diff --git a/app/server/presets.json b/app/server/presets.json index 47d419d..f025ee5 100644 --- a/app/server/presets.json +++ b/app/server/presets.json @@ -95,7 +95,7 @@ { "key": "video_task", "label": "视频生成", - "desc": "WorkBuddy 网关(minimax / kling)", + "desc": "文生视频 / 图生视频", "url": "http://mcp_video_task:3000/mcp", "mode": "skill", "fields": [ @@ -105,12 +105,20 @@ "secret": true, "prefix": "Bearer ", "required": true, - "placeholder": "wb.ai 的 API Key" + "placeholder": "API Key" + }, + { + "header": "X-Video-Task-Base-Url", + "label": "地址", + "required": true, + "placeholder": "https:///v1", + "hint": "OpenAI 兼容 videos / generations" }, { "header": "X-Video-Task-Model", "label": "模型", - "default": "cn:minimax-video-h3" + "required": true, + "placeholder": "上游的模型 id" } ] }, diff --git a/mcp/README.md b/mcp/README.md index 2d73195..efa01f1 100644 --- a/mcp/README.md +++ b/mcp/README.md @@ -6,7 +6,7 @@ |---|---|---|---|---| | `web_search/` | `mcp_web_search` | 网页搜索(默认免费 Parallel / Exa,可换成自建 `/search` 上游) | `3000` | `POST /` | | `image_task/` | `mcp_image_task` | 图片生成 / 编辑(文生图走 `/images/generations`,图生图走 `/images/edits`,单工具 op=create/get) | `3000` | `POST /mcp` | -| `video_task/` | `mcp_video_task` | 视频生成(上游 WorkBuddy 网关 `/v1/videos`,异步,单工具 op=create/get) | `3000` | `POST /mcp` | +| `video_task/` | `mcp_video_task` | 视频生成(任意 OpenAI 兼容 `/videos` 上游,异步,单工具 op=create/get) | `3000` | `POST /mcp` | | `asr/` | `mcp_asr` | 语音转文字(上游 OpenAI 兼容 `/audio/transcriptions`),**额外带一个给输入框用的 `POST /transcribe`** | `3000` | `POST /mcp` | **这些服务一个宿主机端口都不占**——它们只挂在 `agent-mcp-net` 上,由 `agent-core` / `agent-executor` @@ -50,7 +50,9 @@ mcp_servers: > 从旧的「agent + mcp 双仓库」部署升级:先在旧 mcp 仓库目录 `docker compose down`(只需一次,否则同名容器与网络会冲突),再 `up` 本栈。 `image_task` / `video_task` 的 token **不由服务端保存**——每个请求从 `Authorization` header 读取 -(agent 侧配在 `config.yml` 的 `mcp_servers[].headers`)。`video_task` 用 wb.ai 网关的 API Key; +(agent 侧配在 `config.yml` 的 `mcp_servers[].headers`)。`video_task` 走任意 OpenAI 兼容的 videos 上游: +提交 `/videos/generations`、查询 `/videos/{task_id}`,地址与模型由 `X-Video-Task-Base-Url` / +`X-Video-Task-Model` 指定(两者都必填); `image_task` 走任意 OpenAI 兼容的 images 上游:文生图 `/images/generations`、图生图 `/images/edits`, 上游地址与模型分别由 `X-Image-Task-Base-Url` / `X-Image-Task-Model` 指定(两者都必填)。 所以这两个服务不用配任何密钥。 diff --git a/mcp/video_task/README.md b/mcp/video_task/README.md index 9668f40..feadb5a 100644 --- a/mcp/video_task/README.md +++ b/mcp/video_task/README.md @@ -1,13 +1,15 @@ # MCP Video Task -对接 **WorkBuddy 网关**(`wb.ai`)的 OpenAI 兼容视频接口。只有一个工具 `video_task`,用 `op` 区分提交与查询。 -Server **不内部轮询**——AI 拿到 `task_id` 后自己决定什么时候复查。 +对接**任意 OpenAI 兼容的视频生成端点**:提交走 `POST {base}/videos/generations`,查询走 +`GET {base}/videos/{task_id}`。对外只有一个工具 `video_task`,用 `op` 区分动作: -- 协议:MCP **Streamable HTTP**(单端点 `/mcp`,支持 SSE 流) -- 上游:`POST /v1/videos/generations`(提交,`wait=false` 立刻回 `task_id`)、`GET /v1/videos/{task_id}`(查询) -- 默认模型:`cn:minimax-video-h3`;可选 `cn:kling-video-v3`、`global:kling-video-v3` -- 端口:容器内 `3000`(MCP `/mcp` + `/health`,容器外不映射,只在 `agent-mcp-net` 内被 agent 访问) -- 鉴权:**无服务端密钥**,每次请求必须从 header 携带网关 API Key +- `op=create` 提交任务,立刻回 `task_id` +- `op=get` 按 `task_id` 单次查询,**Server 不内部轮询** + +- 协议:MCP **Streamable HTTP**(单端点 `/mcp`) +- 端口:容器内 `3000`(`/mcp` + `/health`) +- 鉴权:**服务端不存密钥**,每个请求从 header 读 +- 上游:**不预设任何厂商**,Base-Url / 模型 / 凭据全部由调用方带入 --- @@ -18,133 +20,162 @@ Server **不内部轮询**——AI 拿到 `task_id` 后自己决定什么时候 仓库根目录 `docker-compose.yml` 已包含本服务: ```bash +cd /root/workspace/tools/agent docker compose up -d --build mcp_video_task ``` +容器名 `mcp_video_task`,不占宿主机端口,由 agent 走 `agent-mcp-net` 按容器名访问。 + ```bash -# 容器内健康检查(无需占端口) +# 容器内健康检查 docker exec mcp_video_task node -e "fetch('http://localhost:3000/health').then(r=>r.text()).then(console.log)" -# {"status":"ok","service":"video_task","upstream":"https://wb.ai.foresh.com","sessions":0} +# {"status":"ok","service":"video_task","sessions":0} ``` ### MCP 客户端配置 -```yaml -mcp_servers: - - name: video_task - url: "http://mcp_video_task:3000/mcp" - mode: skill - headers: - Authorization: Bearer - X-Video-Task-Model: cn:minimax-video-h3 # 可选,覆盖默认模型 +```jsonc +{ + "mcpServers": { + "video_task": { + "type": "http", + "url": "http://mcp_video_task:3000/mcp", + "headers": { + "Authorization": "Bearer <上游 API Key>", + "X-Video-Task-Base-Url": "https:///v1", + "X-Video-Task-Model": "<上游模型 id>" + } + } + } +} ``` --- -## 工具:`video_task` +## 三个 header + +| header | 必填 | 说明 | +|---|---|---| +| `Authorization` | 是 | `Bearer `,原样转给上游 | +| `X-Video-Task-Base-Url` | 是 | 完整 http(s) 地址,填到 `/v1` 即可;直接填到 `/videos/generations` 或 `/videos` 也会被剥回 base | +| `X-Video-Task-Model` | 是 | 上游模型 id,原样透传;不猜缺省值 | + +配置缺失或非法时,工具直接回 `isError` 并说明缺哪个 header,不会去打上游。 + +--- + +## 工具:video_task ### 参数(AI 可见) | 参数 | 类型 | 说明 | |---|---|---| -| `op` | `create` \| `get` | 提交 / 查询,必填 | -| `prompt` | string ≤ 2000 | `create` 必填。图生视频时描述画面如何运动 | -| `image_url` | url | 图生视频的参考图,公网可访问的 png/jpeg/jpg | -| `seconds` | 4–15 | 时长,默认 4。越长越慢越贵 | -| `aspect_ratio` | string | `16:9` / `9:16` / `1:1`。上游不严格校验 | -| `resolution` | string | `768P` / `1080P`,留空用上游默认 | +| `op` | `create` \| `get` | 必填 | +| `prompt` | string | 视频描述,`create` 时与 `image_url` 至少给一个 | +| `image_url` | string | 参考图(图生视频),公开可访问的 png/jpeg URL | +| `seconds` | int | 时长(秒),留空用上游默认 | +| `aspect_ratio` | string | 画幅,如 `16:9` / `9:16` / `1:1` | +| `resolution` | string | 分辨率,如 `768P` / `1080P` | | `negative_prompt` | string | 不希望出现的画面 | -| `enable_audio` | boolean | 是否生成音频,默认 false(别名 `with_audio`) | -| `task_id` | string | `get` 必填,`create` 返回 | +| `enable_audio` | bool | 是否生成音频,默认 false(旧名 `with_audio` 同样接受) | +| `task_id` | string | `create` 返回的 id,`get` 时必填 | -### 返回(`op=create`) +### 返回(op=create) ``` -task_id: 1312831544-minimax_video_h3-5c12… -model: cn:minimax-video-h3 +task_id: +model: <上游模型 id> status: queued -提示:生成是异步的,用 op="get" + task_id 查结果。 + +Video generation is asynchronous. Call this tool again with op="get" and task_id="" to fetch the result. ``` -### 返回(`op=get`) +### 返回(op=get) + +任务完成: ``` -task_id: … -model: cn:minimax-video-h3 +task_id: +model: <上游模型 id> status: completed -Video URL: https://…mp4?q-signature=… +Video URL: https://... credit: 62.29 - -注意:视频是签名直链,约 12 小时失效;要长期保存请立刻下载到本地。 +note: <上游附带的提示,如直链有效期> ``` -`status` 归一化为 `queued` / `in_progress` / `completed` / `failed`。 -`failed` 时返回 `isError: true` 并把上游原话带出来。 +未完成 / 失败:回 `status` 与 `progress`,失败时附上游 `error` 原文,并置 `isError=true`。 --- -## 模型与计费(2026-11-17 实测) +## 上游要求 -| 模型 | 请求 | 实际扣费 | 出片耗时 | 分辨率 | -|---|---|---|---|---| -| `cn:minimax-video-h3` | 4s / 16:9 | **62.29** | **128 秒** | 768P | -| `cn:kling-video-v3` | 4s(实际出 5s) | 102.86 | ≈473 秒 | — | +只要满足这两条,任何实现都能接(自建网关、云厂商、本地服务): -- 扣费发生在**出片那一刻**,提交不扣;重复查询不重复计费 -- 上游 `usage` 只回一次,之后重复查询不带 `credit` 属正常 -- `kling` 慢一个量级,客户端超时要给到 15 分钟以上 +1. `POST {base}/videos/generations` 返回 `task_id` 或 `id` +2. `GET {base}/videos/{task_id}` 返回 `status`;完成时带 `videos[].url` 或 `data[].url` -## 本地文件 → 公网 URL - -本服务不搬字节:图生视频要的是**公网可访问的 URL**。本地文件先由 agent 内置的 -「公开临时 URL」工具(`mint_public_link`)换一个短命链接,再把该链接作为 `image_url` 传进来。 +`status` 按小写比较:`completed` / `succeeded` / `success` 视为完成,`failed` / `error` / +`cancelled` 视为失败,其余按进行中处理。 --- -## 环境变量 +## 本地文件 → 公网 URL(本服务不搬字节) -| 变量 | 默认值 | 说明 | -|---|---|---| -| `WB_BASE_URL` | `https://wb.ai.foresh.com` | 网关地址,末尾斜杠自动去掉 | +图生视频要的是**公网可访问的 URL**,不是字节流。本地文件这样走: -## curl 调试 +``` +AI ──调用 agent 内置的「公开临时 URL」工具(直接传文件路径)──► 拿到一条短命公网 URL + │ + ▼ +AI 把 URL 当 image_url 交给 video_task:3000/mcp(转成上游的 image 字段) + │ + ▼ +上游 ──匿名 GET──► agent 侧提供的公开临时 URL(文件始终躺在任务目录里,零拷贝) +``` + +链接短命:到期或 agent 重启后失效;上游只在提交任务那一刻拉一次图,够用。 + +--- + +## 用 curl 手工调试 ```bash -# 1. initialize 拿 session id -curl -s -D - -X POST http://127.0.0.1:3000/mcp \ - -H 'Content-Type: application/json' -H 'Accept: application/json, text/event-stream' \ - -H "Authorization: Bearer $WB_KEY" \ - -d '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2025-06-18","capabilities":{},"clientInfo":{"name":"c","version":"0"}}}' -# 响应头里的 mcp-session-id 后续要带上 +URL=http://127.0.0.1:3000/mcp # 容器内;宿主机调试需自行映射端口 +H=(-H 'Content-Type: application/json' -H 'Accept: application/json, text/event-stream' + -H 'Authorization: Bearer ' + -H 'X-Video-Task-Base-Url: https:///v1' + -H 'X-Video-Task-Model: ') -# 2. 提交 -curl -s -X POST http://127.0.0.1:3000/mcp \ - -H 'Content-Type: application/json' -H 'Accept: application/json, text/event-stream' \ - -H "mcp-session-id: $SID" -H "Authorization: Bearer $WB_KEY" \ - -d '{"jsonrpc":"2.0","id":2,"method":"tools/call","params":{"name":"video_task","arguments":{"op":"create","prompt":"一只橘猫在草地上奔跑","seconds":4}}}' +# 1. initialize,从响应头取 mcp-session-id +curl -s -D - "${H[@]}" -X POST $URL \ + -d '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2025-06-18","capabilities":{},"clientInfo":{"name":"c","version":"0"}}}' + +# 2. 提交任务 +curl -s "${H[@]}" -H 'mcp-session-id: ' -X POST $URL \ + -d '{"jsonrpc":"2.0","id":2,"method":"tools/call","params":{"name":"video_task","arguments":{"op":"create","prompt":"一只猫在草地上跑"}}}' # 3. 查结果 -curl -s -X POST http://127.0.0.1:3000/mcp \ - -H 'Content-Type: application/json' -H 'Accept: application/json, text/event-stream' \ - -H "mcp-session-id: $SID" -H "Authorization: Bearer $WB_KEY" \ - -d '{"jsonrpc":"2.0","id":3,"method":"tools/call","params":{"name":"video_task","arguments":{"op":"get","task_id":""}}}' +curl -s "${H[@]}" -H 'mcp-session-id: ' -X POST $URL \ + -d '{"jsonrpc":"2.0","id":3,"method":"tools/call","params":{"name":"video_task","arguments":{"op":"get","task_id":""}}}' ``` -响应是 SSE 帧(`event: message` + `data: {...}`),解析时取 `data:` 那一行。 +响应是 SSE(`event: message` + `data: {...}`),取 `data:` 那行解析。 --- ## 架构与文件 -| 文件 | 作用 | -|---|---| -| `src/mcp-server.js` | 全部逻辑:Express + MCP SDK,`buildServer()` 注册工具,`startHttp()` 起 Streamable HTTP | -| `Dockerfile` | node:22-alpine,`node src/mcp-server.js --http --port 3000` | -| `package.json` | `@modelcontextprotocol/sdk` + `express` + `zod` | +``` +mcp/video_task/ +├── Dockerfile +├── package.json +├── README.md +└── src/mcp-server.js # 全部逻辑:header 解析 / 上游调用 / MCP 服务 +``` ### 关键设计 -- **服务端不轮询**:`create` 提交完立刻返回,`get` 只拉一次,由 AI 控制节奏 -- **状态归一化**:上游 `queued` / `in_progress` / `completed` / `failed` 统一成小写,AI 不用猜 -- **错误透传**:上游 4xx/5xx 的原话进 `content[0].text`,模型能读到「模型不存在」这类可行动信息 -- **鉴权随请求走**:服务端不落任何密钥,多实例共用一份镜像也不串号 +- **无状态**:不存密钥、不存任务,重启即忘。任务状态由上游持有,`task_id` 就是全部上下文。 +- **不轮询**:`op=get` 只拉一次。轮询节奏由 AI 决定,避免服务端挂着长任务占资源。 +- **配置错误不打上游**:缺 header 时同步返回 `isError`,省一次无意义的 400。 +- **`wait=false` 提交**:立刻拿 `task_id`,不让工具调用被上游排队拖死。 diff --git a/mcp/video_task/src/mcp-server.js b/mcp/video_task/src/mcp-server.js index 135897d..4777732 100644 --- a/mcp/video_task/src/mcp-server.js +++ b/mcp/video_task/src/mcp-server.js @@ -1,8 +1,4 @@ #!/usr/bin/env node -// MCP 视频任务服务:对接 WorkBuddy 网关(wb.ai)的 OpenAI 兼容视频接口 -// 提交 POST /v1/videos/generations(wait=false,立刻回 task_id) -// 查询 GET /v1/videos/{task_id} -// 只有一条工具 video_task,用 op 区分 create / get;服务端不轮询,AI 自己查。 import express from "express"; import { randomUUID } from "node:crypto"; import { McpServer } from "@modelcontextprotocol/sdk/server/mcp.js"; @@ -10,22 +6,13 @@ import { StreamableHTTPServerTransport } from "@modelcontextprotocol/sdk/server/ import { isInitializeRequest } from "@modelcontextprotocol/sdk/types.js"; import { z } from "zod"; -// 上游网关地址:容器内可用 WB_BASE_URL 覆盖 -const WB_BASE = (process.env.WB_BASE_URL || "https://wb.ai.foresh.com").replace(/\/+$/, ""); -const SUBMIT_URL = `${WB_BASE}/v1/videos/generations`; -const queryUrl = (id) => `${WB_BASE}/v1/videos/${encodeURIComponent(id)}`; +// 上游必须是 OpenAI 兼容的视频端点:提交 POST /videos/generations,查询 GET /videos/{task_id}。 +// Base-Url / 模型 / 凭据全部由调用方随每个请求用 header 带入(X-Video-Task-Base-Url / +// X-Video-Task-Model / Authorization),服务自己不存、也不预设任何厂商。 -// 默认模型:minimax 出片约 2 分钟、扣 57 积分;kling 约 8 分钟、扣 103 积分 -const DEFAULT_MODEL = "cn:minimax-video-h3"; -const DEFAULT_SECONDS = 4; - -// 提交/查询都是快接口(wait=false 立刻返回),卡住基本等于上游挂了 +// 提交与查询都是快接口(wait=false 立刻回 task_id),卡住基本等于上游挂了。 const UPSTREAM_TIMEOUT_MS = 60_000; -// 上游状态归一:命中终态才停轮询 -const DONE = new Set(["completed", "succeeded", "success"]); -const FAILED = new Set(["failed", "error", "cancelled", "canceled"]); - // ── 优雅退出 ────────────────────────────────────────────────────────────── // 容器里 node 以 PID 1 运行且未注册 SIGTERM 处理器时,内核不会对其执行默认 // 终止动作,docker stop/restart 会白等满 10s 宽限期才被 SIGKILL。 @@ -38,7 +25,7 @@ function shutdown(sig) { shuttingDown = true; console.log(`[video] received ${sig}, shutting down`); closeAllSessions(); - if (!httpServer) process.exit(0); + if (!httpServer) process.exit(0); // stdio 模式无监听,直接退 httpServer.close(() => process.exit(0)); setTimeout(() => process.exit(0), 3000).unref(); } @@ -56,65 +43,94 @@ function parseArgs(argv) { return args; } -// 鉴权:每个请求从 header 取 wb.ai 的 API Key,服务端不落盘 -function pickToken(headers) { +// 对外只有三个 header:Authorization / X-Video-Task-Base-Url / X-Video-Task-Model。 +function pickAuthToken(headers) { if (!headers) return null; - const candidates = [ - headers["x-video-task-token"], - headers["x-mcp-token"], - headers["authorization"], - ]; - for (const v of candidates) { - if (!v) continue; - const s = String(v).trim(); - if (!s) continue; - if (/^Bearer\s+/i.test(s)) { - const t = s.replace(/^Bearer\s+/i, "").trim(); - if (t) return t; - } else { - return s; - } + const s = String(headers["authorization"] || "").trim(); + if (!s) return null; + if (/^Bearer\s+/i.test(s)) { + const t = s.replace(/^Bearer\s+/i, "").trim(); + return t || null; } - return null; + return s; } -function pickModel(headers) { +function pickHeaderString(headers, name) { if (!headers) return null; - const v = headers["x-video-task-model"] || headers["x-mcp-model"] || headers["x-model"]; - if (!v) return null; + const v = headers[name]; + if (v === undefined || v === null) return null; const s = String(v).trim(); return s || null; } -// 网关的模型名一定带域前缀(cn: / global:)。旧配置里遗留的智谱模型名 -// (cogvideox-flash 之类)没有冒号,发上去会 400「route config not found」, -// 统一回落到默认模型,老配置不用改也能用。 -function normalizeModel(m) { - if (!m) return DEFAULT_MODEL; - return m.includes(":") ? m : DEFAULT_MODEL; +function pickBaseUrl(headers) { + return pickHeaderString(headers, "x-video-task-base-url"); } -async function callWb({ url, token, method = "POST", body }) { +function pickModel(headers) { + return pickHeaderString(headers, "x-video-task-model"); +} + +// 解析 Base-Url:必须是完整 http(s) URL。允许调用方直接填到 /videos/generations +// 或 /videos(照抄文档的写法),统一剥回 base——否则查询会被拼成两条路径。 +function resolveBaseUrl(rawBase) { + const base = String(rawBase || "").trim().replace(/\/+$/, ""); + if (!base) { + throw new Error( + "Missing X-Video-Task-Base-Url: pass the upstream base URL (e.g. https:///v1)." + ); + } + if (!/^https?:\/\//i.test(base)) { + throw new Error( + `Invalid X-Video-Task-Base-Url "${base}": must be a full http(s) URL.` + ); + } + return base.replace(/\/videos(\/generations)?$/i, ""); +} + +// 模型只由 header 决定,AI 看不到也填不了。不猜缺省值:猜错只会换来含义不明的 400。 +function resolveModel(headers) { + const model = pickModel(headers); + if (!model) { + throw new Error("Missing X-Video-Task-Model: pass the upstream model name."); + } + return model; +} + +// 一次解析出本次请求的上游、凭据、模型。配置问题同步抛错,直接回给调用方。 +function resolveTarget(headers) { + const token = pickAuthToken(headers); + if (!token) { + throw new Error('Missing Authorization: use "Authorization: Bearer ".'); + } + return { + base: resolveBaseUrl(pickBaseUrl(headers)), + token, + model: resolveModel(headers), + }; +} + +// 统一的上游请求:超时、错误体解析都在这层,提交与查询共用。 +async function callUpstream({ url, token, method = "POST", body }) { const init = { method, - headers: { - Authorization: `Bearer ${token}`, - "Content-Type": "application/json", - }, + headers: { Authorization: `Bearer ${token}` }, signal: AbortSignal.timeout(UPSTREAM_TIMEOUT_MS), }; - if (body !== undefined) init.body = JSON.stringify(body); - + if (body !== undefined) { + init.headers["Content-Type"] = "application/json"; + init.body = JSON.stringify(body); + } let resp; try { resp = await fetch(url, init); } catch (e) { + // AbortSignal.timeout 中止时抛 TimeoutError/AbortError,转成明确的超时错误 if (e?.name === "TimeoutError" || e?.name === "AbortError") { - throw new Error(`上游 ${UPSTREAM_TIMEOUT_MS} ms 内没有响应`); + throw new Error(`Upstream request timed out after ${UPSTREAM_TIMEOUT_MS} ms`); } throw e; } - const text = await resp.text(); let data; try { @@ -123,9 +139,7 @@ async function callWb({ url, token, method = "POST", body }) { data = { raw: text }; } if (!resp.ok) { - // 网关的错误体是 {error:{message}},取原话给模型看,便于自行纠错 - const msg = data?.error?.message || data?.error?.code || text.slice(0, 300); - const err = new Error(`网关返回 ${resp.status}: ${msg}`); + const err = new Error(`Upstream error: ${resp.status} ${resp.statusText} - ${text}`); err.status = resp.status; err.detail = data; throw err; @@ -133,62 +147,74 @@ async function callWb({ url, token, method = "POST", body }) { return data; } -// ===== op=create:提交任务,wait=false 立刻返回 task_id ===== -async function submitVideo({ token, model, prompt, imageUrl, seconds, aspectRatio, resolution, negativePrompt, enableAudio }) { - const body = { model, prompt, wait: false }; - if (imageUrl) body.image = imageUrl; - if (seconds) body.seconds = seconds; - if (aspectRatio) body.aspect_ratio = aspectRatio; +// ===== op=create:提交视频生成(wait=false,立刻返回 task_id)===== +async function submitVideo({ base, token, model, prompt, image_url, seconds, aspect_ratio, resolution, negative_prompt, enable_audio }) { + const body = { model, wait: false }; + if (prompt) body.prompt = prompt; + if (image_url) body.image = image_url; // OpenAI Videos 口径的参考图字段 + if (seconds !== undefined) body.seconds = seconds; + if (aspect_ratio) body.aspect_ratio = aspect_ratio; if (resolution) body.resolution = resolution; - if (negativePrompt) body.negative_prompt = negativePrompt; - if (enableAudio !== undefined && enableAudio !== null) body.enable_audio = Boolean(enableAudio); - return await callWb({ url: SUBMIT_URL, token, body }); + if (negative_prompt) body.negative_prompt = negative_prompt; + if (enable_audio !== undefined) body.enable_audio = Boolean(enable_audio); + + return await callUpstream({ url: base + "/videos/generations", token, body }); } -// ===== op=get:查任务(单次拉取) ===== -async function queryVideo({ token, taskId }) { - return await callWb({ url: queryUrl(taskId), token, method: "GET" }); +// ===== op=get:查询任务(单次拉取,不内部轮询)===== +async function queryVideo({ base, token, taskId }) { + return await callUpstream({ + url: base + "/videos/" + encodeURIComponent(taskId), + token, + method: "GET", + }); } -function normStatus(s) { - return String(s || "").trim().toLowerCase(); +const DONE = new Set(["completed", "succeeded", "success"]); +const FAILED = new Set(["failed", "error", "cancelled", "canceled"]); + +function normalizeStatus(s) { + return String(s || "").toLowerCase(); } -async function handleCreate({ args, headers, token }) { +// ---------- op=create 的业务处理 ---------- +async function handleCreate({ args, headers }) { + // 纯空白 prompt 等同没给:schema 只校验类型/长度,不校验内容。 const prompt = typeof args.prompt === "string" ? args.prompt.trim() : ""; - if (!prompt) { + if (!prompt && !args.image_url) { return { isError: true, - content: [{ type: "text", text: 'op=create 需要 "prompt"(视频描述,必填)。' }], + content: [{ type: "text", text: 'op=create requires at least one of "prompt" or "image_url".' }], }; } - const model = normalizeModel(pickModel(headers)); - const seconds = Number.isFinite(args.seconds) ? args.seconds : DEFAULT_SECONDS; + let target; + try { + target = resolveTarget(headers); + } catch (e) { + return { isError: true, content: [{ type: "text", text: e.message }] }; + } try { const data = await submitVideo({ - token, - model, + ...target, prompt, - imageUrl: args.image_url, - seconds, - aspectRatio: args.aspect_ratio, + image_url: args.image_url, + seconds: args.seconds, + aspect_ratio: args.aspect_ratio, resolution: args.resolution, - negativePrompt: args.negative_prompt, - // 兼容旧字段名 - enableAudio: args.enable_audio ?? args.with_audio, + negative_prompt: args.negative_prompt, + enable_audio: args.enable_audio ?? args.with_audio, }); const taskId = data.task_id || data.id; const summary = - `model: ${data.model || model}\n` + `task_id: ${taskId}\n` + - `status: ${normStatus(data.status) || "queued"}\n` + + `model: ${target.model}\n` + + `status: ${data.status || "queued"}\n` + `\n` + - `视频是异步生成(minimax 约 2 分钟,kling 约 8 分钟)。` + - `之后用 op="get" 加 task_id="${taskId}" 查询,直到 status 变成 completed 或 failed。` + - `不要重复提交同一个任务。`; + `Video generation is asynchronous. ` + + `Call this tool again with op="get" and task_id="${taskId}" to fetch the result.`; return { content: [{ type: "text", text: summary }] }; } catch (e) { @@ -197,177 +223,131 @@ async function handleCreate({ args, headers, token }) { content: [ { type: "text", - text: `提交视频任务失败:${e.message}`, + text: `提交视频任务失败:${e.message}${e.detail ? `\nDetails: ${JSON.stringify(e.detail)}` : ""}`, }, ], }; } } -async function handleGet({ args, token }) { +// ---------- op=get 的业务处理 ---------- +async function handleGet({ args, headers }) { + // 纯空白 task_id 不能直接拼进上游路径,否则得到一条含编码后空白的报错。 const taskId = typeof args.task_id === "string" ? args.task_id.trim() : ""; if (!taskId) { - return { - isError: true, - content: [{ type: "text", text: 'op=get 需要 "task_id"。' }], - }; + return { isError: true, content: [{ type: "text", text: 'op=get requires "task_id".' }] }; + } + + let target; + try { + target = resolveTarget(headers); + } catch (e) { + return { isError: true, content: [{ type: "text", text: e.message }] }; } try { - const data = await queryVideo({ token, taskId }); - const status = normStatus(data.status) || "unknown"; + const data = await queryVideo({ ...target, taskId }); + const status = normalizeStatus(data.status); if (DONE.has(status)) { - const url = (data.videos || [])[0]?.url; + const url = (data.videos || [])[0]?.url || (data.data || [])[0]?.url; if (!url) { return { isError: true, - content: [ - { - type: "text", - text: `任务已完成但没有返回视频地址。原始响应:${JSON.stringify(data).slice(0, 500)}`, - }, - ], + content: [{ type: "text", text: `任务已完成但没有视频地址。Raw: ${JSON.stringify(data).slice(0, 500)}` }], }; } const credit = data.usage?.credit; - const summary = + const text = `task_id: ${taskId}\n` + - `model: ${data.model || "-"}\n` + + `model: ${data.model || target.model}\n` + `status: completed\n` + `Video URL: ${url}\n` + (credit !== undefined ? `credit: ${credit}\n` : "") + - `\n注意:视频是签名直链,约 12 小时失效;要长期保存请立刻下载到本地。`; - return { content: [{ type: "text", text: summary }] }; + (data.hint ? `note: ${data.hint}\n` : ""); + return { content: [{ type: "text", text }] }; } - if (FAILED.has(status)) { - return { - isError: true, - content: [ - { - type: "text", - text: - `task_id: ${taskId}\n` + - `status: ${status}\n` + - (data.error ? `error: ${JSON.stringify(data.error)}\n` : "") + - `任务已失败,需要换提示词或模型重新提交。`, - }, - ], - }; - } - - // queued / in_progress:告诉模型继续等,别重复提交 - const progress = data.progress; + const errBlock = data.error ? `\nerror: ${JSON.stringify(data.error)}` : ""; + const progress = data.progress !== undefined ? `\nprogress: ${data.progress}` : ""; return { + isError: FAILED.has(status), content: [ { type: "text", - text: - `task_id: ${taskId}\n` + - `model: ${data.model || "-"}\n` + - `status: ${status}` + - (progress !== undefined && progress !== null ? `(进度 ${progress}%)` : "") + - `\n还没出片,稍后再用 op="get" 查一次。`, + text: `task_id: ${taskId}\nmodel: ${data.model || target.model}\nstatus: ${data.status || "unknown"}${progress}${errBlock}`, }, ], }; } catch (e) { return { isError: true, - content: [{ type: "text", text: `查询视频任务失败:${e.message}` }], + content: [ + { + type: "text", + text: `查询视频任务失败:${e.message}${e.detail ? `\nDetails: ${JSON.stringify(e.detail)}` : ""}`, + }, + ], }; } } function buildServer() { - const server = new McpServer( - { name: "video_task", version: "2.0.0" }, - { capabilities: { tools: {} } } - ); + const server = new McpServer({ name: "video_task", version: "2.0.0" }, { capabilities: { tools: {} } }); server.tool( "video_task", - `AI 视频生成任务(异步)。 + `AI video generation task (async). -操作: -- \`create\`:提交视频生成,必须给 \`prompt\`。立刻返回 \`task_id\`,生成在后台跑。 -- \`get\`:用 \`task_id\` 查进度,出片后返回视频直链。 +Operations: +- \`create\`: Start a new task. Requires at least one of \`prompt\` or \`image_url\`. Returns \`task_id\` immediately. +- \`get\`: Check a task by \`task_id\`. Returns status and, when completed, the video URL. -示例: -- 文生视频 → \`{"op":"create","prompt":"一只橘猫在阳光下的草地上奔跑,电影感"}\` -- 图生视频 → \`{"op":"create","prompt":"画面里的人缓缓转身","image_url":"https://..."}\` -- 查任务 → \`{"op":"get","task_id":""}\` +Examples: +- Text-to-video → \`{"op":"create","prompt":"A cat running in a garden"}\` +- Image-to-video → \`{"op":"create","image_url":"https://example.com/cat.jpg","prompt":"The cat runs forward"}\` +- Check task → \`{"op":"get","task_id":""}\` -工作流:\`create\` 调一次,然后反复用 \`get\` 查同一个 \`task_id\`,直到 status 变成 completed 或 failed。**不要为了看进度重复提交任务。** -耗时参考:minimax 约 2 分钟(扣 57 积分),kling 约 8 分钟(扣 103 积分)。视频直链约 12 小时失效。`, +Workflow: call \`create\` once, then call \`get\` with the returned \`task_id\` until status is \`completed\` or \`failed\`. Never create duplicate tasks just to check status.`, { - op: z - .enum(["create", "get"]) - .describe("create 提交生成;get 按 task_id 查结果。"), + op: z.enum(["create", "get"]).describe("create 提交生成;get 按 task_id 查结果。"), prompt: z .string() .max(2000) .optional() - .describe("视频描述,create 必填。图生视频时描述画面如何运动。"), + .describe("视频描述,create 时 prompt 与 image_url 至少给一个。图生视频时描述画面如何运动。"), image_url: z .string() .url() .optional() .describe( - "图生视频的参考图,公网可访问的 png/jpeg URL。本地文件先用 mint_public_link 换成公网短链。" + "参考图(图生视频):公开可访问的 png/jpeg URL。本地文件先用 agent 内置的公开临时 URL 工具换成短链。" ), seconds: z .number() - .min(4) - .max(15) - .optional() - .describe("视频时长(秒),默认 4。越长越慢越贵。"), - aspect_ratio: z - .string() - .optional() - .describe("画幅,如 16:9 / 9:16 / 1:1。上游不严格校验,写合法值即可。"), - resolution: z - .string() - .optional() - .describe("分辨率,如 768P / 1080P。留空用上游默认。"), - negative_prompt: z - .string() - .optional() - .describe("不希望出现的画面描述。"), - enable_audio: z - .boolean() - .optional() - .describe("是否生成音频,默认 false。"), - with_audio: z - .boolean() - .optional() - .describe("旧字段名,等同 enable_audio。"), - task_id: z - .string() + .int() .min(1) + .max(60) .optional() - .describe("create 返回的 task_id,get 时必填。"), + .describe("视频时长(秒)。留空用上游默认;越长越慢越贵。"), + aspect_ratio: z.string().optional().describe("画幅,如 16:9 / 9:16 / 1:1。"), + resolution: z.string().optional().describe("分辨率,如 768P / 1080P。留空用上游默认。"), + negative_prompt: z.string().optional().describe("不希望出现的画面描述。"), + enable_audio: z.boolean().optional().describe("是否生成音频,默认 false。"), + with_audio: z.boolean().optional().describe("旧字段名,等同 enable_audio。"), + task_id: z.string().min(1).optional().describe("create 返回的 task_id,get 时必填。"), }, async (args, extra) => { const headers = extra?.requestInfo?.headers; - const token = pickToken(headers); - if (!token) { - return { - isError: true, - content: [{ type: "text", text: "缺少鉴权:header 里要带 Authorization: Bearer 。" }], - }; - } - switch (args.op) { case "create": - return await handleCreate({ args, headers, token }); + return await handleCreate({ args, headers }); case "get": - return await handleGet({ args, token }); + return await handleGet({ args, headers }); default: return { isError: true, - content: [{ type: "text", text: `未知 op "${args.op}"(可用:create、get)。` }], + content: [{ type: "text", text: `Unknown op "${args.op}" (available: create, get).` }], }; } } @@ -384,19 +364,15 @@ async function startHttp({ port, host }) { closeAllSessions = () => { for (const { server, transport } of sessions.values()) { - try { - transport.close(); - } catch {} - try { - server.close(); - } catch {} + try { transport.close(); } catch {} + try { server.close(); } catch {} } sessions.clear(); httpServer?.closeAllConnections?.(); }; app.get("/health", (_req, res) => { - res.json({ status: "ok", service: "video_task", upstream: WB_BASE, sessions: sessions.size }); + res.json({ status: "ok", service: "video_task", sessions: sessions.size }); }); async function handleSessionRequest(req, res) { @@ -411,10 +387,7 @@ async function startHttp({ port, host }) { if (!isInitializeRequest(req.body)) { res.status(400).json({ jsonrpc: "2.0", - error: { - code: -32000, - message: "Bad Request: missing valid Mcp-Session-Id for non-initialize request", - }, + error: { code: -32000, message: "Bad Request: missing valid Mcp-Session-Id for non-initialize request" }, id: req.body?.id ?? null, }); return; @@ -424,15 +397,11 @@ async function startHttp({ port, host }) { const server = buildServer(); const transport = new StreamableHTTPServerTransport({ sessionIdGenerator: () => newId, - onsessioninitialized: (id) => { - sessions.set(id, { server, transport }); - }, + onsessioninitialized: (id) => { sessions.set(id, { server, transport }); }, }); transport.onclose = () => { - if (transport.sessionId) { - sessions.delete(transport.sessionId); - } + if (transport.sessionId) sessions.delete(transport.sessionId); }; await server.connect(transport); @@ -454,7 +423,7 @@ async function startHttp({ port, host }) { app.delete("/mcp", handleOther); httpServer = app.listen(port, host, () => { - console.log(`[video] MCP HTTP server listening on http://${host}:${port}/mcp → ${WB_BASE}`); + console.log(`[video] MCP HTTP server listening on http://${host}:${port}/mcp`); }); }