{"$schema":"https://wellknown.network/schemas/agent-record-v1.json","schemaVersion":"1","id":"ag_uvncyzapad73","handle":"vision-bridge-mcp","url":"https://wellknown.network/agents/vision-bridge-mcp","links":{"self":"https://wellknown.network/agents/vision-bridge-mcp/record.json","html":"https://wellknown.network/agents/vision-bridge-mcp","markdown":"https://wellknown.network/agents/vision-bridge-mcp/record.md","api":"https://wellknown.network/api/v1/agents/vision-bridge-mcp","status":"https://wellknown.network/api/v1/agents/vision-bridge-mcp/status","claim":"https://wellknown.network/agents/vision-bridge-mcp/claim","claimApi":"https://wellknown.network/api/v1/claims","claimDescriptor":"https://wellknown.network/agents/vision-bridge-mcp/claim.json","badge":"https://wellknown.network/agents/vision-bridge-mcp/badge.svg","openapi":"https://wellknown.network/openapi.json","history":"https://wellknown.network/api/v1/agents/vision-bridge-mcp/history","tools":"https://wellknown.network/api/v1/agents/vision-bridge-mcp/tools"},"ard":{"identifier":"urn:air::server:vision-bridge-mcp","type":"application/mcp-server-card+json"},"kind":"mcp_server","declared":{"name":"vision-bridge-mcp","summary":"MCP Server：用 MiniMax 多模态模型把图片解析成文字描述，供 DeepSeek 等纯文本模型使用","description":"# vision-bridge MCP Server\n\n图片视觉桥接服务：当对话里出现图片时，先由 **MiniMax 多模态模型** 把图片解析成文字描述，\n再交给 **DeepSeek（纯文本模型）** 继续推理。以 MCP Server 形式接入 ZCode / Trae 客户端。\n\n> 📄 **给其他机器/其他人部署**：请参考 [INTEGRATION.md](./INTEGRATION.md)（通用集成文档）。\n> 本文档描述的是本机已完成的配置。\n\n```\n用户发图 ──► ZCode/Trae (DeepSeek 对话) ──调用 MCP 工具──► vision-bridge ──► MiniMax 多模态\n                                                                    ◄── 图片的文字描述 ──\n         DeepSeek 基于文字描述继续回答 ◄────────────────────────────────────┘\n```\n\n## 提供的工具\n\n| 工具 | 作用 |\n|------|------|\n| `analyze_image(image?, question?)` | 图片理解：返回图片内容的文字描述，或回答关于图片的具体问题 |\n| `extract_text_from_image(image?, language?)` | 图片 OCR：按原始顺序提取图中所有文字（标题/正文/按钮/水印等） |\n\n`image` 参数支持：本地文件路径、`http(s)` 链接、base64 data URI。\n**也可以不传**：工具会自动定位用户在当前 ZCode 会话中最新上传的图片附件\n（`~/.zcode/cli/artifacts/<session>/prompt-attachment-upload-*.txt`，毫秒级定位）。\n定位逻辑：ZCode 的 UserPromptSubmit hook 会把当前会话 id 写入\n`~/.zcode/vision-inbox/current-session.txt`，server 据此**只扫当前会话目录**，\n不受其他会话干扰；当前会话无附件时才兜底取全局最新。\n\n## 安装\n\n```bash\ncd /Users/wjc/mcp-vision-server\npython3 -m venv .venv\n.venv/bin/pip install -r requirements.txt\n.venv/bin/pip install -e .          # 生成本机 vision-bridge 命令（editable，改代码即时生效）\n```\n\n## 配置 API Key\n\n`.env` 有两个位置（server 启动时会依次加载，环境变量优先级最高）：\n\n1. **当前工作目录**的 `.env` —— 本地开发（本机已配置）\n2. **`~/.config/vision-bridge/.env`** —— 全局配置，**推荐**：MCP 客户端启动 server 时\n   cwd 通常是工作区而非项目目录，放这里任何场景都能加载（本机已配置）\n\n```ini\nMINIMAX_API_KEY=你的密钥\nMINIMAX_GROUP_ID=      # 国内版 (api.minimax.chat) 必填；国际版 (api.minimaxi.com) 留空\nMINIMAX_BASE_URL=https://api.minimax.chat/v1   # 国际版改为 https://api.minimaxi.com/v1\nMINIMAX_MODEL=MiniMax-M2                        # 按你账号开通的模型调整（如 MiniMax-M3）\n```\n\n> 密钥只放 `.env`，不要提交到 git（已加入 `.gitignore`）。\n\n## 接入 ZCode（已完成注册）\n\n已写入 `~/.zcode/cli/config.json`（使用安装生成的 `vision-bridge` 命令，无需关心项目路径）：\n\n```json\n{\n  \"mcp\": {\n    \"servers\": {\n      \"vision-bridge\": {\n        \"command\": \"/Users/wjc/mcp-vision-server/.venv/bin/vision-bridge\",\n        \"args\": []\n      }\n    }\n  }\n}\n```\n\n**重启 ZCode 后**在「设置 → MCP」中应能看到 vis…","publisher":null,"homepage":null,"repository":null,"version":"0.1.0","license":"MIT","protocols":["mcp"],"tags":["mcp","minimax","vision","deepseek","图片识别","ocr"],"pricing":null,"endpoints":[{"url":"pypi:vision-bridge-mcp","type":"package_pypi","auth":null,"probeable":false}],"skills":null,"tools":null,"extra":null,"attribution":{"kind":"pypi","name":"pypi","license":"pypi","summary":"pypi","version":"pypi","description":"pypi"}},"derived":{"capabilities":[{"slug":"media.image-understanding","name":"Image Understanding","confidence":1,"provenance":"declared"},{"slug":"documents.ocr","name":"OCR","confidence":1,"provenance":"declared"}],"categories":["documents","media"],"language":"fr"},"observed":{"status":"unknown","statusReason":"Distributed as a package to run locally; no network endpoint to check.","lastOkAt":null,"lastProbedAt":null,"statusComputedAt":null,"reliability30d":null,"latestObservations":[],"tools":null,"package":{"name":"vision-bridge-mcp","registry":"pypi","observedAt":"2026-10-22T23:26:16.504Z","publishedAt":"2026-08-07T06:54:00.259107Z","latestVersion":"0.1.0"},"toolSurface":null,"endpointFacts":[]},"verification":{"claimed":false,"claimedAt":null,"proofs":[]},"provenance":{"sources":[{"source":"pypi","key":"vision-bridge-mcp","url":"https://pypi.org/project/vision-bridge-mcp/","firstSeenAt":"2026-09-10T15:23:41.436Z","fetchedAt":"2026-10-22T23:24:42.425Z","normalizedAt":"2026-10-22T23:24:42.425Z"}]},"firstSeenAt":"2026-09-10T15:23:41.436Z","updatedAt":"2026-10-22T23:26:16.504Z"}