teableio · HynLcc · Jul 17, 2026 · Jul 17, 2026
diff --git a/en/basic/admin-panel/ai-setting.mdx b/en/basic/admin-panel/ai-setting.mdx
@@ -8,11 +8,11 @@ description: "Configure AI Chat, AI fields, AI automation, and App Builder for a
 
 Path: Admin Panel → AI Settings
 
-The **AI Settings** page is used to configure **AI Chat**, **AI fields**, **AI automation**, and **App Builder**. The **Pending configuration** panel on the right shows which required items are still missing. Complete them in order.
+The **AI Settings** page is used to configure **AI Chat**, **AI fields**, **AI automation**, and **App Builder**. The **Pending configuration** panel on the right shows which required items are still missing.
 
 ## Before you start
 
-Before you open the AI Settings page, make sure the following items are ready. For self-hosted AI Chat and App Builder, first deploy the runtime plane — see [Architecture](/en/deploy/architecture).
+Before you open the AI Settings page, make sure the following items are ready.
 
 | Requirement | Required? | Notes |
 |------|------|------|
@@ -23,28 +23,30 @@ Before you open the AI Settings page, make sure the following items are ready. F
 
 ## Setup steps
 
-Follow the **Pending configuration** panel in order. Turn the required items green first, then test the features. If a button is disabled or a feature still does not work, it usually means an earlier requirement has not been completed yet.
-
-Follow these four steps:
+Start with the AI Chat runtime, app deployment, and model setup:
 
 <Steps>
+  <Step title="Configure the AI Chat runtime and app deployment">
+    Choose a [deployment option](/en/deploy/choose) and complete the deployment, then click **Run test** and **Test Connection**.
+  </Step>
   <Step title="Configure LLM API">
     Set up your model connection and complete the initial test.
   </Step>
   <Step title="Configure Recommended Models">
-    Add a clear set of primary models for users to choose from.
+    Choose the recommended models for AI fields and AI automation.
   </Step>
   <Step title="Set chat model">
     Choose the default model for AI Chat.
   </Step>
-  <Step title="Configure Agent Runtime and App Builder">
-    Confirm the Agent Runtime configuration status, then complete the App Builder access checks.
-  </Step>
 </Steps>
 
+### Configure the runtime
+
+Click **Run test** to verify the AI Chat runtime.
+
 ### Configure LLM API
 
-In **LLM API**, add or verify the model provider used by supported model calls. If **AI Chat** or **App Builder** uses your own key (BYOK), choose **OpenAI Compatible** or **Anthropic** as the provider type, add the models it supports, and complete the model test.
+In **LLM API**, add a model provider and test its models. If you use your own key (BYOK), choose **OpenAI Compatible** or **Anthropic** as the provider type.
 
 Open **Model Settings** for a model when you need to review or override these values:
 
@@ -55,38 +57,19 @@ Open **Model Settings** for a model when you need to review or override these va
 
 ### Configure Recommended Models
 
-This step decides which models users can choose from. Start with a clear primary set so users can choose by task complexity. A common setup is:
-
-| Model | Best for |
-|------|------|
-| **Opus** | Complex tasks and high-quality output |
-| **Sonnet** | Everyday tasks with a good balance of quality and cost |
-| **Haiku** | Lightweight tasks and fast responses |
+Choose the recommended models users can select in **AI fields** and **AI automation**.
 
 ### Set chat model
 
-This step sets the default model for sidebar **AI Chat**. The model must come from the recommended model list, so if this step is disabled, the previous step is usually incomplete. If **AI Chat** or **App Builder** uses your own key (BYOK), choose a tested **OpenAI Compatible** or **Anthropic** model.
+Choose the default model for sidebar **AI Chat**. The chat model must come from the recommended model list.
 
-Choose a tested, stable model here.
+### Enable AI features
 
-### Optional. Configure Voice input
-
-**Voice input** lets users enter messages in AI Chat with a microphone.
+In **AI Capabilities**, enable **AI Field** and **AI Chat** as needed. Configure the runtime before using **AI Chat**.
 
-To let users use the microphone in AI Chat, enable voice input and enter the **OpenAI API Key**. Teable uses the default OpenAI transcription endpoint automatically; only change **Transcription Endpoint** when your deployment uses a custom transcription service. You can also adjust **Max recording duration (seconds)** and **Transcription requests per minute per user**. After the setting is saved, users will see the microphone button in the AI Chat input box.
+### Configure App Builder
 
-### Configure Agent Runtime and App Builder
-
-Self-hosted **AI Chat** and **App Builder** require Agent Runtime setup and a compatible chat model. When you use your own key (BYOK), the chat model must come from an **OpenAI Compatible** or **Anthropic** provider. Deploy the runtime plane first — see [Architecture](/en/deploy/architecture) and the [teable-deployment](https://github.com/teableio/teable-deployment) repository. For assistance, contact **support@teable.ai**.
-
-<Steps>
-  <Step title="Confirm Agent Runtime status">
-    Confirm that the current deployment includes the required Agent Runtime configuration.
-  </Step>
-  <Step title="Finish the App Builder setup">
-    Confirm **Test Public Access**, then configure optional App Builder settings such as custom domain and app sign-in if you need them.
-  </Step>
-</Steps>
+Configure the **Deployment Engine**, then click **Test Connection**.
 
 You can add these optional App Builder settings when needed:
 
@@ -106,28 +89,17 @@ In **App Auth**, you can configure sign-in providers for generated apps:
 Click **Test Public Access** in the page and confirm that both your **Teable instance** and **object storage (MinIO / S3)** are publicly accessible.
 </Tip>
 
-**How to verify the setup is complete**
+### Optional. Configure Voice input
+
+To use voice input in AI Chat, enable the feature and enter an **OpenAI API Key**. Configure the **Transcription Endpoint** only when using a custom transcription service. You can also limit the recording length and requests per minute.
+
+### Verify the setup
 
 After the setup is done, first confirm that all required items in the **Pending configuration** panel are green. Then test **AI Chat**, **AI fields**, **AI automation**, and **App Builder** one by one.
 
 ## FAQ
 
 <AccordionGroup>
-<Accordion title="The chat model option is disabled">
-
-Go back and finish **Configure recommended models** first. The chat model can only be selected from the recommended model list.
-</Accordion>
-
-<Accordion title="I entered the API key, but the right-side status is still not green">
-
-Check the **Pending configuration** panel first to see which item is still missing. In many cases, the current step is saved correctly, but an earlier requirement is still incomplete.
-</Accordion>
-
-<Accordion title="The page says Agent Runtime configuration is still missing">
-
-Check whether the runtime plane is deployed and connected (see [Architecture](/en/deploy/architecture)), including the required Teable environment variables — the deployment's doctor script verifies the whole chain. For assistance, contact **support@teable.ai**.
-</Accordion>
-
 <Accordion title="I get a credit exceeded error">
 
 Check the account or provider used by your configured model service. Top up or change the model configuration if needed.

diff --git a/en/basic/admin-panel/computed-outbox.mdx b/en/basic/admin-panel/computed-outbox.mdx
@@ -7,7 +7,7 @@ Path: Admin Panel → Computed Outbox
 
 Use **Computed Outbox** when formula, lookup, or other computed field values stop updating or take much longer than expected. It helps instance administrators determine whether the delay comes from Redis and BullMQ, missing workers, queue failures, or database backlog before checking service logs.
 
-The page samples the BullMQ queue and database task ledger. It is read-only and does not execute or retry tasks.
+The page samples the BullMQ queue and database task ledger. It also lets you recover individual dead-letter or timed-out tasks.
 
 ## Read the Health Status
 
@@ -33,6 +33,10 @@ When the status needs attention, the page lists the specific reasons below the h
 
 Click **Refresh** to request a new sample. The page also refreshes using the configured monitoring interval, which is 30 seconds by default.
 
+## Recover an Anomalous Task
+
+The **Anomaly maintenance** section lists dead-letter and timed-out tasks. Fix the cause, click **Recover**, then click **Recover task**. Teable delivers the task when BullMQ is available or keeps it in the database ledger until BullMQ recovers.
+
 ## Use the Result
 
 | What you see | What to check next |

diff --git a/zh/basic/admin-panel/ai-setting.mdx b/zh/basic/admin-panel/ai-setting.mdx
@@ -8,11 +8,11 @@ description: "为私有化部署实例配置 AI 对话、AI 字段、AI 自动
 
 路径：管理面板 → AI 配置
 
-**AI 配置** 页用于配置 **AI 对话**、**AI 字段**、**AI 自动化** 和 **应用构建器**。右侧的 **待配置项** 会提示当前还缺少哪些关键配置，建议按顺序逐项完成。
+**AI 配置** 页用于配置 **AI 对话**、**AI 字段**、**AI 自动化** 和 **应用构建器**。右侧的 **待配置项** 会提示当前还缺少哪些关键配置。
 
 ## 开始前的准备
 
-在进入 AI 配置页之前，先确认下面这些准备项已经就绪。私有化部署如需使用 AI 对话和应用构建器，请先部署运行平面，见 [架构](/zh/deploy/architecture)。
+在进入 AI 配置页之前，先确认下面这些准备项已经就绪。
 
 | 准备项 | 是否必需 | 说明 |
 |------|------|------|
@@ -23,28 +23,30 @@ description: "为私有化部署实例配置 AI 对话、AI 字段、AI 自动
 
 ## 配置步骤
 
-建议按 **待配置项** 的顺序完成配置，先让必填项变绿，再去测试功能。很多“按钮是灰的”或“功能仍不可用”的情况，通常只是前置项还没完成。
-
-建议按下面四步完成：
+先完成 AI 对话运行环境、应用部署和模型配置：
 
 <Steps>
+  <Step title="配置 AI 对话运行环境和应用部署">
+    先查看[部署方式](/zh/deploy/choose)并完成部署，再点击 **运行测试** 和 **测试连接**。
+  </Step>
   <Step title="配置 LLM API">
     先完成模型接入和基础验证。
   </Step>
   <Step title="配置推荐模型">
-    先给用户一组清晰的主力模型。
+    选择供 AI 字段和 AI 自动化使用的推荐模型。
   </Step>
   <Step title="设置聊天模型">
     为 AI 对话指定默认模型。
   </Step>
-  <Step title="配置 Agent Runtime 与应用构建器">
-    确认 Agent Runtime 配置状态，再完成应用构建器的访问检查。
-  </Step>
 </Steps>
 
+### 配置运行环境
+
+点击 **运行测试** 验证 AI 对话运行环境。
+
 ### 配置 LLM API
 
-在 **LLM API** 中添加或确认已支持的模型调用所使用的模型服务商。如果 **AI 对话** 或 **应用构建器** 使用自带 Key，请将提供商类型选为 **OpenAI Compatible**（OpenAI 兼容）或 **Anthropic**，填入它支持的模型，并完成模型测试。
+在 **LLM API** 中添加模型服务商，并完成模型测试。使用自带 Key 时，请将提供商类型选为 **OpenAI Compatible**（OpenAI 兼容）或 **Anthropic**。
 
 如需检查或调整模型参数，请打开该模型的 **模型配置**：
 
@@ -55,38 +57,19 @@ description: "为私有化部署实例配置 AI 对话、AI 字段、AI 自动
 
 ### 配置推荐模型
 
-这一步决定了用户可以选择哪些模型。建议先提供一组清晰的主力模型，方便用户按任务复杂度选择。一个常见的做法是：
-
-| 模型 | 适合场景 |
-|------|------|
-| **Opus** | 复杂任务和高质量输出 |
-| **Sonnet** | 日常任务，兼顾效果和成本 |
-| **Haiku** | 轻量任务和快速响应 |
+选择用户可以在 **AI 字段** 和 **AI 自动化** 中使用的推荐模型。
 
 ### 设置聊天模型
 
-这里要为侧边栏 **AI 对话** 指定默认模型。这个模型需要从推荐模型列表中选择，所以如果这一步是灰的，通常说明前一步还没完成。**AI 对话** 或 **应用构建器** 使用自带 Key 时，请选择已测试通过的 **OpenAI Compatible**（OpenAI 兼容）或 **Anthropic** 模型。
+为侧边栏 **AI 对话** 选择默认模型。聊天模型只能从推荐模型列表中选择。
 
-这里建议选择已经完成测试、表现稳定的模型。
+### 启用 AI 功能
 
-### 可选：配置语音输入
-
-**语音输入** 可以让用户在 AI 对话中用麦克风输入消息。
+在 **AI 能力** 中按需启用 **AI 字段** 和 **AI 聊天**。使用 **AI 聊天** 前，需要先完成运行环境配置。
 
-如果希望用户在 AI 对话中使用麦克风输入，请启用语音输入并填写 **OpenAI API Key**。Teable 默认使用 OpenAI 官方转写接口，不需要修改 **转写 Endpoint**；只有接入自定义转写服务时，才需要填写对应的接口地址。你也可以按需设置 **单次最长录音时长（秒）** 和 **每用户每分钟转写次数**。保存后，用户会在 AI 对话输入框中看到麦克风按钮。
+### 配置应用构建器
 
-### 配置 Agent Runtime 与应用构建器
-
-私有化部署如需使用 **AI 对话** 和 **应用构建器**，需要先完成 Agent Runtime 配置，并配置兼容的聊天模型。使用自带 Key 时，聊天模型必须来自 **OpenAI Compatible**（OpenAI 兼容）或 **Anthropic** 类型的提供商。请先部署运行平面 —— 见 [架构](/zh/deploy/architecture) 与 [teable-deployment](https://github.com/teableio/teable-deployment) 仓库。如需协助，请联系 **support@teable.ai**。
-
-<Steps>
-  <Step title="确认 Agent Runtime 状态">
-    确认当前部署已包含所需的 Agent Runtime 配置。
-  </Step>
-  <Step title="完成应用构建器配置">
-    确认 **测试公网访问** 通过，再按需配置自定义域名、应用登录等可选项。
-  </Step>
-</Steps>
+在 **部署引擎** 中完成配置，然后点击 **测试连接**。
 
 如有需要，可以继续补充下面这些应用构建器可选配置：
 
@@ -106,28 +89,17 @@ description: "为私有化部署实例配置 AI 对话、AI 字段、AI 自动
 点击页面中的 **测试公网访问**，确认 **Teable 实例** 和 **对象存储（MinIO / S3）** 都已经可以从公网访问。
 </Tip>
 
-**如何验证配置是否完成**
+### 可选：配置语音输入
+
+如需在 AI 对话中使用语音输入，请启用该功能并填写 **OpenAI API Key**。只有使用自定义转写服务时，才需要配置 **转写 Endpoint**。你还可以按需限制录音时长和每分钟请求次数。
+
+### 验证配置
 
 完成配置后，先确认右侧 **待配置项** 的必填项都已变绿。然后依次测试 **AI 对话**、**AI 字段**、**AI 自动化** 和 **应用构建器**。
 
 ## 常见问题
 
 <AccordionGroup>
-<Accordion title="设置聊天模型是灰的">
-
-先回到上一步完成 **配置推荐模型**。聊天模型只能从推荐模型列表中选择。
-</Accordion>
-
-<Accordion title="明明填了 API Key，右侧还是没变绿">
-
-先回看右侧 **待配置项**，确认缺的是哪一项。很多时候不是当前这一步没保存，而是前面的配置还没有补齐。
-</Accordion>
-
-<Accordion title="提示还缺少 Agent Runtime 配置">
-
-请检查运行平面是否已部署并接通（见 [架构](/zh/deploy/architecture)），以及所需的 Teable 环境变量是否配置 —— 部署仓库自带的 doctor 脚本可以整链验证。如需协助，请联系 **support@teable.ai**。
-</Accordion>
-
 <Accordion title="出现 credit exceeded 报错">
 
 请检查当前模型服务对应的账户或服务商额度，必要时充值或调整模型配置。

diff --git a/zh/basic/admin-panel/computed-outbox.mdx b/zh/basic/admin-panel/computed-outbox.mdx
@@ -7,7 +7,7 @@ description: "在管理面板中监控 BullMQ 投递和计算任务积压。"
 
 当公式、Lookup 等计算字段没有更新，或更新时间明显变长时，可以打开 **计算任务 Outbox**。它帮助实例管理员先判断问题来自 Redis 和 BullMQ、Worker 缺失、队列失败还是数据库积压，再查看对应的服务日志。
 
-页面会采样 BullMQ 队列和数据库任务账本。它是只读监控，不会执行或重试任务。
+页面会采样 BullMQ 队列和数据库任务账本，也可以逐条恢复死信或执行超时的任务。
 
 ## 查看健康状态
 
@@ -33,6 +33,10 @@ description: "在管理面板中监控 BullMQ 投递和计算任务积压。"
 
 点击 **刷新** 可以立即请求一次新采样。页面也会按配置的监控间隔自动刷新，默认间隔为 30 秒。
 
+## 恢复异常任务
+
+**异常任务维护** 区域会显示死信和执行超时任务。修复异常原因后，点击 **恢复**，再点击 **确认恢复**。BullMQ 可用时，Teable 会重新投递任务；暂不可用时，任务会保留在数据库账本中，待 BullMQ 恢复后再投递。
+
 ## 根据结果继续排查
 
 | 页面结果 | 后续检查 |