{"$schema":"https://wellknown.network/schemas/agent-record-v1.json","schemaVersion":"1","id":"ag_has9mhwgncsf","handle":"vision-mcp-ms","url":"https://wellknown.network/agents/vision-mcp-ms","links":{"self":"https://wellknown.network/agents/vision-mcp-ms/record.json","html":"https://wellknown.network/agents/vision-mcp-ms","markdown":"https://wellknown.network/agents/vision-mcp-ms/record.md","api":"https://wellknown.network/api/v1/agents/vision-mcp-ms","status":"https://wellknown.network/api/v1/agents/vision-mcp-ms/status","claim":"https://wellknown.network/agents/vision-mcp-ms/claim","claimApi":"https://wellknown.network/api/v1/claims","claimDescriptor":"https://wellknown.network/agents/vision-mcp-ms/claim.json","badge":"https://wellknown.network/agents/vision-mcp-ms/badge.svg","openapi":"https://wellknown.network/openapi.json","history":"https://wellknown.network/api/v1/agents/vision-mcp-ms/history","tools":"https://wellknown.network/api/v1/agents/vision-mcp-ms/tools"},"ard":{"identifier":"urn:air::server:vision-mcp-ms","type":"application/mcp-server-card+json"},"kind":"mcp_server","declared":{"name":"vision-mcp-ms","summary":"云端视觉 MCP：通过 OpenAI 兼容接口（硅基流动等）分析图片，为纯文本大模型提供看图能力。","description":"# vision-mcp-ms\n\n云端视觉 MCP 服务：通过 **OpenAI 兼容接口**（默认硅基流动 SiliconFlow）分析图片，\n为 DeepSeek 这类纯文本大模型提供「看图」能力。\n\n- 单一工具 `analyze_image(image, prompt)`\n- 支持 HTTP(S) 图片 URL 和 base64 data URL\n- 多模型按顺序自动 fallback（限流 / 超时 / 5xx 时切下一个）\n- 纯 API 调用、不依赖本地环境，**可部署到魔搭 MCP 广场云端托管**\n\n## 环境变量\n\n| 变量 | 必填 | 默认值 | 说明 |\n| --- | --- | --- | --- |\n| `OPENAI_API_KEY` | 是* | - | API Key（也兼容 `SILICONFLOW_API_KEY` / `API_KEY`） |\n| `OPENAI_BASE_URL` | 否 | `https://api.siliconflow.cn/v1` | OpenAI 兼容接口地址 |\n| `VISION_MODELS` | 否 | `Qwen/Qwen2.5-VL-7B-Instruct` | 逗号分隔的有序模型列表，靠前的优先 |\n| `VISION_REQUEST_TIMEOUT_MS` | 否 | `60000` | 单次模型请求超时（毫秒） |\n\n\\* 部署时在魔搭 / MCP 客户端的 env 里配置。\n\n## 本地使用（stdio）\n\n在 MCP 客户端（Cherry Studio 桌面版等）里添加：\n\n```json\n{\n  \"mcpServers\": {\n    \"vision-mcp-ms\": {\n      \"command\": \"uvx\",\n      \"args\": [\"vision-mcp-ms\"],\n      \"env\": {\n        \"OPENAI_API_KEY\": \"sk-你的硅基流动Key\",\n        \"OPENAI_BASE_URL\": \"https://api.siliconflow.cn/v1\",\n        \"VISION_MODELS\": \"Qwen/Qwen2.5-VL-7B-Instruct,Qwen/Qwen2.5-VL-72B-Instruct\"\n      }\n    }\n  }\n}\n```\n\n> 本地 stdio 模式下，`image` 参数也支持**本地文件绝对路径**（如 `D:\\Pictures\\a.png`）。\n> 云端托管模式下仅支持 URL / data URL（服务端读不到你的本地文件）。\n\n## 开发 / 测试\n\n```bash\nuv sync                     # 建环境\nuv run python -m vision_mcp # 以 stdio 模式启动\nuv run python -c \"from vision_mcp.server import _to_data_url; print(_to_data_url('https://httpbin.org/image/jpeg')[:40])\"\n```\n\n## 发布到 PyPI\n\n```bash\ncd vision-mcp-ms\nuv run python -m build\nuv run twine upload dist/*.whl\n```\n\n## 部署到魔搭 MCP 广场（云端托管）\n\n1. 把代码推到 GitHub 仓库（或发布到 PyPI 后）。\n2. 打开 <https://www.modelscope.cn/mcp/servers/create?template=customize>\n3. 托管类型选「**可托管部署**」，来源选你的 GitHub 仓库 / PyPI 包。\n4. 配置 `command` / `args`（如 `uvx` / `vision-mcp-ms`），在 env 里填\n   `OPENAI_API_KEY`、`OPENAI_BASE_URL`、`VISION_MODELS`。\n5. 创建后魔搭会自动部署，得到一个 HTTP 地址，填入手机 MCP 客户端即可。","publisher":null,"homepage":null,"repository":null,"version":"0.1.1","license":"MIT","protocols":["mcp"],"tags":["image","mcp","modelcontextprotocol","siliconflow","vision"],"pricing":null,"endpoints":[{"url":"pypi:vision-mcp-ms","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"}],"categories":["media"],"language":"en"},"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-mcp-ms","registry":"pypi","observedAt":"2026-10-22T23:26:16.779Z","publishedAt":"2026-08-12T17:44:57.416130Z","latestVersion":"0.1.1"},"toolSurface":null,"endpointFacts":[]},"verification":{"claimed":false,"claimedAt":null,"proofs":[]},"provenance":{"sources":[{"source":"pypi","key":"vision-mcp-ms","url":"https://pypi.org/project/vision-mcp-ms/","firstSeenAt":"2026-09-10T15:23:42.167Z","fetchedAt":"2026-10-22T23:24:43.142Z","normalizedAt":"2026-10-22T23:24:43.142Z"}]},"firstSeenAt":"2026-09-10T15:23:42.167Z","updatedAt":"2026-10-22T23:26:16.779Z"}