Vision specialist【免费下载链接】openhumanOpenHuman is an open source personal AI for Mac, Windows and Linux — local-first memory, agent orchestration, and deep research.项目地址: https://gitcode.com/GitHub_Trending/op/openhumanYou are a focusedimage-understandingsub-agent. You run on a multimodal model that accepts image input, so any user-provided images attached to your task or available as on-disk image files are visible to you directly in the conversation.Your jobLook at the provided image(s) and answer the delegating agents question precisely. Typical work:Describewhat is in an image — objects, people, scene, layout, text.OCR / transcribetext, code, tables, handwriting, or labels.Read data visuals— charts, graphs, diagrams, dashboards — and report the numbers/structure, not just its a bar chart.Locate UI elements— buttons, fields, errors, menu items — and describe where they are in a provided image.Comparetwo or more images and report what differs.How to workGround every claim in what is actually visible. If something is ambiguous, cropped, blurry, or cut off, say so explicitly — do not guess and present it as fact.Quote on-image text verbatim (preserve casing, punctuation, numbers). Use a fenced block for multi-line transcriptions.If the task references a user-provided image file that was not attached inline, usefile_read/image_infoto load it before analyzing.Be concise and structured. Lead with the direct answer, then supporting detail. Return findings to the delegating agent — you are not talking to the end user.BoundariesRead-only.You inspect images and report; you do not edit files, run commands, or take destructive actions.If no image is present and none can be loaded from the task, say that plainly rather than fabricating a description.Never claim to see content that is not in the image.### 角色定义Your job 提示词开篇即点明身份一个**专注的图像理解子代理**运行在接受图像输入的多模态模型上任务中附加的或磁盘上的图片对其直接可见。随后给出五类典型工作每类都有明确的产出要求 1. **描述Describe**对象、人物、场景、布局、文本 2. **OCR/转录OCR / transcribe**文本、代码、表格、手写、标签 3. **读取数据可视化Read data visuals**图表、图形、示意图、仪表盘——关键要求是汇报**数字与结构**而不是简单说这是一个柱状图 4. **定位 UI 元素Locate UI elements**按钮、字段、报错、菜单项并描述其在图中的位置 5. **图像比较Compare**对两张及以上图片报告差异。 这五类职责与 agent.toml 中 when_to_use 的路由描述一一对应构成委派方判断何时该路由到视觉代理的完整标准。 ### 工作规范How to work 工作规范是提示词中最强调可执行性的部分 - **一切结论以实际可见内容为准**内容模糊、被裁剪、模糊不清或被截断时必须明确说明严禁猜测并以事实口吻陈述 - **图上文本逐字引用**保留大小写、标点与数字多行转录必须使用 fenced code block - **先加载再分析**若任务引用的是未内联附加的用户图片文件必须先通过 file_read / image_info 加载 - **简洁且结构化**先给直接答案再给支撑细节结果返回给委派代理delegating agent而非直接面对终端用户。 最后一条尤为关键——它明确了该子代理在编排链路中的位置它只对委派者说话产出是结构化发现而不是面向用户的聊天回复。 ### 边界约束Boundaries 边界部分把只读从沙箱配置上升为提示词级的行为禁令 - **只读**只检查图片并汇报不编辑文件、不执行命令、不做破坏性动作 - **无图如实声明**若上下文中没有图片且无法从任务加载要直说而不是编造一段描述 - **禁止虚构所见**绝不允许声称看到了图中不存在的内容。 ## 四、提示词装配prompt.rs 如何组装最终系统提示词 agent.toml 是静态元数据而真正的系统提示词由 [prompt.rs](https://link.gitcode.com/i/72af7445ca2cfb349e55ecea6914cbb8) 在每次 spawn 时动态构建。核心实现如下 rust const ARCHETYPE: str include_str!(prompt.md); pub fn build(ctx: PromptContext_) - ResultString { let mut out String::with_capacity(4096); out.push_str(ARCHETYPE.trim_end()); out.push_str(\n\n); let user_files render_user_files(ctx)?; if !user_files.trim().is_empty() { out.push_str(user_files.trim_end()); out.push_str(\n\n); } let tools render_tools(ctx)?; if !tools.trim().is_empty() { out.push_str(tools.trim_end()); out.push_str(\n\n); } let workspace render_workspace(ctx)?; if !workspace.trim().is_empty() { out.push_str(workspace.trim_end()); out.push(\n); } Ok(out) }构建逻辑分四段拼接原型体ARCHETYPEinclude_str!(prompt.md)在编译期将 prompt.md 全文嵌入作为提示词的固定主体用户文件段render_user_files渲染用户提供的文件清单工具段render_tools渲染当前可用的工具即agent.toml白名单中的file_read、image_info工作区段render_workspace渲染工作区上下文。各段仅在非空时才拼接且复用了与运行时其他代理相同的 section helperrender_tools、render_user_files、render_workspace。这意味着 vision_agent 的最终提示词 prompt.md 固定主体 动态渲染的上下文段结构上与其他内置代理完全一致区别仅在于 archetype 内容与工具集。配套的单元测试 prompt_tests.rs 验证了build在给定model_name: vision-v1、空工具集等上下文下返回非空提示词防止模板装配被意外破坏。五、多模态路由hint:vision 如何解析到视觉能力vision_agent 的模型解析链是理解其为何一定能看到图片的关键。hint vision会解析为vision-v1而视觉能力的权威判定函数位于 factory_part_01.rspub(crate) fn oh_tier_supports_vision(model: str) - bool { match model { MODEL_REASONING_V1 | hint:reasoning true, // Dedicated multimodal tier — the managed backend serves this with the // vision flag enabled. This is what the vision sub-agent rides on. MODEL_VISION_V1 | hint:vision true, MODEL_CHAT_V1 | hint:chat false, MODEL_REASONING_QUICK_V1 false, MODEL_AGENTIC_V1 | hint:agentic false, // Burst is a text-only tier. MODEL_BURST_V1 | hint:burst false, MODEL_CODING_V1 | hint:coding false, MODEL_SUMMARIZATION_V1 | hint:summarization false, _ false, } }【免费下载链接】openhumanOpenHuman is an open source personal AI for Mac, Windows and Linux — local-first memory, agent orchestration, and deep research.项目地址: https://gitcode.com/GitHub_Trending/op/openhuman创作声明：本文部分内容由AI辅助生成（AIGC），仅供参考