Agent class name (e.g., KalendarAgent)
`python_version`	`str`	`3.12`	`3.11` / `3.12` / `3.13` — Python version

Framework d’agent et intelligence

Variable	Type	Défaut	Options/Notes	Condition d’activation
`agent_framework`	`str`	`deepagents`	`deepagents` / `langchain` — Agent base: ‘deepagents’ (planning/sub-agents) or ‘langchain’ (plain create_react_agent)
`include_planning`	`bool`	`true`	[deepagents only] Add the planning/todo middleware (write_todos)? Opt-in since deepagents 0.7. Adds tokens; keep ON for multi-step actors.	`agent_framework == 'deepagents'`
`include_subagents`	`bool`	`true`	[deepagents only] Enable scoped sub-agents (the `task` tool)? Highest token cost (fan-out); enable only if delegation is needed.	`agent_framework == 'deepagents'`
`include_reflection`	`bool`	`true`	[deepagents only] Enable the SummarizationMiddleware context/reflection pass? Default ON for parity with current behavior.	`agent_framework == 'deepagents'`
`extra_middleware`	`str`	`""`	[deepagents only] Comma-separated ‘module.path:ClassName’ middleware appended to the create_deep_agent stack. Blank = none.	`agent_framework == 'deepagents'`
`include_lean_channel`	`bool`	`false`	[deepagents only] Route one metadata channel to a lean plain-tool agent (langchain create_agent) for latency-critical surfaces?	`agent_framework == 'deepagents'`
`lean_channel_name`	`str`	`voice`	Metadata channel value that selects the lean agent (invoke with metadata {“channel”: this value}).	`include_lean_channel and agent_framework == 'deepagents'`
`lean_channel_tools`	`str`	`""`	Comma-separated tool names the lean agent may call (subset of the actor’s tools). Blank = every tool.	`include_lean_channel and agent_framework == 'deepagents'`

Intégration

Variable	Type	Défaut	Options/Notes	Condition d’activation
`integration_type`	`str`	`none`	`none` / `n8n` / `custom` — Integration plugin: ‘none’, ‘n8n’, or ‘custom’
`include_n8n_mcp`	`bool`	`true`	[n8n] Connect to n8n-mcp for node docs, templates, and validation?	`integration_type == 'n8n'`
`n8n_mcp_url`	`str`	`http://n8n-mcp:3000/mcp`	[n8n] n8n-mcp HTTP server URL (streamable_http transport)	`integration_type == 'n8n'`
`include_workflow_executor`	`bool`	`true`	[n8n] Enable dynamic workflow-as-tool discovery (calls n8n webhooks)?	`integration_type == 'n8n'`
`n8n_workflow_tag`	`str`	`kaoohi-tool`	[n8n] Tag that marks n8n workflows as callable tools	`integration_type == 'n8n'`
`n8n_allow_workflow_authoring`	`bool`	`false`	[n8n] Let the agent CREATE/UPDATE/DELETE/ACTIVATE workflows? Workflows can run arbitrary code — leave false unless the actor genuinely authors workflows	`integration_type == 'n8n'`
`workflow_sync_interval`	`int`	`5`	[n8n] Minutes between workflow tool re-discovery	`integration_type == 'n8n'`

Infra et fonctionnalités

Variable	Type	Défaut	Options/Notes	Condition d’activation
`fleet_mesh`	`bool`	`false`	Include the inter-actor mesh client (lazy peer resolution via the kaoohi group resolver)? Enable for actors that belong to a suite.
`mesh_depends_on`	`str`	`""`	Peer roles this actor requires, comma-separated (e.g. ‘portefeuille’). Must match the bundle seed’s depends_on_mesh_role. Empty for a standalone actor.	`fleet_mesh`
`include_redis_reply_queue`	`bool`	`true`	Include Redis reply queue for WhatsApp/Evolution API delivery?
`include_scheduler`	`bool`	`false`	Include background scheduler for periodic tasks?
`include_panel`	`bool`	`false`	Include the owner-facing UI panel scaffold (served through the Kaoohi panel relay)?
`panel_slug`	`str`	`{{ domain_module \| replace('_', '-') }}`	[panel] URL slug and ui_panels id (e.g. ‘pipeline’, ‘inbox’) — must match `^[a-z][a-z0-9-]{1,30}$`	`include_panel`
`panel_title`	`str`	`{{ domain_display_name }} Panel`	[panel] Human-readable tab title shown in the chat workspace pane	`include_panel`
`declare_readiness_probe`	`bool`	`true`	Declare a readiness probe at /ready, separate from the liveness path? Without it the platform points liveness, readiness AND startup at /health, which must always return 200 — so readiness gates nothing and a pod that ca…
`include_domain_tools`	`bool`	`true`	Render a domain-tools path in the orchestrator (_create_tools)?
`docker_workers`	`int`	`1`	`1` / `2` / `4` — Number of uvicorn workers in Docker CMD
`resource_profile`	`str`	`auto`	`auto` / `nano` / `small` / `standard` / `large` — Pod size — ‘auto’ derives from your answers (docker_workers=4 → large, else standard: 250m/1000m + 256Mi/512Mi). Manual tiers for known-lean or known-heavy actors.

Auteur

Variable	Type	Défaut	Options/Notes	Condition d’activation
`author_name`	`str`	`Kaoohi Team`	Author name
`author_email`	`str`	`dev@kaoohi.com`	Author email

Avancé

Variable	Type	Défaut	Options/Notes	Condition d’activation
`include_actor_config_mixin`	`bool`	`false`	Keep actor-specific Settings fields in an actor-owned src/config_actor.py (mixed into the config class; copier update never conflicts there)? Scaffolds the file once.
`extra_config_fields_json`	`str`	`""`	[advanced] JSON {sections,items} merged into the emitted config.json (used by kaoohi eject; leave empty)

Métadonnées marketplace

Variable	Type	Défaut	Options/Notes	Condition d’activation
`kaoohi_category`	`str`	`deep_agent`	`deep_agent` / `productivity` / `communication` / `business` / `data` — Marketplace category
`kaoohi_icon`	`str`	`""`	Icon — emoji or asset path (e.g., ‘📅’ or ‘icons/myactor.svg’)
`kaoohi_image_repo`	`str`	`ghcr.io/allcare/{{ project_name \| replace('_', '-') }}`	GHCR image base: ghcr.io/allcare/.
`kaoohi_required_connectors`	`str`	`{% if integration_type == 'n8n' %}n8n{% endif %}`	Required connectors, comma-separated (e.g., ‘google_calendar’ or ‘gmail,whatsapp’). Leave empty for none.
`kaoohi_optional_connectors`	`str`	`whatsapp`	Optional connectors, comma-separated. Leave empty for none.
`kaoohi_capabilities`	`str`	`chat`	Actor capabilities, comma-separated (e.g., ‘scheduling,reminders’)
`kaoohi_expose_external_api`	`bool`	`true`	Expose this actor’s API publicly via the platform-managed edge (https://<PLATFORM_URL>/api/v1/k//…)? Each kompanion gets a public URL + API key. The pod still runs auth-off — kaoohi is the authorization boundary.…
`kaoohi_external_api_operations`	`str`	`chat,status,health`	[external API only] Which operations to expose, comma-separated. Only idempotent, safe operations — settings/oauth are never eligible.	`kaoohi_expose_external_api`
`kaoohi_external_api_protocols`	`str`	`rest,mcp`	[external API only] Transports to expose, comma-separated: rest, mcp. REST is the HTTP API; mcp also exposes this kompanion as an MCP tool for AI agents.	`kaoohi_expose_external_api`

Voir aussi

documentation/05-configuration-reference.md — la configuration côté acteur que le template émet.
documentation/deepagents-intelligence-toggles.md dans le dépôt kaboum (référencé depuis copier.yml) — coûts et latence mesurés des interrupteurs d’intelligence.