From 269fa7d2d50e64cd1057501c6be8588a83ba0b94 Mon Sep 17 00:00:00 2001 From: zhayujie Date: Fri, 27 Feb 2026 18:37:22 +0800 Subject: [PATCH 01/10] feat: 2.0.2 en docs --- docs/en/README.md | 4 ++++ docs/releases/v2.0.2.mdx | 14 +++++++------- 2 files changed, 11 insertions(+), 7 deletions(-) diff --git a/docs/en/README.md b/docs/en/README.md index a17e6c36..a79219c5 100644 --- a/docs/en/README.md +++ b/docs/en/README.md @@ -35,6 +35,10 @@ ## Changelog +> **2026.02.27:** [v2.0.2](https://github.com/zhayujie/chatgpt-on-wechat/releases/tag/2.0.2) — Web console overhaul (streaming chat, model/skill/memory/channel/scheduler/log management), multi-channel concurrent running, session persistence, new models including Gemini 3.1 Pro / Claude 4.6 Sonnet / Qwen3.5 Plus. + +> **2026.02.13:** [v2.0.1](https://github.com/zhayujie/chatgpt-on-wechat/releases/tag/2.0.1) — Built-in Web Search tool, smart context trimming, runtime info dynamic update, Windows compatibility, fixes for scheduler memory loss, Feishu connection issues, and more. + > **2026.02.03:** [v2.0.0](https://github.com/zhayujie/chatgpt-on-wechat/releases/tag/2.0.0) — Full upgrade to AI super assistant with multi-step task planning, long-term memory, built-in tools, Skills framework, new models, and optimized channels. > **2025.05.23:** [v1.7.6](https://github.com/zhayujie/chatgpt-on-wechat/releases/tag/1.7.6) — Web channel optimization, AgentMesh multi-agent plugin, Baidu TTS, claude-4-sonnet/opus support. diff --git a/docs/releases/v2.0.2.mdx b/docs/releases/v2.0.2.mdx index aac2ad3d..cc3626f0 100644 --- a/docs/releases/v2.0.2.mdx +++ b/docs/releases/v2.0.2.mdx @@ -13,43 +13,43 @@ description: CowAgent 2.0.2 - Web 控制台升级、多通道同时运行、会 支持流式输出，可实时展示 Agent 的思考过程（Reasoning）和工具调用过程（Tool Calls），更直观地观察 Agent 的决策过程： -![对话界面](https://cdn.link-ai.tech/doc/20260227180120.png) + #### 模型管理 支持在线管理模型配置，无需手动编辑配置文件： -![模型管理](https://cdn.link-ai.tech/doc/20260227173811.png) + #### 技能管理 支持在线查看和管理 Agent 技能（Skills）： -![技能管理](https://cdn.link-ai.tech/doc/20260227173403.png) + #### 记忆管理 支持在线查看和管理 Agent 记忆： -![记忆管理](https://cdn.link-ai.tech/doc/20260227173349.png) + #### 通道管理 支持在线管理接入通道，支持实时连接/断开操作： -![通道管理](https://cdn.link-ai.tech/doc/20260227173331.png) + #### 定时任务 支持在线查看和管理定时任务，包括一次性任务、固定间隔、Cron 表达式等多种调度方式的可视化管理： -![定时任务](https://cdn.link-ai.tech/doc/20260227173704.png) + #### 日志 支持在线实时查看 Agent 运行日志，便于监控运行状态和排查问题： -![日志](https://cdn.link-ai.tech/doc/20260227173514.png) + 相关提交：[f1a1413](https://github.com/zhayujie/chatgpt-on-wechat/commit/f1a1413), [c0702c8](https://github.com/zhayujie/chatgpt-on-wechat/commit/c0702c8), [394853c](https://github.com/zhayujie/chatgpt-on-wechat/commit/394853c), [1c71c4e](https://github.com/zhayujie/chatgpt-on-wechat/commit/1c71c4e), [5e3eccb](https://github.com/zhayujie/chatgpt-on-wechat/commit/5e3eccb), [e1dc037](https://github.com/zhayujie/chatgpt-on-wechat/commit/e1dc037), [5edbf4c](https://github.com/zhayujie/chatgpt-on-wechat/commit/5edbf4c), [7d258b5](https://github.com/zhayujie/chatgpt-on-wechat/commit/7d258b5) From 8705bf0a70423833d3787736e8917d11b55d851b Mon Sep 17 00:00:00 2001 From: zhayujie Date: Sat, 28 Feb 2026 10:53:16 +0800 Subject: [PATCH 02/10] feat: update docs --- channel/web/chat.html | 16 +++++- docs/channels/dingtalk.mdx | 32 +++++++++--- docs/channels/feishu.mdx | 72 ++++++++++++------------- docs/channels/web.mdx | 56 +++++++++++++++++--- docs/channels/wechatmp.mdx | 40 +++++++++----- docs/channels/wecom.mdx | 47 ++++++++++++++--- docs/docs.json | 1 + docs/en/channels/dingtalk.mdx | 36 ++++++++++--- docs/en/channels/feishu.mdx | 74 +++++++++++++------------- docs/en/channels/web.mdx | 56 +++++++++++++++++--- docs/en/channels/wechatmp.mdx | 44 +++++++++++----- docs/en/channels/wecom.mdx | 53 +++++++++++++++---- docs/en/releases/overview.mdx | 1 + docs/en/releases/v2.0.2.mdx | 98 +++++++++++++++++++++++++++++++++++ docs/guide/manual-install.mdx | 2 +- docs/guide/quick-start.mdx | 2 +- 16 files changed, 485 insertions(+), 145 deletions(-) create mode 100644 docs/en/releases/v2.0.2.mdx diff --git a/channel/web/chat.html b/channel/web/chat.html index 886a8571..37f28ae6 100644 --- a/channel/web/chat.html +++ b/channel/web/chat.html @@ -192,10 +192,24 @@ + + + + + + + + + + + cursor-pointer transition-colors duration-150" title="GitHub"> diff --git a/docs/channels/dingtalk.mdx b/docs/channels/dingtalk.mdx index c01bb367..df11019e 100644 --- a/docs/channels/dingtalk.mdx +++ b/docs/channels/dingtalk.mdx @@ -7,15 +7,29 @@ description: 将 CowAgent 接入钉钉应用 ## 一、创建应用 -1. 进入 [钉钉开发者后台](https://open-dev.dingtalk.com/fe/app#/corp/app)，点击 **创建应用**，填写应用信息 -2. 点击添加应用能力，选择 **机器人** 能力并添加 -3. 配置机器人信息后点击 **发布** +1. 进入 [钉钉开发者后台](https://open-dev.dingtalk.com/fe/app#/corp/app)，登录后点击 **创建应用**，填写应用相关信息： + + + +2. 点击添加应用能力，选择 **机器人** 能力，点击 **添加**： + + + +3. 配置机器人信息后点击 **发布**。发布后，点击 "**点击调试**"，会自动创建测试群聊，可在客户端查看： + + + +4. 点击 **版本管理与发布**，创建新版本发布： + + ## 二、项目配置 -1. 在 **凭证与基础信息** 中获取 `Client ID` 和 `Client Secret` +1. 点击 **凭证与基础信息**，获取 `Client ID` 和 `Client Secret`： -2. 填入 `config.json`： + + +2. 将以下配置加入项目根目录的 `config.json` 文件： ```json { @@ -31,8 +45,12 @@ description: 将 CowAgent 接入钉钉应用 pip3 install dingtalk_stream ``` -4. 启动项目后，在钉钉开发者后台点击 **事件订阅**，点击 **已完成接入，验证连接通道**，显示"连接接入成功"即表示配置完成 +4. 启动项目后，在钉钉开发者后台点击 **事件订阅**，点击 **已完成接入，验证连接通道**，显示 **连接接入成功** 即表示配置完成： + + ## 三、使用 -与机器人私聊或将机器人拉入企业群中均可开启对话。 +与机器人私聊或将机器人拉入企业群中均可开启对话： + + diff --git a/docs/channels/feishu.mdx b/docs/channels/feishu.mdx index f69e3af7..575aa483 100644 --- a/docs/channels/feishu.mdx +++ b/docs/channels/feishu.mdx @@ -3,65 +3,67 @@ title: 飞书 description: 将 CowAgent 接入飞书应用 --- -通过自建应用将 CowAgent 接入飞书，支持 WebSocket 长连接（推荐）和 Webhook 两种事件接收模式。 +通过自建应用将 CowAgent 接入飞书，需要是飞书企业用户且具有企业管理权限。 ## 一、创建企业自建应用 ### 1. 创建应用 -进入 [飞书开发平台](https://open.feishu.cn/app/)，点击 **创建企业自建应用**，填写必要信息后创建。 +进入 [飞书开发平台](https://open.feishu.cn/app/)，点击 **创建企业自建应用**，填写必要信息后点击 **创建**： + + ### 2. 添加机器人能力 -在 **添加应用能力** 菜单中，为应用添加 **机器人** 能力。 +在 **添加应用能力** 菜单中，为应用添加 **机器人** 能力： + + ### 3. 配置应用权限 -点击 **权限管理**，粘贴以下权限配置，全选并批量开通： +点击 **权限管理**，复制以下权限配置，粘贴到 **权限配置** 下方的输入框内，全选筛选出来的权限，点击 **批量开通** 并确认： ``` im:message,im:message.group_at_msg,im:message.group_at_msg:readonly,im:message.p2p_msg,im:message.p2p_msg:readonly,im:message:send_as_bot,im:resource ``` + + ## 二、项目配置 -在 **凭证与基础信息** 中获取 `App ID` 和 `App Secret`，填入 `config.json`： +1. 在 **凭证与基础信息** 中获取 `App ID` 和 `App Secret`： - - - 无需公网 IP，配置如下： + - ```json - { - "channel_type": "feishu", - "feishu_app_id": "YOUR_APP_ID", - "feishu_app_secret": "YOUR_APP_SECRET", - "feishu_event_mode": "websocket" - } - ``` +2. 将以下配置加入项目根目录的 `config.json` 文件： - 需安装依赖：`pip3 install lark-oapi` - - - 需要公网 IP，配置如下： +```json +{ + "channel_type": "feishu", + "feishu_app_id": "YOUR_APP_ID", + "feishu_app_secret": "YOUR_APP_SECRET", + "feishu_bot_name": "YOUR_BOT_NAME" +} +``` - ```json - { - "channel_type": "feishu", - "feishu_app_id": "YOUR_APP_ID", - "feishu_app_secret": "YOUR_APP_SECRET", - "feishu_token": "VERIFICATION_TOKEN", - "feishu_event_mode": "webhook", - "feishu_port": 9891 - } - ``` - - +| 参数 | 说明 | +| --- | --- | +| `feishu_app_id` | 飞书机器人应用 App ID | +| `feishu_app_secret` | 飞书机器人 App Secret | +| `feishu_bot_name` | 飞书机器人名称（创建应用时设置），群聊中使用依赖此配置 | + +配置完成后启动项目。 ## 三、配置事件订阅 -1. 启动项目后，在飞书开放平台点击 **事件与回调**，选择 **长连接** 方式并保存 -2. 点击 **添加事件**，搜索 "接收消息"，选择 "接收消息v2.0"，确认添加 -3. 点击 **版本管理与发布**，创建版本并申请线上发布，审核通过后即可使用 +1. 成功运行项目后，在飞书开放平台点击 **事件与回调**，选择 **长连接** 方式，点击保存： + + + +2. 点击下方的 **添加事件**，搜索 "接收消息"，选择 "**接收消息v2.0**"，确认添加。 + +3. 点击 **版本管理与发布**，创建版本并申请 **线上发布**，在飞书客户端查看审批消息并审核通过： + + 完成后在飞书中搜索机器人名称，即可开始对话。 diff --git a/docs/channels/web.mdx b/docs/channels/web.mdx index b2237aa1..5d1528a2 100644 --- a/docs/channels/web.mdx +++ b/docs/channels/web.mdx @@ -1,9 +1,9 @@ --- -title: Web 网页 -description: 通过 Web 网页端使用 CowAgent +title: Web 控制台 +description: 通过 Web 控制台使用 CowAgent --- -Web 是 CowAgent 的默认通道，启动后会自动运行 Web 控制台，通过浏览器即可与 Agent 对话。 +Web 控制台是 CowAgent 的默认通道，启动后会自动运行，通过浏览器即可与 Agent 对话，并支持在线管理模型、技能、记忆、通道等配置。 ## 配置 @@ -19,13 +19,57 @@ Web 是 CowAgent 的默认通道，启动后会自动运行 Web 控制台，通 | `channel_type` | 设为 `web` | `web` | | `web_port` | Web 服务监听端口 | `9899` | -## 使用 +## 访问地址 启动项目后访问： -- 本地运行：`http://localhost:9899/chat` -- 服务器运行：`http://:9899/chat` +- 本地运行：`http://localhost:9899` +- 服务器运行：`http://:9899` 请确保服务器防火墙和安全组已放行对应端口。 + +## 功能介绍 + +### 对话界面 + +支持流式输出，可实时展示 Agent 的思考过程（Reasoning）和工具调用过程（Tool Calls），更直观地观察 Agent 的决策过程： + + + +### 模型管理 + +支持在线管理模型配置，无需手动编辑配置文件： + + + +### 技能管理 + +支持在线查看和管理 Agent 技能（Skills）： + + + +### 记忆管理 + +支持在线查看和管理 Agent 记忆： + + + +### 通道管理 + +支持在线管理接入通道，支持实时连接/断开操作： + + + +### 定时任务 + +支持在线查看和管理定时任务，包括一次性任务、固定间隔、Cron 表达式等多种调度方式的可视化管理： + + + +### 日志 + +支持在线实时查看 Agent 运行日志，便于监控运行状态和排查问题： + + diff --git a/docs/channels/wechatmp.mdx b/docs/channels/wechatmp.mdx index 69f69579..917df41d 100644 --- a/docs/channels/wechatmp.mdx +++ b/docs/channels/wechatmp.mdx @@ -7,21 +7,22 @@ CowAgent 支持接入个人订阅号和企业服务号两种公众号类型。 | 类型 | 要求 | 特点 | | --- | --- | --- | -| **个人订阅号** | 个人可申请 | 回复生成后需用户主动发消息获取 | +| **个人订阅号** | 个人可申请 | 收到消息时会回复一条提示，回复生成后需用户主动发消息获取 | | **企业服务号** | 企业申请，需通过微信认证开通客服接口 | 回复生成后可主动推送给用户 | - 公众号仅支持服务器和 Docker 部署，需额外安装扩展依赖：`pip3 install -r requirements-optional.txt` + 公众号仅支持服务器和 Docker 部署，不支持本地运行。需额外安装扩展依赖：`pip3 install -r requirements-optional.txt` ## 一、个人订阅号 -在 `config.json` 中配置： +在 `config.json` 中添加以下配置： ```json { "channel_type": "wechatmp", - "wechatmp_app_id": "YOUR_APP_ID", + "single_chat_prefix": [""], + "wechatmp_app_id": "wx73f9******d1e48", "wechatmp_app_secret": "YOUR_APP_SECRET", "wechatmp_aes_key": "", "wechatmp_token": "YOUR_TOKEN", @@ -31,22 +32,37 @@ CowAgent 支持接入个人订阅号和企业服务号两种公众号类型。 ### 配置步骤 -1. 在 [微信公众平台](https://mp.weixin.qq.com/) 的 **设置与开发 → 基本配置 → 服务器配置** 中获取参数 -2. 启用开发者密码，将服务器 IP 加入白名单 -3. 启动程序（监听 80 端口） -4. 在公众号后台 **启用服务器配置**，URL 格式为 `http://{HOST}/wx` +这些配置需要和 [微信公众号后台](https://mp.weixin.qq.com/advanced/advanced?action=dev&t=advanced/dev) 中的保持一致，进入页面后，在左侧菜单选择 **设置与开发 → 基本配置 → 服务器配置**，按下图进行配置： + + + +1. 在公众平台启用开发者密码（对应配置 `wechatmp_app_secret`），并将服务器 IP 填入白名单 +2. 按上图填写 `config.json` 中与公众号相关的配置，要与公众号后台的配置一致 +3. 启动程序，启动后会监听 80 端口（若无权限监听，则在启动命令前加上 `sudo`；若 80 端口已被占用，则关闭该占用进程） +4. 在公众号后台 **启用服务器配置** 并提交，保存成功则表示已成功配置。注意 **"服务器地址(URL)"** 需要配置为 `http://{HOST}/wx` 的格式，其中 `{HOST}` 可以是服务器的 IP 或域名 + +随后关注公众号并发送消息即可看到以下效果： + + + +由于受订阅号限制，回复内容较短的情况下（15s 内），可以立即完成回复，但耗时较长的回复则会先回复一句 "正在思考中"，后续需要用户输入任意文字主动获取答案，而服务号则可以通过客服接口解决这一问题。 + + + **语音识别**：可利用微信自带的语音识别功能，需要在公众号管理页面的 "设置与开发 → 接口权限" 页面开启 "接收语音识别结果"。 + ## 二、企业服务号 -与个人订阅号流程基本相同，差异如下： +企业服务号与上述个人订阅号的接入过程基本相同，差异如下： -1. 在公众平台申请企业服务号并完成微信认证，确认已获得 **客服接口** 权限 -2. 在 `config.json` 中设置 `"channel_type": "wechatmp_service"` -3. 即使是较长耗时的回复，也可以主动推送给用户 +1. 在公众平台申请企业服务号并完成微信认证，在接口权限中确认已获得 **客服接口** 的权限 +2. 在 `config.json` 中设置 `"channel_type": "wechatmp_service"`，其他配置与上述订阅号相同 +3. 交互效果上，即使是较长耗时的回复，也可以主动推送给用户，无需用户手动获取 ```json { "channel_type": "wechatmp_service", + "single_chat_prefix": [""], "wechatmp_app_id": "YOUR_APP_ID", "wechatmp_app_secret": "YOUR_APP_SECRET", "wechatmp_aes_key": "", diff --git a/docs/channels/wecom.mdx b/docs/channels/wecom.mdx index efce1199..d58afd76 100644 --- a/docs/channels/wecom.mdx +++ b/docs/channels/wecom.mdx @@ -19,17 +19,35 @@ description: 将 CowAgent 接入企业微信自建应用 ## 二、创建企业微信应用 -1. 在 [企业微信管理后台](https://work.weixin.qq.com/wework_admin/frame#profile) **我的企业** 中获取 **企业ID** -2. 切换到 **应用管理**，点击创建应用，记录 `AgentId` 和 `Secret` -3. 点击 **设置API接收**，配置应用接口： - - URL 格式为 `http://ip:port/wxcomapp`（认证企业需使用备案域名） - - 随机获取 `Token` 和 `EncodingAESKey` 并保存 +1. 在 [企业微信管理后台](https://work.weixin.qq.com/wework_admin/frame#profile) 点击 **我的企业**，在最下方获取 **企业ID**（后续填写到 `wechatcom_corp_id` 字段中）。 + +2. 切换到 **应用管理**，点击创建应用： + + + +3. 进入应用创建页面，记录 `AgentId` 和 `Secret`： + + + +4. 点击 **设置API接收**，配置应用接口： + + + +- URL 格式为 `http://ip:port/wxcomapp`（认证企业需使用备案域名） +- 随机获取 `Token` 和 `EncodingAESKey` 并保存 + + + 此时保存 API 接收配置会失败，因为程序还未启动，等项目运行后再回来保存。 + ## 三、配置和运行 +在 `config.json` 中添加以下配置（各参数与企业微信后台的对应关系见上方截图）： + ```json { "channel_type": "wechatcom_app", + "single_chat_prefix": [""], "wechatcom_corp_id": "YOUR_CORP_ID", "wechatcomapp_token": "YOUR_TOKEN", "wechatcomapp_secret": "YOUR_SECRET", @@ -48,12 +66,25 @@ description: 将 CowAgent 接入企业微信自建应用 | `wechatcomapp_aes_key` | API 接收配置中的 EncodingAESKey | | `wechatcomapp_port` | 监听端口，默认 9898 | -启动程序后，回到企业微信后台保存 **消息服务器配置**，并将服务器 IP 添加到 **企业可信IP** 中。 +配置完成后启动程序。当后台日志显示 `http://0.0.0.0:9898/` 时说明程序运行成功，需要将该端口对外开放（如在云服务器安全组中放行）。 + +程序启动后，回到企业微信后台保存 **消息服务器配置**，保存成功后还需将服务器 IP 添加到 **企业可信IP** 中，否则无法收发消息： + + - 如遇到配置失败：1. 确保防火墙和安全组已放行端口；2. 检查各参数配置是否一致；3. 认证企业需配置备案域名。 + 如遇到 URL 配置回调不通过或配置失败： + 1. 确保服务器防火墙关闭且安全组放行监听端口 + 2. 仔细检查 Token、Secret Key 等参数配置是否一致，URL 格式是否正确 + 3. 认证企业微信需要配置与主体一致的备案域名 ## 四、使用 -在企业微信中搜索应用名称即可直接对话。如需让外部微信用户使用，可在 **我的企业 → 微信插件** 中分享邀请关注二维码。 +在企业微信中搜索刚创建的应用名称，即可直接对话： + + + +如需让外部个人微信用户使用，可在 **我的企业 → 微信插件** 中分享邀请关注二维码，个人微信扫码关注后即可与应用对话： + + diff --git a/docs/docs.json b/docs/docs.json index 47844b5b..86a42e9c 100644 --- a/docs/docs.json +++ b/docs/docs.json @@ -311,6 +311,7 @@ "group": "Release Notes", "pages": [ "en/releases/overview", + "en/releases/v2.0.2", "en/releases/v2.0.1", "en/releases/v2.0.0" ] diff --git a/docs/en/channels/dingtalk.mdx b/docs/en/channels/dingtalk.mdx index b896469c..1db5f53c 100644 --- a/docs/en/channels/dingtalk.mdx +++ b/docs/en/channels/dingtalk.mdx @@ -7,15 +7,29 @@ Integrate CowAgent into DingTalk by creating an intelligent robot app on the Din ## 1. Create App -1. Go to [DingTalk Developer Console](https://open-dev.dingtalk.com/fe/app#/corp/app), click **Create App**, fill in app information -2. Click **Add App Capability**, select **Robot** capability and add -3. Configure robot information and click **Publish** +1. Go to [DingTalk Developer Console](https://open-dev.dingtalk.com/fe/app#/corp/app), log in and click **Create App**, fill in the app information: + + + +2. Click **Add App Capability**, select **Robot** capability and click **Add**: + + + +3. Configure the robot information and click **Publish**. After publishing, click "**Debug**" to automatically create a test group chat, which can be viewed in the client: + + + +4. Click **Version Management & Release**, create a new version and publish: + + ## 2. Project Configuration -1. Get `Client ID` and `Client Secret` from **Credentials & Basic Info** +1. Click **Credentials & Basic Info**, get the `Client ID` and `Client Secret`: -2. Fill in `config.json`: + + +2. Add the following configuration to `config.json` in the project root: ```json { @@ -25,14 +39,20 @@ Integrate CowAgent into DingTalk by creating an intelligent robot app on the Din } ``` -3. Install dependency: +3. Install the dependency: ```bash pip3 install dingtalk_stream ``` -4. After starting the project, go to DingTalk Developer Console **Event Subscription**, click **Connection verified, verify channel**. When "Connection successful" is displayed, configuration is complete + + +4. After starting the project, go to the DingTalk Developer Console, click **Event Subscription**, then click **Connection verified, verify channel**. When "**Connection successful**" is displayed, the configuration is complete: + + ## 3. Usage -Chat privately with the robot or add it to an enterprise group to start a conversation. +Chat privately with the robot or add it to an enterprise group to start a conversation: + + diff --git a/docs/en/channels/feishu.mdx b/docs/en/channels/feishu.mdx index 7a742744..de90da04 100644 --- a/docs/en/channels/feishu.mdx +++ b/docs/en/channels/feishu.mdx @@ -3,65 +3,67 @@ title: Feishu (Lark) description: Integrate CowAgent into Feishu application --- -Integrate CowAgent into Feishu by creating a custom app. Supports WebSocket (recommended, no public IP required) and Webhook event receiving modes. +Integrate CowAgent into Feishu by creating a custom enterprise app. You need to be a Feishu enterprise user with admin privileges. ## 1. Create Enterprise Custom App ### 1.1 Create App -Go to [Feishu Developer Platform](https://open.feishu.cn/app/), click **Create Enterprise Custom App**, fill in the required information and create. +Go to [Feishu Developer Platform](https://open.feishu.cn/app/), click **Create Enterprise Custom App**, fill in the required information and click **Create**: + + ### 1.2 Add Bot Capability -In **Add App Capabilities**, add **Bot** capability to the app. +In **Add App Capabilities**, add **Bot** capability to the app: + + ### 1.3 Configure App Permissions -Click **Permission Management**, paste the following permission string, select all and enable in batch: +Click **Permission Management**, paste the following permission string into the input box below **Permission Configuration**, select all filtered permissions, click **Batch Enable** and confirm: ``` im:message,im:message.group_at_msg,im:message.group_at_msg:readonly,im:message.p2p_msg,im:message.p2p_msg:readonly,im:message:send_as_bot,im:resource ``` + + ## 2. Project Configuration -Get `App ID` and `App Secret` from **Credentials & Basic Info**, then fill in `config.json`: +1. Get `App ID` and `App Secret` from **Credentials & Basic Info**: - - - No public IP required. Configuration: + - ```json - { - "channel_type": "feishu", - "feishu_app_id": "YOUR_APP_ID", - "feishu_app_secret": "YOUR_APP_SECRET", - "feishu_event_mode": "websocket" - } - ``` +2. Add the following configuration to `config.json` in the project root: - Install dependency: `pip3 install lark-oapi` - - - Requires public IP. Configuration: +```json +{ + "channel_type": "feishu", + "feishu_app_id": "YOUR_APP_ID", + "feishu_app_secret": "YOUR_APP_SECRET", + "feishu_bot_name": "YOUR_BOT_NAME" +} +``` - ```json - { - "channel_type": "feishu", - "feishu_app_id": "YOUR_APP_ID", - "feishu_app_secret": "YOUR_APP_SECRET", - "feishu_token": "VERIFICATION_TOKEN", - "feishu_event_mode": "webhook", - "feishu_port": 9891 - } - ``` - - +| Parameter | Description | +| --- | --- | +| `feishu_app_id` | Feishu bot App ID | +| `feishu_app_secret` | Feishu bot App Secret | +| `feishu_bot_name` | Bot name (set when creating the app), required for group chat usage | + +Start the project after configuration is complete. ## 3. Configure Event Subscription -1. After starting the project, go to Feishu Developer Platform **Events & Callbacks**, select **Long Connection** and save -2. Click **Add Event**, search for "Receive Message", select "Receive Message v2.0", confirm and add -3. Click **Version Management & Release**, create a version and apply for production release. After approval, you can use it +1. After the project is running successfully, go to the Feishu Developer Platform, click **Events & Callbacks**, select **Long Connection** mode, and click save: -Search for the bot name in Feishu to start chatting. + + +2. Click **Add Event** below, search for "Receive Message", select "**Receive Message v2.0**", and confirm. + +3. Click **Version Management & Release**, create a new version and apply for **Production Release**. Check the approval message in the Feishu client and approve: + + + +Once completed, search for the bot name in Feishu to start chatting. diff --git a/docs/en/channels/web.mdx b/docs/en/channels/web.mdx index d8be9263..845b2196 100644 --- a/docs/en/channels/web.mdx +++ b/docs/en/channels/web.mdx @@ -1,9 +1,9 @@ --- -title: Web -description: Use CowAgent through the web interface +title: Web Console +description: Use CowAgent through the web console --- -Web is CowAgent's default channel. The web console starts automatically after launch, allowing you to chat with the Agent through a browser. +The Web Console is CowAgent's default channel. It starts automatically after launch, allowing you to chat with the Agent through a browser and manage models, skills, memory, channels, and other configurations online. ## Configuration @@ -19,13 +19,57 @@ Web is CowAgent's default channel. The web console starts automatically after la | `channel_type` | Set to `web` | `web` | | `web_port` | Web service listen port | `9899` | -## Usage +## Access URL After starting the project, visit: -- Local: `http://localhost:9899/chat` -- Server: `http://:9899/chat` +- Local: `http://localhost:9899` +- Server: `http://:9899` Ensure the server firewall and security group allow the corresponding port. + +## Features + +### Chat Interface + +Supports streaming output with real-time display of the Agent's reasoning process and tool calls, providing intuitive observation of the Agent's decision-making: + + + +### Model Management + +Manage model configurations online without manually editing config files: + + + +### Skill Management + +View and manage Agent skills (Skills) online: + + + +### Memory Management + +View and manage Agent memory online: + + + +### Channel Management + +Manage connected channels online with real-time connect/disconnect operations: + + + +### Scheduled Tasks + +View and manage scheduled tasks online, including one-time tasks, fixed intervals, and Cron expressions: + + + +### Logs + +View Agent runtime logs in real-time for monitoring and troubleshooting: + + diff --git a/docs/en/channels/wechatmp.mdx b/docs/en/channels/wechatmp.mdx index 1de9c8ae..3c6c2c8b 100644 --- a/docs/en/channels/wechatmp.mdx +++ b/docs/en/channels/wechatmp.mdx @@ -7,19 +7,22 @@ CowAgent supports both personal subscription accounts and enterprise service acc | Type | Requirements | Features | | --- | --- | --- | -| **Personal Subscription** | Available to individuals | Users must send a message to retrieve replies | +| **Personal Subscription** | Available to individuals | Sends a placeholder reply first; users must send a message to retrieve the full response | | **Enterprise Service** | Enterprise with verified customer service API | Can proactively push replies to users | - Official Accounts only support server and Docker deployment. Install extended dependencies: `pip3 install -r requirements-optional.txt` + Official Accounts only support server and Docker deployment, not local run mode. Install extended dependencies: `pip3 install -r requirements-optional.txt` -## Personal Subscription Account +## 1. Personal Subscription Account + +Add the following configuration to `config.json`: ```json { "channel_type": "wechatmp", - "wechatmp_app_id": "YOUR_APP_ID", + "single_chat_prefix": [""], + "wechatmp_app_id": "wx73f9******d1e48", "wechatmp_app_secret": "YOUR_APP_SECRET", "wechatmp_aes_key": "", "wechatmp_token": "YOUR_TOKEN", @@ -29,22 +32,37 @@ CowAgent supports both personal subscription accounts and enterprise service acc ### Setup Steps -1. Get parameters from [WeChat Official Account Platform](https://mp.weixin.qq.com/) under **Settings & Development → Basic Configuration → Server Configuration** -2. Enable developer secret and add server IP to the whitelist -3. Start the program (listens on port 80) -4. Enable server configuration with URL format `http://{HOST}/wx` +These configurations must be consistent with the [WeChat Official Account Platform](https://mp.weixin.qq.com/advanced/advanced?action=dev&t=advanced/dev). Navigate to **Settings & Development → Basic Configuration → Server Configuration** and configure as shown below: -## Enterprise Service Account + -Same setup with these differences: +1. Enable the developer secret on the platform (corresponds to `wechatmp_app_secret`), and add the server IP to the whitelist +2. Fill in the `config.json` with the official account parameters matching the platform configuration +3. Start the program, which listens on port 80 (use `sudo` if you don't have permission; stop any process occupying port 80) +4. **Enable server configuration** on the official account platform and submit. A successful save means the configuration is complete. Note that the **"Server URL"** must be in the format `http://{HOST}/wx`, where `{HOST}` can be the server IP or domain -1. Register an enterprise service account with verified **Customer Service API** permission -2. Set `"channel_type": "wechatmp_service"` in `config.json` -3. Replies can be proactively pushed to users +After following the account and sending a message, you should see the following result: + + + +Due to subscription account limitations, short replies (within 15s) can be returned immediately, but longer replies will first send a "Thinking..." placeholder, requiring users to send any text to retrieve the answer. Enterprise service accounts can solve this with the customer service API. + + + **Voice Recognition**: You can use WeChat's built-in voice recognition. Enable "Receive Voice Recognition Results" under "Settings & Development → API Permissions" on the official account management page. + + +## 2. Enterprise Service Account + +The setup process for enterprise service accounts is essentially the same as personal subscription accounts, with the following differences: + +1. Register an enterprise service account on the platform and complete WeChat certification. Confirm that the **Customer Service API** permission has been granted +2. Set `"channel_type": "wechatmp_service"` in `config.json`; other configurations remain the same +3. Even for longer replies, they can be proactively pushed to users without requiring manual retrieval ```json { "channel_type": "wechatmp_service", + "single_chat_prefix": [""], "wechatmp_app_id": "YOUR_APP_ID", "wechatmp_app_secret": "YOUR_APP_SECRET", "wechatmp_aes_key": "", diff --git a/docs/en/channels/wecom.mdx b/docs/en/channels/wecom.mdx index 19608396..7b07cd06 100644 --- a/docs/en/channels/wecom.mdx +++ b/docs/en/channels/wecom.mdx @@ -13,23 +13,41 @@ Integrate CowAgent into WeCom through a custom enterprise app, supporting one-on Required resources: -1. A server with public IP -2. A registered WeCom account (individual registration is possible, but cannot be certified) -3. Certified WeCom requires a domain with corresponding entity filing +1. A server with public IP (overseas server, or domestic server with a proxy for international API access) +2. A registered WeCom account (individual registration is possible but cannot be certified) +3. Certified WeCom accounts additionally require a domain filed under the corresponding entity ## 2. Create WeCom App -1. Get **Corp ID** from **My Enterprise** in [WeCom Admin Console](https://work.weixin.qq.com/wework_admin/frame#profile) -2. Switch to **Application Management**, click Create Application, record `AgentId` and `Secret` -3. Click **Set API Reception**, configure application interface: - - URL format: `http://ip:port/wxcomapp` (certified enterprises must use filed domain) - - Generate random `Token` and `EncodingAESKey` and save +1. In the [WeCom Admin Console](https://work.weixin.qq.com/wework_admin/frame#profile), click **My Enterprise** and find the **Corp ID** at the bottom of the page. Save this ID for the `wechatcom_corp_id` configuration field. + +2. Switch to **Application Management** and click Create Application: + + + +3. On the application creation page, record the `AgentId` and `Secret`: + + + +4. Click **Set API Reception** to configure the application interface: + + + +- URL format: `http://ip:port/wxcomapp` (certified enterprises must use a filed domain) +- Generate random `Token` and `EncodingAESKey` and save them for the configuration file + + + The API reception configuration cannot be saved at this point because the program hasn't started yet. Come back to save it after the project is running. + ## 3. Configuration and Run +Add the following configuration to `config.json` (the mapping between each parameter and the WeCom console is shown in the screenshots above): + ```json { "channel_type": "wechatcom_app", + "single_chat_prefix": [""], "wechatcom_corp_id": "YOUR_CORP_ID", "wechatcomapp_token": "YOUR_TOKEN", "wechatcomapp_secret": "YOUR_SECRET", @@ -48,12 +66,25 @@ Required resources: | `wechatcomapp_aes_key` | EncodingAESKey from API reception config | | `wechatcomapp_port` | Listen port, default 9898 | -After starting the program, return to WeCom Admin Console to save **Message Server Configuration**, and add the server IP to **Enterprise Trusted IPs**. +After configuration, start the program. When the log shows `http://0.0.0.0:9898/`, the program is running successfully. You need to open this port externally (e.g., allow it in the cloud server security group). + +After the program starts, return to the WeCom Admin Console to save the **Message Server Configuration**. After saving successfully, you also need to add the server IP to **Enterprise Trusted IPs**, otherwise messages cannot be sent or received: + + - If configuration fails: 1. Ensure firewall and security group allow the port; 2. Verify all parameters are consistent; 3. Certified enterprises must configure a filed domain. + If the URL configuration callback fails or the configuration is unsuccessful: + 1. Ensure the server firewall is disabled and the security group allows the listening port + 2. Carefully check that Token, Secret Key and other parameter configurations are consistent, and that the URL format is correct + 3. Certified WeCom accounts must configure a filed domain matching the entity ## 4. Usage -Search for the app name in WeCom to start chatting. To allow external WeChat users, share the invite QR code from **My Enterprise → WeChat Plugin**. +Search for the app name you just created in WeCom to start chatting directly. You can run multiple instances listening on different ports to create multiple WeCom apps: + + + +To allow external personal WeChat users to use the app, go to **My Enterprise → WeChat Plugin**, share the invite QR code. After scanning and following, personal WeChat users can join and chat with the app: + + diff --git a/docs/en/releases/overview.mdx b/docs/en/releases/overview.mdx index 8d6019c3..38445fd4 100644 --- a/docs/en/releases/overview.mdx +++ b/docs/en/releases/overview.mdx @@ -5,6 +5,7 @@ description: CowAgent version history | Version | Date | Description | | --- | --- | --- | +| [2.0.2](/en/releases/v2.0.2) | 2026.02.27 | Web Console upgrade, multi-channel concurrency, session persistence | | [2.0.1](/en/releases/v2.0.1) | 2026.02.27 | Built-in Web Search tool, smart context management, multiple fixes | | [2.0.0](/en/releases/v2.0.0) | 2026.02.03 | Full upgrade to AI super assistant | | 1.7.6 | 2025.05.23 | Web Channel optimization, AgentMesh plugin | diff --git a/docs/en/releases/v2.0.2.mdx b/docs/en/releases/v2.0.2.mdx new file mode 100644 index 00000000..b833adfa --- /dev/null +++ b/docs/en/releases/v2.0.2.mdx @@ -0,0 +1,98 @@ +--- +title: v2.0.2 +description: CowAgent 2.0.2 - Web Console upgrade, multi-channel concurrency, session persistence +--- + +**Release Date**: 2026.02.27 | [Full Changelog](https://github.com/zhayujie/chatgpt-on-wechat/compare/2.0.1...master) + +## Highlights + +### 🖥️ Web Console Upgrade + +The Web Console has been fully upgraded with streaming conversation output, visual display of tool execution and reasoning processes, and online management of **models, skills, memory, channels, and Agent configuration**. + +#### Chat Interface + +Supports streaming output with real-time display of the Agent's reasoning process and tool calls, providing intuitive observation of the Agent's decision-making: + + + +#### Model Management + +Manage model configurations online without manually editing config files: + + + +#### Skill Management + +View and manage Agent skills (Skills) online: + + + +#### Memory Management + +View and manage Agent memory online: + + + +#### Channel Management + +Manage connected channels online with real-time connect/disconnect operations: + + + +#### Scheduled Tasks + +View and manage scheduled tasks online, including one-time tasks, fixed intervals, and Cron expressions: + + + +#### Logs + +View Agent runtime logs in real-time for monitoring and troubleshooting: + + + +Related commits: [f1a1413](https://github.com/zhayujie/chatgpt-on-wechat/commit/f1a1413), [c0702c8](https://github.com/zhayujie/chatgpt-on-wechat/commit/c0702c8), [394853c](https://github.com/zhayujie/chatgpt-on-wechat/commit/394853c), [1c71c4e](https://github.com/zhayujie/chatgpt-on-wechat/commit/1c71c4e), [5e3eccb](https://github.com/zhayujie/chatgpt-on-wechat/commit/5e3eccb), [e1dc037](https://github.com/zhayujie/chatgpt-on-wechat/commit/e1dc037), [5edbf4c](https://github.com/zhayujie/chatgpt-on-wechat/commit/5edbf4c), [7d258b5](https://github.com/zhayujie/chatgpt-on-wechat/commit/7d258b5) + +### 🔀 Multi-Channel Concurrency + +Multiple channels (e.g., Feishu, DingTalk, WeCom, Web) can now run simultaneously, each in an independent thread without interference. + +Configuration: Set multiple channels in `config.json` via `channel_type` separated by commas, or connect/disconnect channels in real-time from the Web Console's channel management page. + +```json +{ + "channel_type": "web,feishu,dingtalk" +} +``` + +Related commits: [4694594](https://github.com/zhayujie/chatgpt-on-wechat/commit/4694594), [7cce224](https://github.com/zhayujie/chatgpt-on-wechat/commit/7cce224), [7d258b5](https://github.com/zhayujie/chatgpt-on-wechat/commit/7d258b5), [c9adddb](https://github.com/zhayujie/chatgpt-on-wechat/commit/c9adddb) + +### 💾 Session Persistence + +Session history is now persisted to a local SQLite database. Conversation context is automatically restored after service restarts. Historical conversations in the Web Console are also restored. + +Related commits: [29bfbec](https://github.com/zhayujie/chatgpt-on-wechat/commit/29bfbec), [9917552](https://github.com/zhayujie/chatgpt-on-wechat/commit/9917552), [925d728](https://github.com/zhayujie/chatgpt-on-wechat/commit/925d728) + +## New Models + +- **Gemini 3.1 Pro Preview**: Added `gemini-3.1-pro-preview` model support ([52d7cad](https://github.com/zhayujie/chatgpt-on-wechat/commit/52d7cad)) +- **Claude 4.6 Sonnet**: Added `claude-4.6-sonnet` model support ([52d7cad](https://github.com/zhayujie/chatgpt-on-wechat/commit/52d7cad)) +- **Qwen3.5 Plus**: Added `qwen3.5-plus` model support ([e59a289](https://github.com/zhayujie/chatgpt-on-wechat/commit/e59a289)) +- **MiniMax M2.5**: Added `Minimax-M2.5` model support ([48db538](https://github.com/zhayujie/chatgpt-on-wechat/commit/48db538)) +- **GLM-5**: Added `glm-5` model support ([48db538](https://github.com/zhayujie/chatgpt-on-wechat/commit/48db538)) +- **Kimi K2.5**: Added `kimi-k2.5` model support ([48db538](https://github.com/zhayujie/chatgpt-on-wechat/commit/48db538)) +- **Doubao 2.0 Code**: Added `doubao-2.0-code` coding-specialized model ([ab28ee5](https://github.com/zhayujie/chatgpt-on-wechat/commit/ab28ee5)) +- **DashScope Models**: Added Alibaba Cloud DashScope model name support ([ce58f23](https://github.com/zhayujie/chatgpt-on-wechat/commit/ce58f23)) + +## Website & Documentation + +- **Official Website**: [cowagent.ai](https://cowagent.ai/) +- **Documentation**: [docs.cowagent.ai](https://docs.cowagent.ai/) + +## Bug Fixes + +- **Gemini DingTalk image recognition**: Fixed Gemini unable to process image markers in DingTalk channel ([05a3304](https://github.com/zhayujie/chatgpt-on-wechat/commit/05a3304)) ([#2670](https://github.com/zhayujie/chatgpt-on-wechat/pull/2670)) Thanks [@SgtPepper114](https://github.com/SgtPepper114) +- **Startup script dependencies**: Fixed dependency installation issue in `run.sh` script ([b6fc9fa](https://github.com/zhayujie/chatgpt-on-wechat/commit/b6fc9fa)) +- **Bare except cleanup**: Replaced `bare except` with `except Exception` for better exception handling ([adca89b](https://github.com/zhayujie/chatgpt-on-wechat/commit/adca89b)) ([#2674](https://github.com/zhayujie/chatgpt-on-wechat/pull/2674)) Thanks [@haosenwang1018](https://github.com/haosenwang1018) diff --git a/docs/guide/manual-install.mdx b/docs/guide/manual-install.mdx index 51bb15c1..311bbe28 100644 --- a/docs/guide/manual-install.mdx +++ b/docs/guide/manual-install.mdx @@ -48,7 +48,7 @@ cp config-template.json config.json python3 app.py ``` -运行后默认启动 Web 服务，访问 `http://localhost:9899/chat` 开始对话。 +运行后默认启动 Web 控制台，访问 `http://localhost:9899` 开始对话和管理Agent。 **服务器后台运行：** diff --git a/docs/guide/quick-start.mdx b/docs/guide/quick-start.mdx index f9529d1c..2143941e 100644 --- a/docs/guide/quick-start.mdx +++ b/docs/guide/quick-start.mdx @@ -22,7 +22,7 @@ bash <(curl -sS https://cdn.link-ai.tech/code/cow/run.sh) 5. 引导配置 AI 模型和通信渠道 6. 启动服务 -运行后默认启动 Web 服务，访问 `http://localhost:9899/chat` 开始对话。 +运行后默认启动 Web 控制台，访问 `http://localhost:9899` 开始对话和管理Agent。 ## 管理命令 From fccfa92d7eabb5916161da4867fac6b40373bb4d Mon Sep 17 00:00:00 2001 From: zhayujie Date: Sat, 28 Feb 2026 14:50:55 +0800 Subject: [PATCH 03/10] docs: update channel docs --- docs/en/intro/architecture.mdx | 8 ++++---- docs/en/skills/image-vision.mdx | 2 -- docs/en/skills/linkai-agent.mdx | 2 -- docs/en/skills/skill-creator.mdx | 2 -- docs/en/skills/web-fetch.mdx | 2 -- docs/en/tools/bash.mdx | 2 -- docs/en/tools/browser.mdx | 2 -- docs/en/tools/edit.mdx | 2 -- docs/en/tools/env-config.mdx | 2 -- docs/en/tools/ls.mdx | 2 -- docs/en/tools/memory.mdx | 2 -- docs/en/tools/read.mdx | 2 -- docs/en/tools/scheduler.mdx | 2 -- docs/en/tools/send.mdx | 2 -- docs/en/tools/web-search.mdx | 2 -- docs/en/tools/write.mdx | 2 -- docs/intro/architecture.mdx | 8 ++++---- docs/models/doubao.mdx | 2 -- docs/models/glm.mdx | 2 -- docs/models/kimi.mdx | 2 -- docs/models/qwen.mdx | 2 -- docs/skills/image-vision.mdx | 2 -- docs/skills/linkai-agent.mdx | 2 -- docs/skills/skill-creator.mdx | 2 -- docs/skills/web-fetch.mdx | 2 -- docs/tools/bash.mdx | 2 -- docs/tools/browser.mdx | 2 -- docs/tools/edit.mdx | 2 -- docs/tools/env-config.mdx | 2 -- docs/tools/ls.mdx | 2 -- docs/tools/memory.mdx | 2 -- docs/tools/read.mdx | 2 -- docs/tools/scheduler.mdx | 2 -- docs/tools/send.mdx | 2 -- docs/tools/web-search.mdx | 2 -- docs/tools/write.mdx | 2 -- 36 files changed, 8 insertions(+), 76 deletions(-) diff --git a/docs/en/intro/architecture.mdx b/docs/en/intro/architecture.mdx index 8958da86..3cdf619d 100644 --- a/docs/en/intro/architecture.mdx +++ b/docs/en/intro/architecture.mdx @@ -33,10 +33,10 @@ When Agent mode is enabled, CowAgent runs as an autonomous agent with the follow ## Workspace Directory Structure -The Agent workspace is located at `~/cow` by default and stores system prompts, memory files, and skill files: +The Agent workspace is located at `~/.cow` by default and stores system prompts, memory files, and skill files: ``` -~/cow/ +~/.cow/ ├── system.md # Agent system prompt ├── user.md # User profile ├── memory/ # Long-term memory storage @@ -55,7 +55,7 @@ Configure Agent mode parameters in `config.json`: ```json { "agent": true, - "agent_workspace": "~/cow", + "agent_workspace": "~/.cow", "agent_max_context_tokens": 40000, "agent_max_context_turns": 30, "agent_max_steps": 15 @@ -65,7 +65,7 @@ Configure Agent mode parameters in `config.json`: | Parameter | Description | Default | | --- | --- | --- | | `agent` | Enable Agent mode | `true` | -| `agent_workspace` | Workspace path | `~/cow` | +| `agent_workspace` | Workspace path | `~/.cow` | | `agent_max_context_tokens` | Max context tokens | `40000` | | `agent_max_context_turns` | Max context turns | `30` | | `agent_max_steps` | Max decision steps per task | `15` | diff --git a/docs/en/skills/image-vision.mdx b/docs/en/skills/image-vision.mdx index 63543f4f..8cc5c37e 100644 --- a/docs/en/skills/image-vision.mdx +++ b/docs/en/skills/image-vision.mdx @@ -3,8 +3,6 @@ title: Image Vision description: Recognize images using OpenAI vision models --- -# openai-image-vision - Analyze image content using OpenAI's GPT-4 Vision API, understanding objects, text, colors, and other elements in images. ## Dependencies diff --git a/docs/en/skills/linkai-agent.mdx b/docs/en/skills/linkai-agent.mdx index db891341..d0edc05a 100644 --- a/docs/en/skills/linkai-agent.mdx +++ b/docs/en/skills/linkai-agent.mdx @@ -3,8 +3,6 @@ title: LinkAI Agent description: Integrate LinkAI platform multi-agent skill --- -# linkai-agent - Use agents from the [LinkAI](https://link-ai.tech/) platform as Skills for multi-agent decision-making. The Agent intelligently selects based on agent names and descriptions, calling the corresponding application or workflow via `app_code`. ## Dependencies diff --git a/docs/en/skills/skill-creator.mdx b/docs/en/skills/skill-creator.mdx index d3ff1689..6c5c3f73 100644 --- a/docs/en/skills/skill-creator.mdx +++ b/docs/en/skills/skill-creator.mdx @@ -3,8 +3,6 @@ title: Skill Creator description: Create custom skills through conversation --- -# skill-creator - Quickly create, install, or update skills through natural language conversation. ## Dependencies diff --git a/docs/en/skills/web-fetch.mdx b/docs/en/skills/web-fetch.mdx index d37923a0..f52077de 100644 --- a/docs/en/skills/web-fetch.mdx +++ b/docs/en/skills/web-fetch.mdx @@ -3,8 +3,6 @@ title: Web Fetch description: Fetch web page text content --- -# web-fetch - Use curl to fetch web pages and extract readable text content. A lightweight web access method without browser automation. ## Dependencies diff --git a/docs/en/tools/bash.mdx b/docs/en/tools/bash.mdx index db107552..60b20918 100644 --- a/docs/en/tools/bash.mdx +++ b/docs/en/tools/bash.mdx @@ -3,8 +3,6 @@ title: bash - Terminal description: Execute system commands --- -# bash - Execute Bash commands in the current working directory, returns stdout and stderr. API keys configured via `env_config` are automatically injected into the environment. ## Dependencies diff --git a/docs/en/tools/browser.mdx b/docs/en/tools/browser.mdx index 6efce9f4..b8e7829e 100644 --- a/docs/en/tools/browser.mdx +++ b/docs/en/tools/browser.mdx @@ -3,8 +3,6 @@ title: browser - Browser description: Access and interact with web pages --- -# browser - Use a browser to access and interact with web pages, supports JavaScript-rendered dynamic pages. ## Dependencies diff --git a/docs/en/tools/edit.mdx b/docs/en/tools/edit.mdx index accc9275..f231c6b9 100644 --- a/docs/en/tools/edit.mdx +++ b/docs/en/tools/edit.mdx @@ -3,8 +3,6 @@ title: edit - File Edit description: Edit files via precise text replacement --- -# edit - Edit files via precise text replacement. If `oldText` is empty, appends to the end of the file. ## Dependencies diff --git a/docs/en/tools/env-config.mdx b/docs/en/tools/env-config.mdx index 1bd1c6c7..23f75bf8 100644 --- a/docs/en/tools/env-config.mdx +++ b/docs/en/tools/env-config.mdx @@ -3,8 +3,6 @@ title: env_config - Environment description: Manage API keys and secrets --- -# env_config - Manage environment variables (API keys and secrets) in the workspace `.env` file, with secure conversational updates. Built-in security protection and desensitization. ## Dependencies diff --git a/docs/en/tools/ls.mdx b/docs/en/tools/ls.mdx index 984a3efe..e9a5f656 100644 --- a/docs/en/tools/ls.mdx +++ b/docs/en/tools/ls.mdx @@ -3,8 +3,6 @@ title: ls - Directory List description: List directory contents --- -# ls - List directory contents, sorted alphabetically, directories suffixed with `/`, includes hidden files. ## Dependencies diff --git a/docs/en/tools/memory.mdx b/docs/en/tools/memory.mdx index 94b4e5ab..a7526336 100644 --- a/docs/en/tools/memory.mdx +++ b/docs/en/tools/memory.mdx @@ -3,8 +3,6 @@ title: memory - Memory description: Search and read long-term memory --- -# memory - The memory tool contains two sub-tools: `memory_search` (search memory) and `memory_get` (read memory files). ## Dependencies diff --git a/docs/en/tools/read.mdx b/docs/en/tools/read.mdx index 9ac43dac..56b56570 100644 --- a/docs/en/tools/read.mdx +++ b/docs/en/tools/read.mdx @@ -3,8 +3,6 @@ title: read - File Read description: Read file content --- -# read - Read file content. Supports text files, PDF files, images (returns metadata), and more. ## Dependencies diff --git a/docs/en/tools/scheduler.mdx b/docs/en/tools/scheduler.mdx index a79d4536..0a5520c9 100644 --- a/docs/en/tools/scheduler.mdx +++ b/docs/en/tools/scheduler.mdx @@ -3,8 +3,6 @@ title: scheduler - Scheduler description: Create and manage scheduled tasks --- -# scheduler - Create and manage dynamic scheduled tasks with flexible scheduling and execution modes. ## Dependencies diff --git a/docs/en/tools/send.mdx b/docs/en/tools/send.mdx index f1d7cfcc..1cf089ac 100644 --- a/docs/en/tools/send.mdx +++ b/docs/en/tools/send.mdx @@ -3,8 +3,6 @@ title: send - File Send description: Send files to user --- -# send - Send files to the user (images, videos, audio, documents, etc.), used when the user explicitly requests to send/share a file. ## Dependencies diff --git a/docs/en/tools/web-search.mdx b/docs/en/tools/web-search.mdx index 5e73491d..d9d088c7 100644 --- a/docs/en/tools/web-search.mdx +++ b/docs/en/tools/web-search.mdx @@ -3,8 +3,6 @@ title: web_search - Web Search description: Search the internet for real-time information --- -# web_search - Search the internet for real-time information, news, research, and more. Supports two search backends with automatic fallback. ## Dependencies diff --git a/docs/en/tools/write.mdx b/docs/en/tools/write.mdx index 42f81f21..2c0a10f0 100644 --- a/docs/en/tools/write.mdx +++ b/docs/en/tools/write.mdx @@ -3,8 +3,6 @@ title: write - File Write description: Create or overwrite files --- -# write - Write content to a file. Creates the file if it doesn't exist, overwrites if it does. Automatically creates parent directories. ## Dependencies diff --git a/docs/intro/architecture.mdx b/docs/intro/architecture.mdx index 39ff09a1..67aff375 100644 --- a/docs/intro/architecture.mdx +++ b/docs/intro/architecture.mdx @@ -33,10 +33,10 @@ CowAgent 的整体架构由以下核心模块组成： ## 工作空间 -Agent 的工作空间默认位于 `~/cow` 目录，用于存储系统提示词、记忆文件、技能文件等： +Agent 的工作空间默认位于 `~/.cow` 目录，用于存储系统提示词、记忆文件、技能文件等： ``` -~/cow/ +~/.cow/ ├── system.md # Agent system prompt ├── user.md # User profile ├── memory/ # Long-term memory storage @@ -55,7 +55,7 @@ Agent 的工作空间默认位于 `~/cow` 目录，用于存储系统提示词 ```json { "agent": true, - "agent_workspace": "~/cow", + "agent_workspace": "~/.cow", "agent_max_context_tokens": 40000, "agent_max_context_turns": 30, "agent_max_steps": 15 @@ -65,7 +65,7 @@ Agent 的工作空间默认位于 `~/cow` 目录，用于存储系统提示词 | 参数 | 说明 | 默认值 | | --- | --- | --- | | `agent` | 是否启用 Agent 模式 | `true` | -| `agent_workspace` | 工作空间路径 | `~/cow` | +| `agent_workspace` | 工作空间路径 | `~/.cow` | | `agent_max_context_tokens` | 最大上下文 token 数 | `40000` | | `agent_max_context_turns` | 最大上下文记忆轮次 | `30` | | `agent_max_steps` | 单次任务最大决策步数 | `15` | diff --git a/docs/models/doubao.mdx b/docs/models/doubao.mdx index 6895916d..e7440434 100644 --- a/docs/models/doubao.mdx +++ b/docs/models/doubao.mdx @@ -3,8 +3,6 @@ title: 豆包 Doubao description: 豆包 (火山方舟) 模型配置 --- -# 豆包 (Doubao) - ```json { "model": "doubao-seed-2-0-code-preview-260215", diff --git a/docs/models/glm.mdx b/docs/models/glm.mdx index bee5f4bf..bc393db8 100644 --- a/docs/models/glm.mdx +++ b/docs/models/glm.mdx @@ -3,8 +3,6 @@ title: 智谱 GLM description: 智谱AI GLM 模型配置 --- -# 智谱AI (GLM) - ```json { "model": "glm-5", diff --git a/docs/models/kimi.mdx b/docs/models/kimi.mdx index 32dd5007..1ddf8a6f 100644 --- a/docs/models/kimi.mdx +++ b/docs/models/kimi.mdx @@ -3,8 +3,6 @@ title: Kimi description: Kimi (Moonshot) 模型配置 --- -# Kimi (Moonshot) - ```json { "model": "kimi-k2.5", diff --git a/docs/models/qwen.mdx b/docs/models/qwen.mdx index a8aae922..f3c75063 100644 --- a/docs/models/qwen.mdx +++ b/docs/models/qwen.mdx @@ -3,8 +3,6 @@ title: 通义千问 Qwen description: 通义千问模型配置 --- -# 通义千问 (Qwen) - ```json { "model": "qwen3.5-plus", diff --git a/docs/skills/image-vision.mdx b/docs/skills/image-vision.mdx index c5bb4b49..1c71f620 100644 --- a/docs/skills/image-vision.mdx +++ b/docs/skills/image-vision.mdx @@ -3,8 +3,6 @@ title: 图像识别 description: 使用 OpenAI 视觉模型识别图片 --- -# openai-image-vision - 使用 OpenAI 的 GPT-4 Vision API 分析图片内容，理解图像中的物体、文字、颜色等元素。 ## 依赖 diff --git a/docs/skills/linkai-agent.mdx b/docs/skills/linkai-agent.mdx index a0abc3be..38fc735e 100644 --- a/docs/skills/linkai-agent.mdx +++ b/docs/skills/linkai-agent.mdx @@ -3,8 +3,6 @@ title: LinkAI 智能体 description: 对接 LinkAI 平台的多智能体技能 --- -# linkai-agent - 将 [LinkAI](https://link-ai.tech/) 平台上的智能体作为 Skill 使用，实现多智能体决策。Agent 根据智能体的名称和描述智能选择，通过 `app_code` 调用对应的应用或工作流。 ## 依赖 diff --git a/docs/skills/skill-creator.mdx b/docs/skills/skill-creator.mdx index 5f35da00..a530195b 100644 --- a/docs/skills/skill-creator.mdx +++ b/docs/skills/skill-creator.mdx @@ -3,8 +3,6 @@ title: 创建技能 description: 通过对话创建自定义技能 --- -# skill-creator - 通过自然语言对话快速创建、安装或更新技能。 ## 依赖 diff --git a/docs/skills/web-fetch.mdx b/docs/skills/web-fetch.mdx index 71c239a4..eedf49b8 100644 --- a/docs/skills/web-fetch.mdx +++ b/docs/skills/web-fetch.mdx @@ -3,8 +3,6 @@ title: 网页抓取 description: 抓取网页文本内容 --- -# web-fetch - 使用 curl 抓取网页并提取可读文本内容，轻量级的网页访问方式，无需浏览器自动化。 ## 依赖 diff --git a/docs/tools/bash.mdx b/docs/tools/bash.mdx index 6ea31769..0090fe14 100644 --- a/docs/tools/bash.mdx +++ b/docs/tools/bash.mdx @@ -3,8 +3,6 @@ title: bash - 终端 description: 执行系统命令 --- -# bash - 在当前工作目录执行 Bash 命令，返回 stdout 和 stderr。`env_config` 中配置的 API Key 会自动注入到环境变量中。 ## 依赖 diff --git a/docs/tools/browser.mdx b/docs/tools/browser.mdx index 88e0fcfd..8c473daf 100644 --- a/docs/tools/browser.mdx +++ b/docs/tools/browser.mdx @@ -3,8 +3,6 @@ title: browser - 浏览器 description: 访问和操作网页 --- -# browser - 使用浏览器访问和操作网页，支持 JavaScript 渲染的动态页面。 ## 依赖 diff --git a/docs/tools/edit.mdx b/docs/tools/edit.mdx index 2a8f2916..717af2ba 100644 --- a/docs/tools/edit.mdx +++ b/docs/tools/edit.mdx @@ -3,8 +3,6 @@ title: edit - 文件编辑 description: 通过精确文本替换编辑文件 --- -# edit - 通过精确文本替换编辑文件。如果 `oldText` 为空则追加到文件末尾。 ## 依赖 diff --git a/docs/tools/env-config.mdx b/docs/tools/env-config.mdx index 479ef214..d5d52c68 100644 --- a/docs/tools/env-config.mdx +++ b/docs/tools/env-config.mdx @@ -3,8 +3,6 @@ title: env_config - 环境变量 description: 管理 API Key 等秘钥配置 --- -# env_config - 管理工作空间 `.env` 文件中的环境变量（API Key 等秘钥），支持通过对话安全地添加和更新。内置安全保护和脱敏策略。 ## 依赖 diff --git a/docs/tools/ls.mdx b/docs/tools/ls.mdx index fff91be1..e4d25fc5 100644 --- a/docs/tools/ls.mdx +++ b/docs/tools/ls.mdx @@ -3,8 +3,6 @@ title: ls - 目录列表 description: 列出目录内容 --- -# ls - 列出目录内容，按字母排序，目录名带 `/` 后缀，包含隐藏文件。 ## 依赖 diff --git a/docs/tools/memory.mdx b/docs/tools/memory.mdx index 4fa09360..0b60624a 100644 --- a/docs/tools/memory.mdx +++ b/docs/tools/memory.mdx @@ -3,8 +3,6 @@ title: memory - 记忆 description: 搜索和读取长期记忆 --- -# memory - 记忆工具包含两个子工具：`memory_search`（搜索记忆）和 `memory_get`（读取记忆文件）。 ## 依赖 diff --git a/docs/tools/read.mdx b/docs/tools/read.mdx index 04541251..07e08b88 100644 --- a/docs/tools/read.mdx +++ b/docs/tools/read.mdx @@ -3,8 +3,6 @@ title: read - 文件读取 description: 读取文件内容 --- -# read - 读取文件内容。支持文本文件、PDF 文件、图片（返回元数据）等格式。 ## 依赖 diff --git a/docs/tools/scheduler.mdx b/docs/tools/scheduler.mdx index 20ca4df1..91b1789e 100644 --- a/docs/tools/scheduler.mdx +++ b/docs/tools/scheduler.mdx @@ -3,8 +3,6 @@ title: scheduler - 定时任务 description: 创建和管理定时任务 --- -# scheduler - 创建和管理动态定时任务，支持灵活的调度方式和执行模式。 ## 依赖 diff --git a/docs/tools/send.mdx b/docs/tools/send.mdx index 4f30c617..05f73a6b 100644 --- a/docs/tools/send.mdx +++ b/docs/tools/send.mdx @@ -3,8 +3,6 @@ title: send - 文件发送 description: 向用户发送文件 --- -# send - 向用户发送文件（图片、视频、音频、文档等），当用户明确要求发送/分享文件时使用。 ## 依赖 diff --git a/docs/tools/web-search.mdx b/docs/tools/web-search.mdx index 0eda6d5a..2622d0c5 100644 --- a/docs/tools/web-search.mdx +++ b/docs/tools/web-search.mdx @@ -3,8 +3,6 @@ title: web_search - 联网搜索 description: 搜索互联网获取实时信息 --- -# web_search - 搜索互联网获取实时信息、新闻、研究等内容。支持两个搜索后端，自动选择可用的后端。 ## 依赖 diff --git a/docs/tools/write.mdx b/docs/tools/write.mdx index 63f77a5e..51cf66f1 100644 --- a/docs/tools/write.mdx +++ b/docs/tools/write.mdx @@ -3,8 +3,6 @@ title: write - 文件写入 description: 创建或覆盖写入文件 --- -# write - 写入内容到文件。文件不存在则自动创建，已存在则覆盖。自动创建父目录。 ## 依赖 From b788a3dd4e0821193f9cf6289316a642099e328e Mon Sep 17 00:00:00 2001 From: zhayujie Date: Sat, 28 Feb 2026 15:03:33 +0800 Subject: [PATCH 04/10] fix: incomplete historical session messages --- agent/protocol/agent_stream.py | 178 +++++++++++++++++++++++++++++---- bridge/agent_bridge.py | 30 +++++- docs/en/intro/architecture.mdx | 20 ++-- docs/intro/architecture.mdx | 20 ++-- 4 files changed, 210 insertions(+), 38 deletions(-) diff --git a/agent/protocol/agent_stream.py b/agent/protocol/agent_stream.py index 31c93592..117d6f97 100644 --- a/agent/protocol/agent_stream.py +++ b/agent/protocol/agent_stream.py @@ -636,11 +636,16 @@ class AgentStreamExecutor: ]) # Check if error is message format error (incomplete tool_use/tool_result pairs) - # This happens when previous conversation had tool failures + # This happens when previous conversation had tool failures or context trimming + # broke tool_use/tool_result pairs. is_message_format_error = any(keyword in error_str_lower for keyword in [ 'tool_use', 'tool_result', 'without', 'immediately after', - 'corresponding', 'must have', 'each' - ]) and 'status: 400' in error_str_lower + 'corresponding', 'must have', 'each', + 'tool_call_id', 'is not found', 'tool_calls', + 'must be a response to a preceeding message' + ]) and ('400' in error_str_lower or 'status: 400' in error_str_lower + or 'invalid_request' in error_str_lower + or 'invalidparameter' in error_str_lower) if is_context_overflow or is_message_format_error: error_type = "context overflow" if is_context_overflow else "message format error" @@ -659,9 +664,10 @@ class AgentStreamExecutor: ) # Aggressive trim didn't help or this is a message format error - # -> clear everything + # -> clear everything and also purge DB to prevent reload of dirty data logger.warning("🔄 Clearing conversation history to recover") self.messages.clear() + self._clear_session_db() if is_context_overflow: raise Exception( "抱歉，对话历史过长导致上下文溢出。我已清空历史记录，请重新描述你的需求。" @@ -907,23 +913,127 @@ class AgentStreamExecutor: def _validate_and_fix_messages(self): """ Validate message history and fix incomplete tool_use/tool_result pairs. - Claude API requires each tool_use to have a corresponding tool_result immediately after. + + All LLM APIs (OpenAI, Claude, Moonshot, DashScope) require: + 1. Each tool_use in an assistant message must have a matching tool_result + in the immediately following user message. + 2. Each tool_result in a user message must reference a tool_use_id that + exists in the preceding assistant message. + + This method performs a full scan and removes any messages that would + cause a 400 error due to broken tool_use/tool_result pairing. """ if not self.messages: return - - # Check last message for incomplete tool_use - if len(self.messages) > 0: + + # Pass 1: remove trailing incomplete tool_use (assistant with tool_use + # but no following tool_result) + while self.messages: last_msg = self.messages[-1] if last_msg.get("role") == "assistant": - # Check if assistant message has tool_use blocks content = last_msg.get("content", []) - if isinstance(content, list): - has_tool_use = any(block.get("type") == "tool_use" for block in content) - if has_tool_use: - # This is incomplete - remove it - logger.warning(f"⚠️ Removing incomplete tool_use message from history") - self.messages.pop() + if isinstance(content, list) and any( + isinstance(b, dict) and b.get("type") == "tool_use" + for b in content + ): + logger.warning("⚠️ Removing trailing incomplete tool_use assistant message") + self.messages.pop() + continue + break + + # Pass 2: full scan for orphaned tool_result and missing tool_result + removed = 0 + i = 0 + while i < len(self.messages): + msg = self.messages[i] + role = msg.get("role") + content = msg.get("content", []) + + if role == "assistant" and isinstance(content, list): + tool_use_ids = { + b.get("id") + for b in content + if isinstance(b, dict) and b.get("type") == "tool_use" and b.get("id") + } + if tool_use_ids: + # There must be a following user message with matching tool_results + next_idx = i + 1 + if next_idx >= len(self.messages): + # No following message at all — remove + logger.warning(f"⚠️ Removing assistant tool_use at index {i} (no following tool_result)") + self.messages.pop(i) + removed += 1 + continue + + next_msg = self.messages[next_idx] + next_content = next_msg.get("content", []) + if next_msg.get("role") != "user" or not isinstance(next_content, list): + # Next message is not a user message with tool_results + logger.warning(f"⚠️ Removing assistant tool_use at index {i} (next message is not tool_result)") + self.messages.pop(i) + removed += 1 + continue + + result_ids = { + b.get("tool_use_id") + for b in next_content + if isinstance(b, dict) and b.get("type") == "tool_result" + } + if not tool_use_ids.issubset(result_ids): + # Some tool_use ids have no matching result — remove both + logger.warning( + f"⚠️ Removing mismatched tool_use/result pair at index {i},{next_idx} " + f"(use_ids={tool_use_ids}, result_ids={result_ids})" + ) + self.messages.pop(next_idx) + self.messages.pop(i) + removed += 2 + continue + + elif role == "user" and isinstance(content, list): + has_tool_results = any( + isinstance(b, dict) and b.get("type") == "tool_result" + for b in content + ) + if has_tool_results: + # Check that the preceding message is an assistant with matching tool_use + if i == 0: + logger.warning(f"⚠️ Removing orphaned tool_result at index {i} (no preceding assistant)") + self.messages.pop(i) + removed += 1 + continue + + prev_msg = self.messages[i - 1] + prev_content = prev_msg.get("content", []) + if prev_msg.get("role") != "assistant" or not isinstance(prev_content, list): + logger.warning(f"⚠️ Removing orphaned tool_result at index {i} (prev is not assistant)") + self.messages.pop(i) + removed += 1 + continue + + prev_use_ids = { + b.get("id") + for b in prev_content + if isinstance(b, dict) and b.get("type") == "tool_use" and b.get("id") + } + result_ids = { + b.get("tool_use_id") + for b in content + if isinstance(b, dict) and b.get("type") == "tool_result" + } + if not result_ids.issubset(prev_use_ids): + logger.warning( + f"⚠️ Removing orphaned tool_result at index {i} " + f"(result_ids={result_ids} not in prev use_ids={prev_use_ids})" + ) + self.messages.pop(i) + removed += 1 + continue + + i += 1 + + if removed > 0: + logger.info(f"🔧 Message validation: removed {removed} broken message(s)") def _identify_complete_turns(self) -> List[Dict]: """ @@ -946,24 +1056,30 @@ class AgentStreamExecutor: content = msg.get('content', []) if role == 'user': - # 检查是否是用户查询（不是工具结果） + # Determine if this is a real user query (not a tool_result injection + # or an internal hint message injected by the agent loop). is_user_query = False + has_tool_result = False if isinstance(content, list): - is_user_query = any( - block.get('type') == 'text' - for block in content - if isinstance(block, dict) + has_text = any( + isinstance(block, dict) and block.get('type') == 'text' + for block in content ) + has_tool_result = any( + isinstance(block, dict) and block.get('type') == 'tool_result' + for block in content + ) + # A message with tool_result is always internal, even if it + # also contains text blocks (shouldn't happen, but be safe). + is_user_query = has_text and not has_tool_result elif isinstance(content, str): is_user_query = True if is_user_query: - # 开始新轮次 if current_turn['messages']: turns.append(current_turn) current_turn = {'messages': [msg]} else: - # 工具结果，属于当前轮次 current_turn['messages'].append(msg) else: # AI 回复，属于当前轮次 @@ -1252,6 +1368,24 @@ class AgentStreamExecutor: f"~{current_tokens + system_tokens} -> ~{accumulated_tokens + system_tokens} tokens)" ) + def _clear_session_db(self): + """ + Clear the current session's persisted messages from SQLite DB. + + This prevents dirty data (broken tool_use/tool_result pairs) from being + reloaded on the next request or after a restart. + """ + try: + session_id = getattr(self.agent, '_current_session_id', None) + if not session_id: + return + from agent.memory import get_conversation_store + store = get_conversation_store() + store.clear_session(session_id) + logger.info(f"🗑️ Cleared dirty session data from DB: {session_id}") + except Exception as e: + logger.warning(f"Failed to clear session DB: {e}") + def _prepare_messages(self) -> List[Dict[str, Any]]: """ Prepare messages to send to LLM diff --git a/bridge/agent_bridge.py b/bridge/agent_bridge.py index d7401ef3..a2338521 100644 --- a/bridge/agent_bridge.py +++ b/bridge/agent_bridge.py @@ -332,9 +332,10 @@ class AgentBridge: Returns: Reply object """ + session_id = None + agent = None try: # Extract session_id from context for user isolation - session_id = None if context: session_id = context.kwargs.get("session_id") or context.get("session_id") @@ -371,6 +372,9 @@ class AgentBridge: if context and hasattr(agent, 'model'): agent.model.channel_type = context.get("channel_type", "") + # Store session_id on agent so executor can clear DB on fatal errors + agent._current_session_id = session_id + # Record message count before execution so we can diff new messages with agent.messages_lock: pre_run_len = len(agent.messages) @@ -395,7 +399,17 @@ class AgentBridge: channel_type = (context.get("channel_type") or "") if context else "" with agent.messages_lock: new_messages = agent.messages[pre_run_len:] - self._persist_messages(session_id, list(new_messages), channel_type) + if new_messages: + self._persist_messages(session_id, list(new_messages), channel_type) + elif pre_run_len > 0 and len(agent.messages) == 0: + # Agent cleared its messages (recovery from format error / overflow) + # Also clear the DB to prevent reloading dirty data + try: + from agent.memory import get_conversation_store + get_conversation_store().clear_session(session_id) + logger.info(f"[AgentBridge] Cleared DB for recovered session: {session_id}") + except Exception as e: + logger.warning(f"[AgentBridge] Failed to clear DB after recovery: {e}") # Check if there are files to send (from read tool) if hasattr(agent, 'stream_executor') and hasattr(agent.stream_executor, 'files_to_send'): @@ -415,6 +429,18 @@ class AgentBridge: except Exception as e: logger.error(f"Agent reply error: {e}") + # If the agent cleared its messages due to format error / overflow, + # also purge the DB so the next request starts clean. + if session_id and agent: + try: + with agent.messages_lock: + msg_count = len(agent.messages) + if msg_count == 0: + from agent.memory import get_conversation_store + get_conversation_store().clear_session(session_id) + logger.info(f"[AgentBridge] Cleared DB for session after error: {session_id}") + except Exception as db_err: + logger.warning(f"[AgentBridge] Failed to clear DB after error: {db_err}") return Reply(ReplyType.ERROR, f"Agent error: {str(e)}") def _create_file_reply(self, file_info: dict, text_response: str, context: Context = None) -> Reply: diff --git a/docs/en/intro/architecture.mdx b/docs/en/intro/architecture.mdx index 3cdf619d..cb1c7d76 100644 --- a/docs/en/intro/architecture.mdx +++ b/docs/en/intro/architecture.mdx @@ -33,18 +33,24 @@ When Agent mode is enabled, CowAgent runs as an autonomous agent with the follow ## Workspace Directory Structure -The Agent workspace is located at `~/.cow` by default and stores system prompts, memory files, and skill files: +The Agent workspace is located at `~/cow` by default and stores system prompts, memory files, and skill files: ``` -~/.cow/ +~/cow/ ├── system.md # Agent system prompt ├── user.md # User profile ├── memory/ # Long-term memory storage │ ├── core.md # Core memory │ └── daily/ # Daily memory -├── skills/ # Custom skills -│ ├── skill-1/ -│ └── skill-2/ +└── skills/ # Custom skills + ├── skill-1/ + └── skill-2/ +``` + +Secret keys are stored separately in `~/.cow` directory for security: + +``` +~/.cow/ └── .env # Secret keys for skills ``` @@ -55,7 +61,7 @@ Configure Agent mode parameters in `config.json`: ```json { "agent": true, - "agent_workspace": "~/.cow", + "agent_workspace": "~/cow", "agent_max_context_tokens": 40000, "agent_max_context_turns": 30, "agent_max_steps": 15 @@ -65,7 +71,7 @@ Configure Agent mode parameters in `config.json`: | Parameter | Description | Default | | --- | --- | --- | | `agent` | Enable Agent mode | `true` | -| `agent_workspace` | Workspace path | `~/.cow` | +| `agent_workspace` | Workspace path | `~/cow` | | `agent_max_context_tokens` | Max context tokens | `40000` | | `agent_max_context_turns` | Max context turns | `30` | | `agent_max_steps` | Max decision steps per task | `15` | diff --git a/docs/intro/architecture.mdx b/docs/intro/architecture.mdx index 67aff375..deae6456 100644 --- a/docs/intro/architecture.mdx +++ b/docs/intro/architecture.mdx @@ -33,18 +33,24 @@ CowAgent 的整体架构由以下核心模块组成： ## 工作空间 -Agent 的工作空间默认位于 `~/.cow` 目录，用于存储系统提示词、记忆文件、技能文件等： +Agent 的工作空间默认位于 `~/cow` 目录，用于存储系统提示词、记忆文件、技能文件等： ``` -~/.cow/ +~/cow/ ├── system.md # Agent system prompt ├── user.md # User profile ├── memory/ # Long-term memory storage │ ├── core.md # Core memory │ └── daily/ # Daily memory -├── skills/ # Custom skills -│ ├── skill-1/ -│ └── skill-2/ +└── skills/ # Custom skills + ├── skill-1/ + └── skill-2/ +``` + +秘钥文件单独存储在 `~/.cow` 目录（出于安全考虑）： + +``` +~/.cow/ └── .env # Secret keys for skills ``` @@ -55,7 +61,7 @@ Agent 的工作空间默认位于 `~/.cow` 目录，用于存储系统提示词 ```json { "agent": true, - "agent_workspace": "~/.cow", + "agent_workspace": "~/cow", "agent_max_context_tokens": 40000, "agent_max_context_turns": 30, "agent_max_steps": 15 @@ -65,7 +71,7 @@ Agent 的工作空间默认位于 `~/.cow` 目录，用于存储系统提示词 | 参数 | 说明 | 默认值 | | --- | --- | --- | | `agent` | 是否启用 Agent 模式 | `true` | -| `agent_workspace` | 工作空间路径 | `~/.cow` | +| `agent_workspace` | 工作空间路径 | `~/cow` | | `agent_max_context_tokens` | 最大上下文 token 数 | `40000` | | `agent_max_context_turns` | 最大上下文记忆轮次 | `30` | | `agent_max_steps` | 单次任务最大决策步数 | `15` | From a33ce97ed9987041cb4ff9a06c78d1302b4e3c96 Mon Sep 17 00:00:00 2001 From: zhayujie Date: Sat, 28 Feb 2026 15:14:56 +0800 Subject: [PATCH 05/10] fix: restore only user/assistant text from history, strip tool calls Made-with: Cursor --- agent/protocol/agent_stream.py | 125 +++++++-------------------------- bridge/agent_initializer.py | 93 ++++++++++++++++++++---- 2 files changed, 106 insertions(+), 112 deletions(-) diff --git a/agent/protocol/agent_stream.py b/agent/protocol/agent_stream.py index 117d6f97..71ba3841 100644 --- a/agent/protocol/agent_stream.py +++ b/agent/protocol/agent_stream.py @@ -912,22 +912,21 @@ class AgentStreamExecutor: def _validate_and_fix_messages(self): """ - Validate message history and fix incomplete tool_use/tool_result pairs. + Validate message history and fix broken tool_use/tool_result pairs. - All LLM APIs (OpenAI, Claude, Moonshot, DashScope) require: - 1. Each tool_use in an assistant message must have a matching tool_result - in the immediately following user message. - 2. Each tool_result in a user message must reference a tool_use_id that - exists in the preceding assistant message. - - This method performs a full scan and removes any messages that would - cause a 400 error due to broken tool_use/tool_result pairing. + Historical messages restored from DB are text-only (no tool calls), + so this method only needs to handle edge cases in the current session: + - Trailing assistant message with tool_use but no following tool_result + (e.g. process was interrupted mid-execution) + - Orphaned tool_result at the start of messages (e.g. after context + trimming removed the preceding assistant tool_use) """ if not self.messages: return - # Pass 1: remove trailing incomplete tool_use (assistant with tool_use - # but no following tool_result) + removed = 0 + + # Remove trailing incomplete tool_use assistant messages while self.messages: last_msg = self.messages[-1] if last_msg.get("role") == "assistant": @@ -938,99 +937,27 @@ class AgentStreamExecutor: ): logger.warning("⚠️ Removing trailing incomplete tool_use assistant message") self.messages.pop() + removed += 1 continue break - # Pass 2: full scan for orphaned tool_result and missing tool_result - removed = 0 - i = 0 - while i < len(self.messages): - msg = self.messages[i] - role = msg.get("role") - content = msg.get("content", []) - - if role == "assistant" and isinstance(content, list): - tool_use_ids = { - b.get("id") - for b in content - if isinstance(b, dict) and b.get("type") == "tool_use" and b.get("id") - } - if tool_use_ids: - # There must be a following user message with matching tool_results - next_idx = i + 1 - if next_idx >= len(self.messages): - # No following message at all — remove - logger.warning(f"⚠️ Removing assistant tool_use at index {i} (no following tool_result)") - self.messages.pop(i) - removed += 1 - continue - - next_msg = self.messages[next_idx] - next_content = next_msg.get("content", []) - if next_msg.get("role") != "user" or not isinstance(next_content, list): - # Next message is not a user message with tool_results - logger.warning(f"⚠️ Removing assistant tool_use at index {i} (next message is not tool_result)") - self.messages.pop(i) - removed += 1 - continue - - result_ids = { - b.get("tool_use_id") - for b in next_content - if isinstance(b, dict) and b.get("type") == "tool_result" - } - if not tool_use_ids.issubset(result_ids): - # Some tool_use ids have no matching result — remove both - logger.warning( - f"⚠️ Removing mismatched tool_use/result pair at index {i},{next_idx} " - f"(use_ids={tool_use_ids}, result_ids={result_ids})" - ) - self.messages.pop(next_idx) - self.messages.pop(i) - removed += 2 - continue - - elif role == "user" and isinstance(content, list): - has_tool_results = any( + # Remove leading orphaned tool_result user messages + while self.messages: + first_msg = self.messages[0] + if first_msg.get("role") == "user": + content = first_msg.get("content", []) + if isinstance(content, list) and any( isinstance(b, dict) and b.get("type") == "tool_result" for b in content - ) - if has_tool_results: - # Check that the preceding message is an assistant with matching tool_use - if i == 0: - logger.warning(f"⚠️ Removing orphaned tool_result at index {i} (no preceding assistant)") - self.messages.pop(i) - removed += 1 - continue - - prev_msg = self.messages[i - 1] - prev_content = prev_msg.get("content", []) - if prev_msg.get("role") != "assistant" or not isinstance(prev_content, list): - logger.warning(f"⚠️ Removing orphaned tool_result at index {i} (prev is not assistant)") - self.messages.pop(i) - removed += 1 - continue - - prev_use_ids = { - b.get("id") - for b in prev_content - if isinstance(b, dict) and b.get("type") == "tool_use" and b.get("id") - } - result_ids = { - b.get("tool_use_id") - for b in content - if isinstance(b, dict) and b.get("type") == "tool_result" - } - if not result_ids.issubset(prev_use_ids): - logger.warning( - f"⚠️ Removing orphaned tool_result at index {i} " - f"(result_ids={result_ids} not in prev use_ids={prev_use_ids})" - ) - self.messages.pop(i) - removed += 1 - continue - - i += 1 + ) and not any( + isinstance(b, dict) and b.get("type") == "text" + for b in content + ): + logger.warning("⚠️ Removing leading orphaned tool_result user message") + self.messages.pop(0) + removed += 1 + continue + break if removed > 0: logger.info(f"🔧 Message validation: removed {removed} broken message(s)") diff --git a/bridge/agent_initializer.py b/bridge/agent_initializer.py index 62df5c85..f4c83296 100644 --- a/bridge/agent_initializer.py +++ b/bridge/agent_initializer.py @@ -130,8 +130,14 @@ class AgentInitializer: Load persisted conversation messages from SQLite and inject them into the agent's in-memory message list. - Only runs when conversation persistence is enabled (default: True). - Respects agent_max_context_turns to limit how many turns are loaded. + Only user text and assistant text are restored. Tool call chains + (tool_use / tool_result) are stripped out because: + 1. They are intermediate process, the value is already in the final + assistant text reply. + 2. They consume massive context tokens (often 80%+ of history). + 3. Different models have incompatible tool message formats, so + restoring tool chains across model switches causes 400 errors. + 4. Eliminates the entire class of tool_use/tool_result pairing bugs. """ from config import conf if not conf().get("conversation_persistence", True): @@ -140,25 +146,86 @@ class AgentInitializer: try: from agent.memory import get_conversation_store store = get_conversation_store() - # On restore, load at most min(10, max_turns // 2) turns so that - # a long-running session does not immediately fill the context window - # after a restart. The full max_turns budget is reserved for the - # live conversation that follows. max_turns = conf().get("agent_max_context_turns", 30) - restore_turns = max(4, max_turns // 5) + restore_turns = max(6, max_turns // 3) saved = store.load_messages(session_id, max_turns=restore_turns) if saved: - with agent.messages_lock: - agent.messages = saved - logger.debug( - f"[AgentInitializer] Restored {len(saved)} messages " - f"({restore_turns} turns cap) for session={session_id}" - ) + filtered = self._filter_text_only_messages(saved) + if filtered: + with agent.messages_lock: + agent.messages = filtered + logger.debug( + f"[AgentInitializer] Restored {len(filtered)} text messages " + f"(from {len(saved)} total, {restore_turns} turns cap) " + f"for session={session_id}" + ) except Exception as e: logger.warning( f"[AgentInitializer] Failed to restore conversation history for " f"session={session_id}: {e}" ) + + @staticmethod + def _filter_text_only_messages(messages: list) -> list: + """ + Filter messages to keep only user text and assistant text. + + Strips out: + - assistant messages that only contain tool_use (no text) + - user messages that only contain tool_result (no text) + - internal hint messages injected by the agent loop + + Keeps: + - user messages with actual text content + - assistant messages with text content (tool_use blocks removed, + only the text portion is kept) + """ + filtered = [] + for msg in messages: + role = msg.get("role") + content = msg.get("content") + + if isinstance(content, str): + if content.strip(): + filtered.append(msg) + continue + + if not isinstance(content, list): + continue + + if role == "user": + text_parts = [ + b.get("text", "") + for b in content + if isinstance(b, dict) and b.get("type") == "text" + ] + has_tool_result = any( + isinstance(b, dict) and b.get("type") == "tool_result" + for b in content + ) + if has_tool_result: + continue + text = "\n".join(p for p in text_parts if p).strip() + if text: + filtered.append({ + "role": "user", + "content": [{"type": "text", "text": text}] + }) + + elif role == "assistant": + text_parts = [ + b.get("text", "") + for b in content + if isinstance(b, dict) and b.get("type") == "text" + ] + text = "\n".join(p for p in text_parts if p).strip() + if text: + filtered.append({ + "role": "assistant", + "content": [{"type": "text", "text": text}] + }) + + return filtered def _load_env_file(self): """Load environment variables from .env file""" From e9c57ddf4decf8b97ea41217effcc29a40d9684b Mon Sep 17 00:00:00 2001 From: zhayujie Date: Sat, 28 Feb 2026 15:25:20 +0800 Subject: [PATCH 06/10] fix: adjust default turns --- agent/chat/service.py | 2 +- agent/protocol/agent.py | 2 +- bridge/agent_initializer.py | 4 ++-- channel/web/web_channel.py | 2 +- 4 files changed, 5 insertions(+), 5 deletions(-) diff --git a/agent/chat/service.py b/agent/chat/service.py index af50ce8b..b0d49fb0 100644 --- a/agent/chat/service.py +++ b/agent/chat/service.py @@ -111,7 +111,7 @@ class ChatService: logger.info(f"[ChatService] Starting agent run: session={session_id}, query={query[:80]}") from config import conf - max_context_turns = conf().get("agent_max_context_turns", 30) + max_context_turns = conf().get("agent_max_context_turns", 20) # Get full system prompt with skills full_system_prompt = agent.get_full_system_prompt() diff --git a/agent/protocol/agent.py b/agent/protocol/agent.py index 3f0167d3..3342eea9 100644 --- a/agent/protocol/agent.py +++ b/agent/protocol/agent.py @@ -480,7 +480,7 @@ class Agent: # Get max_context_turns from config from config import conf - max_context_turns = conf().get("agent_max_context_turns", 30) + max_context_turns = conf().get("agent_max_context_turns", 20) # Create stream executor with copied message history executor = AgentStreamExecutor( diff --git a/bridge/agent_initializer.py b/bridge/agent_initializer.py index f4c83296..44f28b92 100644 --- a/bridge/agent_initializer.py +++ b/bridge/agent_initializer.py @@ -146,8 +146,8 @@ class AgentInitializer: try: from agent.memory import get_conversation_store store = get_conversation_store() - max_turns = conf().get("agent_max_context_turns", 30) - restore_turns = max(6, max_turns // 3) + max_turns = conf().get("agent_max_context_turns", 20) + restore_turns = max(6, max_turns // 5) saved = store.load_messages(session_id, max_turns=restore_turns) if saved: filtered = self._filter_text_only_messages(saved) diff --git a/channel/web/web_channel.py b/channel/web/web_channel.py index e0ce3019..1f4aa077 100644 --- a/channel/web/web_channel.py +++ b/channel/web/web_channel.py @@ -524,7 +524,7 @@ class ConfigHandler: "model": local_config.get("model", ""), "channel_type": local_config.get("channel_type", ""), "agent_max_context_tokens": local_config.get("agent_max_context_tokens", 50000), - "agent_max_context_turns": local_config.get("agent_max_context_turns", 30), + "agent_max_context_turns": local_config.get("agent_max_context_turns", 20), "agent_max_steps": local_config.get("agent_max_steps", 15), "api_bases": api_bases, "api_keys": api_keys_masked, From 6ed85029c5209cb96ebfced2b5c841224e51700f Mon Sep 17 00:00:00 2001 From: zhayujie Date: Sat, 28 Feb 2026 16:46:49 +0800 Subject: [PATCH 07/10] fix: agent skills --- agent/tools/bash/bash.py | 10 ++ agent/tools/read/read.py | 4 +- docs/en/models/linkai.mdx | 4 +- docs/models/linkai.mdx | 4 +- skills/linkai-agent/README.md | 91 +++++----------- skills/linkai-agent/SKILL.md | 154 +++++++--------------------- skills/linkai-agent/scripts/call.sh | 138 ------------------------- 7 files changed, 77 insertions(+), 328 deletions(-) delete mode 100755 skills/linkai-agent/scripts/call.sh diff --git a/agent/tools/bash/bash.py b/agent/tools/bash/bash.py index 7b7ec2db..4c8430b4 100644 --- a/agent/tools/bash/bash.py +++ b/agent/tools/bash/bash.py @@ -100,6 +100,12 @@ SAFETY: else: logger.debug(f"[Bash] Process User: {os.environ.get('USERNAME', os.environ.get('USER', 'unknown'))}") + # On Windows, set console codepage to UTF-8 and prepend chcp for shell commands + if sys.platform == "win32": + env["PYTHONIOENCODING"] = "utf-8" + if command and not command.strip().lower().startswith("chcp"): + command = f"chcp 65001 >nul 2>&1 && {command}" + # Execute command with inherited environment variables result = subprocess.run( command, @@ -108,6 +114,8 @@ SAFETY: stdout=subprocess.PIPE, stderr=subprocess.PIPE, text=True, + encoding="utf-8", + errors="replace", timeout=timeout, env=env ) @@ -131,6 +139,8 @@ SAFETY: stdout=subprocess.PIPE, stderr=subprocess.PIPE, text=True, + encoding="utf-8", + errors="replace", timeout=timeout, env=env ) diff --git a/agent/tools/read/read.py b/agent/tools/read/read.py index cd1baa22..a6368ea6 100644 --- a/agent/tools/read/read.py +++ b/agent/tools/read/read.py @@ -240,8 +240,8 @@ class Read(BaseTool): "message": f"文件过大 ({format_size(file_size)} > 50MB)，无法读取内容。文件路径: {absolute_path}" }) - # Read file - with open(absolute_path, 'r', encoding='utf-8') as f: + # Read file (utf-8-sig strips BOM automatically on Windows) + with open(absolute_path, 'r', encoding='utf-8-sig') as f: content = f.read() # Truncate content if too long (20K characters max for model context) diff --git a/docs/en/models/linkai.mdx b/docs/en/models/linkai.mdx index 37ce8503..677f0051 100644 --- a/docs/en/models/linkai.mdx +++ b/docs/en/models/linkai.mdx @@ -8,8 +8,7 @@ The [LinkAI](https://link-ai.tech) platform lets you flexibly switch between Ope ```json { "use_linkai": true, - "linkai_api_key": "YOUR_API_KEY", - "linkai_app_code": "YOUR_APP_CODE" + "linkai_api_key": "YOUR_API_KEY" } ``` @@ -17,7 +16,6 @@ The [LinkAI](https://link-ai.tech) platform lets you flexibly switch between Ope | --- | --- | | `use_linkai` | Set to `true` to enable LinkAI interface | | `linkai_api_key` | Create at [LinkAI Console](https://link-ai.tech/console/interface) | -| `linkai_app_code` | Optional. Code of the LinkAI agent (app or workflow) | | `model` | Leave empty to use the agent's default model. Can be switched flexibly on the platform. All models in the [model list](https://link-ai.tech/console/models) are supported | See the [API documentation](https://docs.link-ai.tech/platform/api) for more details. diff --git a/docs/models/linkai.mdx b/docs/models/linkai.mdx index 1a84017f..4cf97780 100644 --- a/docs/models/linkai.mdx +++ b/docs/models/linkai.mdx @@ -8,8 +8,7 @@ description: 通过 LinkAI 平台统一接入多种模型 ```json { "use_linkai": true, - "linkai_api_key": "YOUR_API_KEY", - "linkai_app_code": "YOUR_APP_CODE" + "linkai_api_key": "YOUR_API_KEY" } ``` @@ -17,7 +16,6 @@ description: 通过 LinkAI 平台统一接入多种模型 | --- | --- | | `use_linkai` | 设为 `true` 启用 LinkAI 接口 | | `linkai_api_key` | 在 [控制台](https://link-ai.tech/console/interface) 创建 | -| `linkai_app_code` | LinkAI 智能体（应用或工作流）的 code，选填 | | `model` | 留空则使用智能体默认模型，可在平台中灵活切换，[模型列表](https://link-ai.tech/console/models) 中的全部模型均可使用 | 参考 [接口文档](https://docs.link-ai.tech/platform/api) 了解更多。 diff --git a/skills/linkai-agent/README.md b/skills/linkai-agent/README.md index d801ff6b..aa7bcb55 100644 --- a/skills/linkai-agent/README.md +++ b/skills/linkai-agent/README.md @@ -59,7 +59,7 @@ cp config.json.template config.json ### 3. 调用应用 ```bash -bash scripts/call.sh "G7z6vKwp" "What is artificial intelligence?" +bash(command='curl -sS --max-time 120 -X POST "https://api.link-ai.tech/v1/chat/completions" -H "Content-Type: application/json" -H "Authorization: Bearer $LINKAI_API_KEY" -d "{\"app_code\":\"G7z6vKwp\",\"messages\":[{\"role\":\"user\",\"content\":\"What is artificial intelligence?\"}],\"stream\":false}"', timeout=130) ``` ## 使用示例 @@ -67,46 +67,32 @@ bash scripts/call.sh "G7z6vKwp" "What is artificial intelligence?" ### 基础调用 ```bash -# 调用默认模型 -bash scripts/call.sh "G7z6vKwp" "解释一下量子计算" +# 调用默认模型 (通过 bash + curl) +bash(command='curl -sS --max-time 120 -X POST "https://api.link-ai.tech/v1/chat/completions" -H "Content-Type: application/json" -H "Authorization: Bearer $LINKAI_API_KEY" -d "{\"app_code\":\"G7z6vKwp\",\"messages\":[{\"role\":\"user\",\"content\":\"解释一下量子计算\"}],\"stream\":false}"', timeout=130) ``` ### 指定模型 -```bash -# 使用 GPT-4.1 模型 -bash scripts/call.sh "G7z6vKwp" "写一篇关于AI的文章" "LinkAI-4.1" +在 JSON body 中添加 `model` 字段： -# 使用 DeepSeek 模型 -bash scripts/call.sh "G7z6vKwp" "帮我写代码" "deepseek-chat" - -# 使用 Claude 模型 -bash scripts/call.sh "G7z6vKwp" "分析这段文本" "claude-4-sonnet" +```json +{ + "app_code": "G7z6vKwp", + "model": "LinkAI-4.1", + "messages": [{"role": "user", "content": "写一篇关于AI的文章"}], + "stream": false +} ``` ### 调用工作流 -```bash -# 工作流会按照配置的节点顺序执行 -bash scripts/call.sh "workflow_code" "输入数据或问题" -``` +工作流的 app_code 从 LinkAI 控制台获取，调用方式与普通应用相同。 ## ⚠️ 重要提示 ### 超时配置 -LinkAI 应用（特别是视频/图片生成、复杂工作流）可能需要较长时间处理。 - -**脚本内置超时**： -- 默认：120 秒（适合大多数场景） -- 可通过第 5 个参数自定义：`bash scripts/call.sh "" "false" "180"` - -**推荐超时时间**： -- **文本问答**：120 秒（默认） -- **图片生成**：120-180 秒 -- **视频生成**：180-300 秒 - -Agent 调用时会自动设置合适的超时时间。 +LinkAI 应用（特别是视频/图片生成、复杂工作流）可能需要较长时间处理。在 curl 命令中加入 `--max-time 180`，并相应增加 bash 工具的 `timeout` 参数。 ## 配置说明 @@ -125,38 +111,6 @@ Agent 调用时会自动设置合适的超时时间。 3. 选择要集成的应用/工作流 4. 在应用详情页找到 `app_code` -## 支持的模型 - -LinkAI 支持多种主流 AI 模型： - -**OpenAI 系列：** -- `LinkAI-4.1` - GPT-4.1 (1000K 上下文) -- `LinkAI-4.1-mini` - GPT-4.1 mini (1000K) -- `LinkAI-4.1-nano` - GPT-4.1 nano (1000K) -- `LinkAI-4o` - GPT-4o (128K) -- `LinkAI-4o-mini` - GPT-4o mini (128K) - -**DeepSeek 系列：** -- `deepseek-chat` - DeepSeek-V3 对话模型 (64K) -- `deepseek-reasoner` - DeepSeek-R1 推理模型 (64K) - -**Claude 系列：** -- `claude-4-sonnet` - Claude 4 Sonnet (200K) -- `claude-3-7-sonnet` - Claude 3.7 (200K) -- `claude-3-5-sonnet` - Claude 3.5 (200K) - -**Google 系列：** -- `gemini-2.5-pro` - Gemini 2.5 Pro (1000K) -- `gemini-2.0-flash` - Gemini 2.0 Flash (1000K) - -**国产模型：** -- `qwen3` - 通义千问3 (128K) -- `wenxin-4.5` - 文心一言4.5 (8K) -- `doubao-1.5-pro-256k` - 豆包1.5 (256K) -- `glm-4-plus` - 智谱GLM-4-Plus (4K) - -完整模型列表：https://link-ai.tech/console/models - ## 应用类型 ### 1. 普通应用 @@ -185,10 +139,16 @@ LinkAI 支持多种主流 AI 模型： ### 成功响应 +API 返回 OpenAI 兼容格式，从 `choices[0].message.content` 获取回复内容： + ```json { - "app_code": "G7z6vKwp", - "content": "人工智能（AI）是计算机科学的一个分支...", + "choices": [{ + "message": { + "role": "assistant", + "content": "人工智能（AI）是计算机科学的一个分支..." + } + }], "usage": { "prompt_tokens": 10, "completion_tokens": 150, @@ -201,9 +161,10 @@ LinkAI 支持多种主流 AI 模型： ```json { - "error": "LinkAI API error", - "message": "应用不存在", - "response": { ... } + "error": { + "message": "应用不存在", + "code": "xxx" + } } ``` @@ -259,7 +220,7 @@ Agent 看到所有可用应用的完整信息 ↓ Agent 根据描述选择合适的应用 ↓ -调用 call.sh +通过 bash + curl 调用 LinkAI API ↓ LinkAI API 处理并返回结果 ``` diff --git a/skills/linkai-agent/SKILL.md b/skills/linkai-agent/SKILL.md index 364d5cc6..dfd37fc3 100644 --- a/skills/linkai-agent/SKILL.md +++ b/skills/linkai-agent/SKILL.md @@ -1,6 +1,6 @@ --- name: linkai-agent -description: Call LinkAI applications and workflows. Use bash command to execute like 'bash /scripts/call.sh '. +description: Call LinkAI applications and workflows. Use bash with curl to invoke the chat completions API. homepage: https://link-ai.tech metadata: emoji: 🤖 @@ -10,110 +10,61 @@ metadata: primaryEnv: "LINKAI_API_KEY" --- -# LinkAI Agent Caller +# LinkAI Agent -Call LinkAI applications and workflows through API. Supports multiple apps/workflows configured in config.json. - -The available apps are dynamically loaded from `config.json` at skill loading time. +Call LinkAI applications and workflows through the chat completions API. Available apps are loaded from config.json. ## Setup -This skill requires a LinkAI API key. If not configured: +This skill requires a LinkAI API key. -1. Get your API key from https://link-ai.tech/console/api-keys -2. Set the key using: `env_config(action="set", key="LINKAI_API_KEY", value="your-key")` +1. Get your API key from [LinkAI Console](https://link-ai.tech/console/interface) +2. Set the environment variable: `export LINKAI_API_KEY=Link_xxxxxxxxxxxx` (or use env_config tool) ## Configuration 1. Copy `config.json.template` to `config.json` -2. Configure your apps/workflows: - -```json -{ - "apps": [ - { - "app_code": "your_app_code", - "app_name": "App Name", - "app_description": "What this app does" - } - ] -} -``` - -3. The skill description will be automatically updated when the agent loads this skill +2. Add your apps/workflows in config.json. The skill description is auto-generated from this config when loaded. ## Usage -**Important**: Scripts are located relative to this skill's base directory. - -When you see this skill in ``, note the `` path. - -**CRITICAL**: Always use `bash` command to execute the script: +Use the bash tool with curl to call the API. **Prefer curl** to avoid encoding issues on Windows PowerShell. ```bash -# General pattern (MUST start with bash): -bash "/scripts/call.sh" "" "" [model] [stream] [timeout] - -# DO NOT execute the script directly like this (WRONG): -# "/scripts/call.sh" ... - -# Parameters: -# - app_code: LinkAI app or workflow code (required) -# - question: User question (required) -# - model: Override model (optional, uses app default if not specified) -# - stream: Enable streaming (true/false, default: false) -# - timeout: curl timeout in seconds (default: 120, recommended for video/image generation) +curl -X POST "https://api.link-ai.tech/v1/chat/completions" \ + -H "Content-Type: application/json" \ + -H "Authorization: Bearer $LINKAI_API_KEY" \ + -d '{ + "app_code": "", + "messages": [{"role": "user", "content": ""}], + "stream": false + }' ``` -**IMPORTANT - Timeout Configuration**: -- The script has a **default timeout of 120 seconds** (suitable for most cases) -- For complex tasks (video generation, large workflows), pass a longer timeout as the 5th parameter -- The bash tool also needs sufficient timeout - set its `timeout` parameter accordingly -- Example: `bash(command="bash