iobroker.pi-agent 0.0.0-stage → 0.0.2

package/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Alex <sadam6752@gmail.com>
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

package/README.md

@@ -1,3 +1,134 @@
-# Temporary Holding Version
+![Logo](admin/pi-agent.png)
 
-This version is a temporary placeholder for this package. An operational version to replace this has been submitted for review and is awaiting a staged release.
+# ioBroker.pi-agent
+
+[![NPM version](https://img.shields.io/npm/v/iobroker.pi-agent.svg)](https://www.npmjs.com/package/iobroker.pi-agent)
+[![Downloads](https://img.shields.io/npm/dm/iobroker.pi-agent.svg)](https://www.npmjs.com/package/iobroker.pi-agent)
+[![License](https://img.shields.io/github/license/sadam6752-tech/ioBroker.pi-agent.svg)](LICENSE)
+
+[![NPM](https://nodei.co/npm/iobroker.pi-agent.png?downloads=true)](https://nodei.co/npm/iobroker.pi-agent/)
+
+![Test and Release](https://github.com/sadam6752-tech/ioBroker.pi-agent/workflows/Test%20and%20Release/badge.svg)
+
+A language model agent for ioBroker that survives restarts and crashes. It is built on
+[Pi Durable](https://earendil.com/posts/pi-durable/): every model call and tool call is stored as a checkpoint
+before it runs, so an interrupted request continues after the adapter starts again instead of starting over.
+
+## What it does
+
+- Reads and searches states through the tools `search_objects`, `get_state` (one id or a `*` pattern) and
+  `list_instances`. These tools may be repeated safely.
+- `list_instances` starts from the instance objects and reads `alive` and `connected` of each instance by its exact id.
+  A `*` pattern only finds states that have an object, so a stale `alive` value of an instance whose state object is
+  gone is missing from a pattern query but not from `list_instances`.
+- Logs every tool call as one line: `Tool <name> <arguments> -> <outcome>` (blocked writes as a warning).
+- Writes states through `set_state`. This tool is **never repeated** after an interruption: the model is told that the
+  call was interrupted and checks the actual value first. A light cannot be switched twice by a restart.
+- Writes only to states on the **write allowlist**. Everything else is blocked before it runs.
+- Answers each request exactly once. The request id is derived from the triggering state, so a repeated trigger finds the
+  original request.
+
+## Requirements
+
+- Node.js 22.19.0 or newer, js-controller 6.0.11 or newer, admin 7.6.20 or newer.
+- An API key of the chosen provider: Anthropic, OpenAI or DeepSeek.
+- The adapter runs as its own process (not in compact mode). It keeps long-running model requests and a conversation
+  store open, and its dependencies are large; running it apart keeps a fault or a memory peak away from other adapters.
+  The API key is kept in memory of this instance only and is never put into the process environment.
+
+## Configuration
+
+| Setting         | Meaning                                                                          |
+| --------------- | -------------------------------------------------------------------------------- |
+| Provider, Model | The language model. Empty model: the default of the provider (`claude-sonnet-5-5`, `gpt-6-sol`, `deepseek-flash`). |
+| API key         | Stored encrypted.                                                                |
+| Dry run         | On by default. Write requests are logged and answered, but not executed.         |
+| Write allowlist | One regular expression per line. Empty means the agent cannot write anything.   |
+| Instructions    | Text appended to the system prompt, for example room names and house rules.     |
+| Telegram instance | A `telegram` instance whose bot talks to the agent. Empty: Telegram is not used. |
+| Prefix          | Only messages that start with it go to the agent, e.g. `/ai`. Empty: every message of an allowed user. |
+| Allowed Telegram users | Telegram user ID, user name or first name (as in the telegram adapter), one per line. Messages from anyone else are ignored. |
+
+Example allowlist:
+
+```
+^hm-rpc\.0\.Wohnzimmer\.
+^javascript\.0\.agent\.
+```
+
+Start with dry run on, watch the log, and only then switch it off.
+
+## Usage
+
+Write a text to `pi-agent.0.chat.input`. The answer appears in `pi-agent.0.chat.answer`. From a script:
+
+```js
+sendTo('pi-agent.0', 'ask', { text: 'Is the living room light on?' }, result => log(result.answer));
+```
+
+The answer of a `sendTo` request comes back only through the callback (`result.answer`, or `result.error`); it is not
+written to `chat.answer`.
+
+`pi-agent.0.chat.pending` lists the requests that were accepted but not answered yet. If the adapter stops in the
+middle, they continue on the next start. The answer of a request from `chat.input` or `sendTo` is then written to
+`chat.answer` (the `sendTo` caller is no longer waiting), the answer of a Telegram request goes to its chat. A
+`sendTo` caller that was interrupted gets `{ error: "Interrupted by shutdown; the request continues after the next start" }`.
+
+### Telegram
+
+Select a `telegram` instance, enter the allowed users and, if the bot also serves other scripts, a prefix such as
+`/ai`. Then write to the bot: `/ai Is the living room light on?`
+
+- Every chat has its own conversation with its own history; `pi-agent.0.telegram.chats` maps chat ids to them.
+- Messages from users who are not on the list are ignored and logged as a warning. An empty list allows nobody.
+- A first name such as `Alex` counts only for users the telegram adapter knows (authenticated there), because anybody
+  can choose any first name in Telegram. Two known users with the same first name are both allowed; use the user name
+  or the ID to tell them apart.
+- A message that the telegram adapter delivers twice is answered once.
+- Long answers are split into several messages.
+
+### DeepSeek
+
+Select **DeepSeek** as provider and enter a key from the DeepSeek platform. Available models: `deepseek-flash`
+(default, fast) and `deepseek-v4-pro`. Requests go to `https://api.deepseek.com`.
+
+## Security
+
+- The model decides which tools to call. The allowlist and the dry run are your protection, not the prompt.
+- The text of your requests and the values the agent reads are sent to the provider. Do not give it access to states
+  that contain secrets.
+- The conversation is stored unencrypted in the instance data directory.
+
+## Changelog
+
+### 0.0.2 (2026-10-25)
+
+- Telegram answers keep their formatting: Markdown of the model is sent as Telegram HTML (bold, italic, code, links;
+  tables as a monospace block), and Telegram chats are asked for short answers without tables
+- Released through the CI workflow with npm provenance (trusted publishing)
+
+### 0.0.1 (2026-10-25)
+
+- Initial version: durable agent with the tools `search_objects`, `get_state`, `list_instances` and `set_state`
+- Write allowlist and dry run; an interrupted write is never repeated; requests continue after a restart
+- Telegram: talk to the agent through a telegram instance, with allowed users, prefix and one conversation per chat
+- Providers: Anthropic, OpenAI, DeepSeek; every tool call is logged
+
+## License
+
+MIT License
+
+Copyright (c) 2026 Alex <sadam6752@gmail.com>
+
+Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated
+documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to
+whom the Software is furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all copies or substantial portions of the
+Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE
+WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR
+COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

package/admin/i18n/de.json

@@ -0,0 +1,19 @@
+{
+	"Provider": "Anbieter",
+	"Model": "Modell",
+	"API key": "API-Schlüssel",
+	"Dry run": "Probelauf",
+	"Log write requests instead of executing them": "Schreibanfragen nur protokollieren, nicht ausführen",
+	"Leave empty to use the default model of the provider": "Leer lassen, um das Standardmodell des Anbieters zu verwenden",
+	"Write allowlist": "Schreib-Freigabeliste",
+	"One regular expression per line. The agent may only write states whose ID matches one of them.": "Ein regulärer Ausdruck pro Zeile. Der Agent darf nur Datenpunkte schreiben, deren ID zu einem davon passt.",
+	"Instructions": "Anweisungen",
+	"Additional instructions for the agent": "Zusätzliche Anweisungen für den Agenten",
+	"Telegram": "Telegram",
+	"Telegram instance": "Telegram-Instanz",
+	"Empty: the agent does not use Telegram": "Leer: Der Agent nutzt Telegram nicht",
+	"Prefix": "Präfix",
+	"Only messages that start with this text go to the agent, e.g. /ai. Empty: every message of an allowed user": "Nur Nachrichten, die mit diesem Text beginnen, gehen an den Agenten, z. B. /ai. Leer: jede Nachricht eines erlaubten Nutzers",
+	"Allowed Telegram users": "Erlaubte Telegram-Nutzer",
+	"Telegram user ID, user name or first name (as in the telegram adapter), one per line. Messages from anyone else are ignored.": "Telegram-Nutzer-ID, Benutzername oder Vorname (wie im Telegram-Adapter), einer pro Zeile. Nachrichten von allen anderen werden ignoriert."
+}

package/admin/i18n/en.json

@@ -0,0 +1,19 @@
+{
+	"Provider": "Provider",
+	"Model": "Model",
+	"API key": "API key",
+	"Dry run": "Dry run",
+	"Log write requests instead of executing them": "Log write requests instead of executing them",
+	"Leave empty to use the default model of the provider": "Leave empty to use the default model of the provider",
+	"Write allowlist": "Write allowlist",
+	"One regular expression per line. The agent may only write states whose ID matches one of them.": "One regular expression per line. The agent may only write states whose ID matches one of them.",
+	"Instructions": "Instructions",
+	"Additional instructions for the agent": "Additional instructions for the agent",
+	"Telegram": "Telegram",
+	"Telegram instance": "Telegram instance",
+	"Empty: the agent does not use Telegram": "Empty: the agent does not use Telegram",
+	"Prefix": "Prefix",
+	"Only messages that start with this text go to the agent, e.g. /ai. Empty: every message of an allowed user": "Only messages that start with this text go to the agent, e.g. /ai. Empty: every message of an allowed user",
+	"Allowed Telegram users": "Allowed Telegram users",
+	"Telegram user ID, user name or first name (as in the telegram adapter), one per line. Messages from anyone else are ignored.": "Telegram user ID, user name or first name (as in the telegram adapter), one per line. Messages from anyone else are ignored."
+}

package/admin/i18n/es.json

@@ -0,0 +1,19 @@
+{
+	"Provider": "Proveedor",
+	"Model": "Modelo",
+	"API key": "Clave de API",
+	"Dry run": "Ejecución de prueba",
+	"Log write requests instead of executing them": "Registrar las solicitudes de escritura en lugar de ejecutarlas",
+	"Leave empty to use the default model of the provider": "Dejar vacío para usar el modelo predeterminado del proveedor",
+	"Write allowlist": "Lista de escritura permitida",
+	"One regular expression per line. The agent may only write states whose ID matches one of them.": "Una expresión regular por línea. El agente solo puede escribir estados cuyo ID coincida con alguna de ellas.",
+	"Instructions": "Instrucciones",
+	"Additional instructions for the agent": "Instrucciones adicionales para el agente",
+	"Telegram": "Telegram",
+	"Telegram instance": "Instancia de Telegram",
+	"Empty: the agent does not use Telegram": "Vacío: el agente no usa Telegram",
+	"Prefix": "Prefijo",
+	"Only messages that start with this text go to the agent, e.g. /ai. Empty: every message of an allowed user": "Solo los mensajes que empiezan con este texto van al agente, p. ej. /ai. Vacío: todos los mensajes de un usuario permitido",
+	"Allowed Telegram users": "Usuarios de Telegram permitidos",
+	"Telegram user ID, user name or first name (as in the telegram adapter), one per line. Messages from anyone else are ignored.": "ID, nombre de usuario o nombre de pila de Telegram (como en el adaptador telegram), uno por línea. Los mensajes de cualquier otra persona se ignoran."
+}

package/admin/i18n/fr.json

@@ -0,0 +1,19 @@
+{
+	"Provider": "Fournisseur",
+	"Model": "Modèle",
+	"API key": "Clé API",
+	"Dry run": "Essai à blanc",
+	"Log write requests instead of executing them": "Journaliser les demandes d'écriture au lieu de les exécuter",
+	"Leave empty to use the default model of the provider": "Laisser vide pour utiliser le modèle par défaut du fournisseur",
+	"Write allowlist": "Liste d'écriture autorisée",
+	"One regular expression per line. The agent may only write states whose ID matches one of them.": "Une expression régulière par ligne. L'agent ne peut écrire que les états dont l'ID correspond à l'une d'elles.",
+	"Instructions": "Consignes",
+	"Additional instructions for the agent": "Consignes supplémentaires pour l'agent",
+	"Telegram": "Telegram",
+	"Telegram instance": "Instance Telegram",
+	"Empty: the agent does not use Telegram": "Vide : l'agent n'utilise pas Telegram",
+	"Prefix": "Préfixe",
+	"Only messages that start with this text go to the agent, e.g. /ai. Empty: every message of an allowed user": "Seuls les messages commençant par ce texte vont à l'agent, p. ex. /ai. Vide : tous les messages d'un utilisateur autorisé",
+	"Allowed Telegram users": "Utilisateurs Telegram autorisés",
+	"Telegram user ID, user name or first name (as in the telegram adapter), one per line. Messages from anyone else are ignored.": "ID, nom d'utilisateur ou prénom Telegram (comme dans l'adaptateur telegram), un par ligne. Les messages de toute autre personne sont ignorés."
+}

package/admin/i18n/it.json

@@ -0,0 +1,19 @@
+{
+	"Provider": "Fornitore",
+	"Model": "Modello",
+	"API key": "Chiave API",
+	"Dry run": "Prova a secco",
+	"Log write requests instead of executing them": "Registra le richieste di scrittura invece di eseguirle",
+	"Leave empty to use the default model of the provider": "Lasciare vuoto per usare il modello predefinito del fornitore",
+	"Write allowlist": "Elenco di scrittura consentita",
+	"One regular expression per line. The agent may only write states whose ID matches one of them.": "Un'espressione regolare per riga. L'agente può scrivere solo gli stati il cui ID corrisponde a una di esse.",
+	"Instructions": "Istruzioni",
+	"Additional instructions for the agent": "Istruzioni aggiuntive per l'agente",
+	"Telegram": "Telegram",
+	"Telegram instance": "Istanza Telegram",
+	"Empty: the agent does not use Telegram": "Vuoto: l'agente non usa Telegram",
+	"Prefix": "Prefisso",
+	"Only messages that start with this text go to the agent, e.g. /ai. Empty: every message of an allowed user": "Solo i messaggi che iniziano con questo testo vanno all'agente, ad es. /ai. Vuoto: ogni messaggio di un utente consentito",
+	"Allowed Telegram users": "Utenti Telegram consentiti",
+	"Telegram user ID, user name or first name (as in the telegram adapter), one per line. Messages from anyone else are ignored.": "ID, nome utente o nome Telegram (come nell'adattatore telegram), uno per riga. I messaggi di chiunque altro vengono ignorati."
+}

package/admin/i18n/nl.json

@@ -0,0 +1,19 @@
+{
+	"Provider": "Aanbieder",
+	"Model": "Taalmodel",
+	"API key": "API-sleutel",
+	"Dry run": "Proefrun",
+	"Log write requests instead of executing them": "Schrijfverzoeken alleen loggen en niet uitvoeren",
+	"Leave empty to use the default model of the provider": "Leeg laten om het standaardmodel van de aanbieder te gebruiken",
+	"Write allowlist": "Toegestane schrijflijst",
+	"One regular expression per line. The agent may only write states whose ID matches one of them.": "Eén reguliere expressie per regel. De agent mag alleen states schrijven waarvan de ID overeenkomt met een ervan.",
+	"Instructions": "Instructies",
+	"Additional instructions for the agent": "Aanvullende instructies voor de agent",
+	"Telegram": "Telegram",
+	"Telegram instance": "Telegram-instantie",
+	"Empty: the agent does not use Telegram": "Leeg: de agent gebruikt Telegram niet",
+	"Prefix": "Voorvoegsel",
+	"Only messages that start with this text go to the agent, e.g. /ai. Empty: every message of an allowed user": "Alleen berichten die met deze tekst beginnen gaan naar de agent, bijv. /ai. Leeg: elk bericht van een toegestane gebruiker",
+	"Allowed Telegram users": "Toegestane Telegram-gebruikers",
+	"Telegram user ID, user name or first name (as in the telegram adapter), one per line. Messages from anyone else are ignored.": "Telegram-gebruikers-ID, gebruikersnaam of voornaam (zoals in de telegram-adapter), één per regel. Berichten van anderen worden genegeerd."
+}

package/admin/i18n/pl.json

@@ -0,0 +1,19 @@
+{
+	"Provider": "Dostawca",
+	"Model": "Model językowy",
+	"API key": "Klucz API",
+	"Dry run": "Próba bez zapisu",
+	"Log write requests instead of executing them": "Zapisuj żądania zapisu w dzienniku zamiast je wykonywać",
+	"Leave empty to use the default model of the provider": "Pozostaw puste, aby użyć domyślnego modelu dostawcy",
+	"Write allowlist": "Lista dozwolonych zapisów",
+	"One regular expression per line. The agent may only write states whose ID matches one of them.": "Jedno wyrażenie regularne w wierszu. Agent może zapisywać tylko stany, których ID pasuje do jednego z nich.",
+	"Instructions": "Instrukcje",
+	"Additional instructions for the agent": "Dodatkowe instrukcje dla agenta",
+	"Telegram": "Telegram",
+	"Telegram instance": "Instancja Telegram",
+	"Empty: the agent does not use Telegram": "Puste: agent nie korzysta z Telegrama",
+	"Prefix": "Prefiks",
+	"Only messages that start with this text go to the agent, e.g. /ai. Empty: every message of an allowed user": "Do agenta trafiają tylko wiadomości zaczynające się od tego tekstu, np. /ai. Puste: każda wiadomość dozwolonego użytkownika",
+	"Allowed Telegram users": "Dozwoleni użytkownicy Telegrama",
+	"Telegram user ID, user name or first name (as in the telegram adapter), one per line. Messages from anyone else are ignored.": "ID, nazwa użytkownika lub imię w Telegramie (jak w adapterze telegram), jedno w wierszu. Wiadomości od innych osób są ignorowane."
+}

package/admin/i18n/pt.json

@@ -0,0 +1,19 @@
+{
+	"Provider": "Provedor",
+	"Model": "Modelo",
+	"API key": "Chave de API",
+	"Dry run": "Execução de teste",
+	"Log write requests instead of executing them": "Registar os pedidos de escrita em vez de os executar",
+	"Leave empty to use the default model of the provider": "Deixe vazio para usar o modelo padrão do provedor",
+	"Write allowlist": "Lista de escrita permitida",
+	"One regular expression per line. The agent may only write states whose ID matches one of them.": "Uma expressão regular por linha. O agente só pode escrever estados cujo ID corresponda a uma delas.",
+	"Instructions": "Instruções",
+	"Additional instructions for the agent": "Instruções adicionais para o agente",
+	"Telegram": "Telegram",
+	"Telegram instance": "Instância do Telegram",
+	"Empty: the agent does not use Telegram": "Vazio: o agente não usa o Telegram",
+	"Prefix": "Prefixo",
+	"Only messages that start with this text go to the agent, e.g. /ai. Empty: every message of an allowed user": "Apenas mensagens que começam com este texto vão para o agente, p. ex. /ai. Vazio: todas as mensagens de um utilizador permitido",
+	"Allowed Telegram users": "Utilizadores do Telegram permitidos",
+	"Telegram user ID, user name or first name (as in the telegram adapter), one per line. Messages from anyone else are ignored.": "ID, nome de utilizador ou primeiro nome do Telegram (como no adaptador telegram), um por linha. As mensagens de qualquer outra pessoa são ignoradas."
+}

package/admin/i18n/ru.json

@@ -0,0 +1,19 @@
+{
+	"Provider": "Провайдер",
+	"Model": "Модель",
+	"API key": "API-ключ",
+	"Dry run": "Пробный запуск",
+	"Log write requests instead of executing them": "Только записывать запросы на запись в журнал, не выполняя их",
+	"Leave empty to use the default model of the provider": "Оставьте пустым, чтобы использовать модель провайдера по умолчанию",
+	"Write allowlist": "Список разрешённых записей",
+	"One regular expression per line. The agent may only write states whose ID matches one of them.": "Одно регулярное выражение в строке. Агент может записывать только состояния, ID которых соответствует одному из них.",
+	"Instructions": "Инструкции",
+	"Additional instructions for the agent": "Дополнительные инструкции для агента",
+	"Telegram": "Telegram",
+	"Telegram instance": "Инстанс Telegram",
+	"Empty: the agent does not use Telegram": "Пусто: агент не использует Telegram",
+	"Prefix": "Префикс",
+	"Only messages that start with this text go to the agent, e.g. /ai. Empty: every message of an allowed user": "Агенту передаются только сообщения, начинающиеся с этого текста, например /ai. Пусто: все сообщения разрешённого пользователя",
+	"Allowed Telegram users": "Разрешённые пользователи Telegram",
+	"Telegram user ID, user name or first name (as in the telegram adapter), one per line. Messages from anyone else are ignored.": "ID, имя пользователя или имя Telegram (как в адаптере telegram), по одному в строке. Сообщения всех остальных игнорируются."
+}

package/admin/i18n/uk.json

@@ -0,0 +1,19 @@
+{
+	"Provider": "Провайдер",
+	"Model": "Модель",
+	"API key": "API-ключ",
+	"Dry run": "Пробний запуск",
+	"Log write requests instead of executing them": "Лише записувати запити на запис у журнал, не виконуючи їх",
+	"Leave empty to use the default model of the provider": "Залиште порожнім, щоб використовувати модель провайдера за замовчуванням",
+	"Write allowlist": "Список дозволених записів",
+	"One regular expression per line. The agent may only write states whose ID matches one of them.": "Один регулярний вираз у рядку. Агент може записувати лише стани, ID яких відповідає одному з них.",
+	"Instructions": "Інструкції",
+	"Additional instructions for the agent": "Додаткові інструкції для агента",
+	"Telegram": "Telegram",
+	"Telegram instance": "Інстанс Telegram",
+	"Empty: the agent does not use Telegram": "Порожньо: агент не використовує Telegram",
+	"Prefix": "Префікс",
+	"Only messages that start with this text go to the agent, e.g. /ai. Empty: every message of an allowed user": "Агенту передаються лише повідомлення, що починаються з цього тексту, напр. /ai. Порожньо: усі повідомлення дозволеного користувача",
+	"Allowed Telegram users": "Дозволені користувачі Telegram",
+	"Telegram user ID, user name or first name (as in the telegram adapter), one per line. Messages from anyone else are ignored.": "ID, імʼя користувача або імʼя в Telegram (як в адаптері telegram), по одному в рядку. Повідомлення всіх інших ігноруються."
+}

package/admin/i18n/zh-cn.json

@@ -0,0 +1,19 @@
+{
+	"Provider": "提供商",
+	"Model": "模型",
+	"API key": "API 密钥",
+	"Dry run": "试运行",
+	"Log write requests instead of executing them": "仅记录写入请求而不执行",
+	"Leave empty to use the default model of the provider": "留空以使用提供商的默认模型",
+	"Write allowlist": "写入白名单",
+	"One regular expression per line. The agent may only write states whose ID matches one of them.": "每行一个正则表达式。代理只能写入 ID 与其中之一匹配的状态。",
+	"Instructions": "指令",
+	"Additional instructions for the agent": "给代理的附加指令",
+	"Telegram": "Telegram",
+	"Telegram instance": "Telegram 实例",
+	"Empty: the agent does not use Telegram": "留空：代理不使用 Telegram",
+	"Prefix": "前缀",
+	"Only messages that start with this text go to the agent, e.g. /ai. Empty: every message of an allowed user": "只有以此文本开头的消息才会发给代理，例如 /ai。留空：允许用户的所有消息",
+	"Allowed Telegram users": "允许的 Telegram 用户",
+	"Telegram user ID, user name or first name (as in the telegram adapter), one per line. Messages from anyone else are ignored.": "Telegram 用户 ID、用户名或名字（与 telegram 适配器中相同），每行一个。其他人的消息将被忽略。"
+}

package/admin/jsonConfig.json

@@ -0,0 +1,133 @@
+{
+	"i18n": true,
+	"type": "panel",
+	"items": {
+		"provider": {
+			"type": "select",
+			"label": "Provider",
+			"options": [
+				{
+					"label": "Anthropic",
+					"value": "anthropic"
+				},
+				{
+					"label": "OpenAI",
+					"value": "openai"
+				},
+				{
+					"label": "DeepSeek",
+					"value": "deepseek"
+				}
+			],
+			"xs": 12,
+			"sm": 6,
+			"md": 4,
+			"lg": 4,
+			"xl": 4
+		},
+		"modelId": {
+			"type": "text",
+			"label": "Model",
+			"xs": 12,
+			"sm": 6,
+			"md": 4,
+			"lg": 4,
+			"xl": 4,
+			"help": "Leave empty to use the default model of the provider"
+		},
+		"apiKey": {
+			"type": "password",
+			"label": "API key",
+			"repeat": false,
+			"xs": 12,
+			"sm": 12,
+			"md": 4,
+			"lg": 4,
+			"xl": 4
+		},
+		"dryRun": {
+			"type": "checkbox",
+			"label": "Dry run",
+			"help": "Log write requests instead of executing them",
+			"newLine": true,
+			"xs": 12,
+			"sm": 12,
+			"md": 12,
+			"lg": 12,
+			"xl": 12
+		},
+		"writeAllow": {
+			"type": "text",
+			"label": "Write allowlist",
+			"help": "One regular expression per line. The agent may only write states whose ID matches one of them.",
+			"minRows": 4,
+			"xs": 12,
+			"sm": 12,
+			"md": 12,
+			"lg": 12,
+			"xl": 12,
+			"maxRows": 12
+		},
+		"instructions": {
+			"type": "text",
+			"label": "Instructions",
+			"help": "Additional instructions for the agent",
+			"minRows": 3,
+			"xs": 12,
+			"sm": 12,
+			"md": 12,
+			"lg": 12,
+			"xl": 12,
+			"maxRows": 12
+		},
+		"telegramHeader": {
+			"type": "header",
+			"text": "Telegram",
+			"size": 3,
+			"newLine": true,
+			"xs": 12,
+			"sm": 12,
+			"md": 12,
+			"lg": 12,
+			"xl": 12
+		},
+		"telegramInstance": {
+			"type": "instance",
+			"adapter": "telegram",
+			"allowDeactivate": true,
+			"label": "Telegram instance",
+			"help": "Empty: the agent does not use Telegram",
+			"newLine": true,
+			"xs": 12,
+			"sm": 6,
+			"md": 4,
+			"lg": 4,
+			"xl": 4
+		},
+		"telegramPrefix": {
+			"type": "text",
+			"label": "Prefix",
+			"help": "Only messages that start with this text go to the agent, e.g. /ai. Empty: every message of an allowed user",
+			"hidden": "!data.telegramInstance",
+			"xs": 12,
+			"sm": 6,
+			"md": 4,
+			"lg": 4,
+			"xl": 4
+		},
+		"telegramUsers": {
+			"type": "text",
+			"label": "Allowed Telegram users",
+			"help": "Telegram user ID, user name or first name (as in the telegram adapter), one per line. Messages from anyone else are ignored.",
+			"minRows": 3,
+			"maxRows": 10,
+			"hidden": "!data.telegramInstance",
+			"newLine": true,
+			"xs": 12,
+			"sm": 12,
+			"md": 12,
+			"lg": 12,
+			"xl": 12
+		}
+	}
+}