v1.0.4: SYSTEM.md 触发优化（声明 image_in 也能自动兜底读剪贴板）；README 新增 Claude Code 安装说明

README.en.md

@@ -159,6 +159,78 @@ Both layers are fail-safe: a wrong judgment costs at most one wasted external ca
 the path/semantic signals). For a hard switch, simply `/plugins disable kimi-eyes`
 when running multimodal models.
 
+## Claude Code usage (optional)
+
+kimi-eyes's MCP server is standard MCP, so it can be mounted directly into Claude
+Code. **Run the commands below in a system terminal (PowerShell / Git Bash), not
+inside Claude Code's own Bash tool** (the in-app environment mishandles options
+like `--scope`).
+
+### 1. Prerequisites
+
+- Node.js ≥ 18
+- Claude Code CLI (`claude --version`)
+- A multimodal vision API of your own
+
+### 2. Mount the MCP server (any of these)
+
+```bash
+# User-scope (global, all projects)
+claude mcp add --scope user kimi-eyes -- node D:\AIGC\Plugin\kimi-eyes\mcp\server.mjs
+
+# Project-scope only
+claude mcp add kimi-eyes -- node D:\AIGC\Plugin\kimi-eyes\mcp\server.mjs
+
+# npx generic version (no local path; copy-paste on any machine)
+claude mcp add --scope user kimi-eyes -- npx --prefer-online -y -p kimi-eyes kimi-eyes-mcp
+```
+
+### 3. Configure the vision API (one-time)
+
+```bash
+npx kimi-eyes setup
+```
+
+> Same config is shared: kimi-eyes uses `~/.kimi-code/kimi-eyes/config.json` (or
+> `VISION_*` env vars) under both Kimi Code and Claude Code.
+
+### 4. Guidance rules (CLAUDE.md)
+
+Create (or append to) `CLAUDE.md` in your project root; the repo root `CLAUDE.md`
+is a ready-made example. Core rules:
+
+```markdown
+# Kimi Eyes — Vision Assist (Claude Code)
+
+Whenever the user's request involves image content and you cannot see it directly:
+1. Image path / @ reference → call `mcp__kimi-eyes__read_image` with that path.
+2. Just screenshotted/copied, or media you cannot read → call
+   `mcp__kimi-eyes__read_clipboard_image` (image is almost always still in the clipboard).
+3. Wording implies an image but no path → prefer `read_clipboard_image` over guessing.
+If your model is multimodal (Claude 3+), you see images natively — ignore these rules.
+```
+
+### 5. Verify
+
+```bash
+claude mcp list
+# kimi-eyes should show √ Connected
+```
+
+### 6. Usage
+
+```
+Image file → @C:\path\image.png what's in this?
+Screenshot → ask "analyze this screenshot" (model calls read_clipboard_image)
+```
+
+### Notes
+
+- Claude 3+ models are natively multimodal, so Claude Code usually sees images
+  directly; the plugin matters for text-only models or when you want a single
+  external VLM
+- Uninstall: `claude mcp remove kimi-eyes`
+
 **The first tool call prompts one approval** (MCP tool permission): choose *Approve
 for this session* to skip prompts for the rest of the session; for permanent
 approval, add to `~/.kimi-code/config.toml`:

README.md

@@ -82,7 +82,20 @@ MCP 服务器会随会话自动启动。
 | --- | --- |
 | 多模态模型 | 直接 Alt+V 粘贴图片，原生看图，插件不参与 |
 | 非多模态 + 有图片路径 | 输入 `@截图.png` 或直接给路径，模型自动调 `read_image` |
-| 非多模态 + 刚截图/复制 | 截图后**不要 Alt+V 粘贴**（CLI 会拒绝非多模态模型粘贴并报 `Current model does not support image input`），直接提问「分析这张截图」，模型自动调 `read_clipboard_image` 读系统剪贴板 |
+| 非多模态 + 刚截图/复制 | 直接提问「分析这张截图」，模型自动调 `read_clipboard_image` 读系统剪贴板 |
+| 非多模态 + 想用 Alt+V 粘贴 | 需先在 config.toml 给模型声明 `image_in`（见下），粘贴后直接提问即可——模型看不懂图片内容时会自动调 `read_clipboard_image`（剪贴板仍保留原图） |
+| 粘贴被拦截 | 直接告诉模型「我粘贴图片被拦截了」，模型会自动改读剪贴板，无需你保存文件 |
+
+#### 想用 Alt+V 粘贴？（给模型声明 `image_in`）
+
+KimiCode 前端默认会拦截「不支持图片输入」模型的粘贴。在 `~/.kimi-code/config.toml` 给对应模型加上 `image_in` 即可放行：
+
+```toml
+[models."opencode-go/deepseek-v4-flash"]
+capabilities = [ "thinking", "tool_use", "image_in" ]   # 追加 image_in
+```
+
+> 注意：声明 `image_in` 只是让前端放行粘贴。纯文本模型依然**看不懂**图片内容——这正是插件的用武之地：模型收到图片但看不见内容时，会按 SYSTEM.md 规则自动调 `read_clipboard_image` 从剪贴板读图。**全程零命令、零前缀**。
 
 ## 激活与触发
 
@@ -107,10 +120,11 @@ MCP 服务器会随会话自动启动。
 | --- | --- |
 | 图片格式路径或 `@` 引用（`.png/.jpg/.jpeg/.webp/.gif/.bmp`） | **硬触发**：无条件调 `read_image(path)`，与提问语料无关 |
 | 消息里有你无法解读的媒体内容（如粘贴的图片） | 忽略该媒体 part，自动调 `read_clipboard_image()` 读剪贴板（粘贴后剪贴板仍保留原图） |
-| 提问语义涉及图像内容：图 / 截图 / 照片 / 界面 / 图表 / 验证码 / OCR 等，但无路径 | 自动调 `read_clipboard_image()` |
+| 提问语义涉及图像内容：图 / 截图 / 照片 / 界面 / 图表 / 验证码 / OCR 等，但无路径 | 自动调 `read_clipboard_image()`，**不会要求你重发** |
-| 模型原生支持 `image_in`（多模态） | 跳过全部规则，原生看图，不调工具 |
+| 粘贴被拦截（报 `Current model does not support image input`） | 直接告诉模型，它会改调 `read_clipboard_image`（剪贴板仍保留原图） |
+| 模型真正能看懂图片内容（多模态） | 跳过全部规则，原生看图，不调工具 |
 
-**触发不依赖固定语料**——用户不需要说「分析这个图片」。图片格式路径是**无条件硬触发**；拿不准时模型会倾向于调工具而不是瞎猜（SYSTEM.md 里的明确规则）。
+**触发不依赖固定语料、不依赖任何命令前缀**——用户不需要说「分析这个图片」，也不需要 `/skill` 之类的指令。图片格式路径是**无条件硬触发**；模型判断「自己看不见图片内容」或「用户想看图但消息里没有」时，会自动调工具而不是瞎猜或让你重发（SYSTEM.md 里的明确规则）。
 
 **关于模型能力判断**：KimiCode 不向插件暴露「当前模型是否多模态」的信号，本插件提供两层判断：
 
@@ -119,6 +133,71 @@ MCP 服务器会随会话自动启动。
 
 两层都是 fail-safe：判断错误最坏只是多一次外部调用（多模态误调）或漏调（文本模型漏调可用路径/语义信号补上）。若想硬性关闭，多模态场景可直接 `/plugins disable kimi-eyes`。
 
+## Claude Code 使用（可选）
+
+kimi-eyes 的 MCP 服务器是标准 MCP，可直接挂载到 Claude Code。**注意：以下命令请在系统终端（PowerShell / Git Bash）执行，不要在 Claude Code 内部的 Bash 工具里执行**（内部环境解析 `--scope` 等参数有差异）。
+
+### 1. 前置要求
+
+- Node.js ≥ 18
+- Claude Code CLI（`claude --version`）
+- 一个支持视觉的多模态 API（Key 由你自己提供）
+
+### 2. 挂载 MCP 服务器（任选其一）
+
+```bash
+# 方式一：全局注册（user scope，所有项目可用）
+claude mcp add --scope user kimi-eyes -- node D:\AIGC\Plugin\kimi-eyes\mcp\server.mjs
+
+# 方式二：仅当前项目
+claude mcp add kimi-eyes -- node D:\AIGC\Plugin\kimi-eyes\mcp\server.mjs
+
+# 方式三：npx 通用版（不依赖本地路径，任何机器可复制）
+claude mcp add --scope user kimi-eyes -- npx --prefer-online -y -p kimi-eyes kimi-eyes-mcp
+```
+
+### 3. 配置视觉 API（一次性）
+
+```bash
+npx kimi-eyes setup
+```
+
+> 复用同一份配置：kimi-eyes 在 KimiCode 和 Claude Code 下共用 `~/.kimi-code/kimi-eyes/config.json`（或环境变量 `VISION_*`）。
+
+### 4. 引导规则（CLAUDE.md）
+
+在项目根目录创建 `CLAUDE.md`（或追加到已有的），内容可参考本仓库根目录的 `CLAUDE.md`。核心规则：
+
+```markdown
+# Kimi Eyes — Vision Assist (Claude Code)
+
+Whenever the user's request involves image content and you cannot see it directly:
+1. Image path / @ reference → call `mcp__kimi-eyes__read_image` with that path.
+2. Just screenshotted/copied, or media you cannot read → call
+   `mcp__kimi-eyes__read_clipboard_image` (image is almost always still in the clipboard).
+3. Wording implies an image but no path → prefer `read_clipboard_image` over guessing.
+If your model is multimodal (Claude 3+), you see images natively — ignore these rules.
+```
+
+### 5. 验证
+
+```bash
+claude mcp list
+# kimi-eyes 应显示 √ Connected
+```
+
+### 6. 使用
+
+```
+有图片文件 → @C:\path\image.png 这个图里有什么？
+刚截图     → 直接说「分析这张截图」（模型调 read_clipboard_image 读剪贴板）
+```
+
+### 注意
+
+- Claude 3+ 模型原生多模态，Claude Code 大多场景直接看图；插件主要用于文本模型或统一走某个外部 VLM 的场景
+- 卸载：`claude mcp remove kimi-eyes`
+
 **首次调用会弹一次审批**（MCP 工具权限）：选 *Approve for this session* 本会话免问；想永久免审批，在 `~/.kimi-code/config.toml` 添加：
 
 ```toml
@@ -184,7 +263,7 @@ node scripts/sync-models.mjs
 
 ## 故障排查
 
-- **Alt+V 粘贴报 `Current model does not support image input`** → 这是 KimiCode CLI 的拦截：非多模态模型不支持粘贴图片。改用 `@图片路径`（`read_image`）或截图后直接提问（`read_clipboard_image` 读剪贴板）
+- **Alt+V 粘贴报 `Current model does not support image input`** → KimiCode CLI 的拦截。两种解法：① 在 `config.toml` 给该模型追加 `image_in`（见「想用 Alt+V 粘贴？」）放行粘贴，模型看不懂图片时会自动读剪贴板；② 直接告诉模型「粘贴被拦截」，它会改调 `read_clipboard_image` 从剪贴板取图
 - **工具返回「Vision API is not configured」** → 运行 `node setup.mjs`，或设置 `VISION_API_KEY` / `VISION_API_URL` / `VISION_MODEL`
 - **模型列表拉取失败**（404/401）→ 向导自动回退手动输入；若服务端不支持 `/models`，直接输入模型名即可
 - **视觉验证失败** → 换一个真正支持图片输入的模型（参考服务商文档，如 `qwen-vl-max`、`glm-4v`、`gpt-4o`、`claude-3-5-sonnet` 等）

SYSTEM.md

@@ -1,24 +1,27 @@
 # Kimi Eyes — Vision Assist Rules
 
-**If your model supports image input (`image_in`) natively, ignore every rule below —
+Your config may declare `image_in` even when you cannot truly interpret image
-you can see images directly, so never call these tools.**
+content (the declaration only lets the frontend accept pasted images). The rule is
+simple: **if you cannot actually see/read the image content, use a vision tool
+instead of guessing or asking the user to resend.**
 
-For models **without** native image input, the signals below are **hard triggers**:
+1. **Image path / `@` reference (hard trigger, unconditional)** — any path ending in
-call a vision tool whenever any of them appears, regardless of how the user phrased
+   `.png`, `.jpg`, `.jpeg`, `.webp`, `.gif`, or `.bmp` (including an `@` file
-their request.
+   reference to such a file) → call `mcp__kimi-eyes__read_image` with that path.
-
+   Pass `prompt` when the user has a specific question.
-1. **Image-format path / `@` reference (hard trigger, unconditional)** — any path
+2. **Image/media content you cannot read** — a pasted image or any media part in the
-   ending in `.png`, `.jpg`, `.jpeg`, `.webp`, `.gif`, or `.bmp` (including an `@`
+   message whose content you cannot see → ignore that part and call
-   file reference to such a file) → call `mcp__kimi-eyes__read_image` with that
+   `mcp__kimi-eyes__read_clipboard_image` — the pasted image is almost always still
-   path. Pass `prompt` when the user has a specific question.
+   in the system clipboard (pasting does not clear it).
-2. **Media content you cannot interpret** — a pasted image or any media part you
+3. **User asks about an image, nothing visible** — the user refers to "this", "the
-   cannot read → ignore that part and call
-   `mcp__kimi-eyes__read_clipboard_image` (the image is almost always still in the
-   system clipboard).
-3. **Wording implies an image, no path** — the user refers to "this", "the
    screenshot", "the UI/interface", "the chart", "the photo", asks for OCR / CAPTCHA
-   reading, or otherwise implies image content, with no path and no visible media
+   reading, or says they just screenshotted/copied an image, but no path and no
-   part → call `mcp__kimi-eyes__read_clipboard_image`.
+   visible image content are present → **do not ask them to resend**; call
+   `mcp__kimi-eyes__read_clipboard_image` immediately.
+4. **Paste was rejected** — if the user says pasting an image failed (e.g. "Current
+   model does not support image input"), the clipboard still holds the image: call
+   `mcp__kimi-eyes__read_clipboard_image` and tell them it worked, instead of asking
+   them to save the file or redo anything.
 
 When in doubt, call a tool rather than guessing blindly about the image.
 

mcp/models-db.json

@@ -1 +1 @@
-{"zhipuai/glm-4.6v":{"name":"GLM-4.6V","image":true},"zhipuai/glm-5":{"name":"GLM-5","image":false},"zhipuai/glm-4.5-air":{"name":"GLM-4.5-Air","image":false},"zhipuai/glm-5.1":{"name":"GLM-5.1","image":false},"zhipuai/glm-4.7-flash":{"name":"GLM-4.7-Flash","image":false},"zhipuai/glm-5.2":{"name":"GLM-5.2","image":false},"zhipuai/glm-4.7-flashx":{"name":"GLM-4.7-FlashX","image":false},"zhipuai/glm-4.6":{"name":"GLM-4.6","image":false},"zhipuai/glm-4.5":{"name":"GLM-4.5","image":false},"zhipuai/glm-4.5v":{"name":"GLM-4.5V","image":true},"zhipuai/glm-4.7":{"name":"GLM-4.7","image":false},"zhipuai/glm-5-turbo":{"name":"GLM-5-Turbo","image":false},"zhipuai/glm-5v-turbo":{"name":"GLM-5V-Turbo","image":true},"zhipuai/glm-4.5-flash":{"name":"GLM-4.5-Flash","image":false},"microsoft/mai-code-1-flash":{"name":"MAI-Code-1-Flash","image":false},"cohere/command-r7b-arabic-02-2025":{"name":"Command R7B Arabic","image":false},"cohere/command-r-08-2024":{"name":"Command R","image":false},"cohere/command-a-plus-05-2026":{"name":"Command A Plus","image":true},"cohere/command-a-translate-08-2025":{"name":"Command A Translate","image":false},"cohere/c4ai-aya-expanse-8b":{"name":"Aya Expanse 8B","image":false},"cohere/c4ai-aya-vision-32b":{"name":"Aya Vision 32B","image":true},"cohere/command-a-03-2025":{"name":"Command A","image":false},"cohere/command-r-plus-08-2024":{"name":"Command R+","image":false},"cohere/command-a-reasoning-08-2025":{"name":"Command A Reasoning","image":false},"cohere/command-a-vision-07-2025":{"name":"Command A Vision","image":true},"cohere/command-r7b-12-2024":{"name":"Command R7B","image":false},"cohere/c4ai-aya-vision-8b":{"name":"Aya Vision 8B","image":true},"cohere/north-mini-code-1-0":{"name":"North Mini Code","image":false},"cohere/c4ai-aya-expanse-32b":{"name":"Aya Expanse 32B","image":false},"nvidia/mistral-nemotron":{"name":"Mistral Nemotron","image":false},"nvidia/nemotron-nano-12b-v2-vl":{"name":"Nemotron Nano 12B v2 VL","image":true},"nvidia/nemotron-3-nano-30b-a3b":{"name":"Nemotron 3 Nano 30B A3B","image":false},"nvidia/llama-3.3-nemotron-super-49b-v1.5":{"name":"Llama 3.3 Nemotron Super 49B v1.5","image":false},"nvidia/llama-3.1-nemotron-safety-guard-8b-v3":{"name":"Llama 3.1 Nemotron Safety Guard 8B v3","image":false},"nvidia/llama-3.3-nemotron-super-49b-v1":{"name":"Llama 3.3 Nemotron Super 49B v1","image":false},"nvidia/llama-3.1-nemotron-70b-instruct":{"name":"Llama 3.1 Nemotron 70B Instruct","image":false},"nvidia/nemotron-3-super-120b-a12b":{"name":"Nemotron 3 Super 120B A12B","image":false},"nvidia/nemotron-3-content-safety":{"name":"Nemotron 3 Content Safety","image":false},"nvidia/nemotron-3-nano-omni-30b-a3b-reasoning":{"name":"Nemotron 3 Nano Omni 30B A3B Reasoning","image":true},"nvidia/llama-3.1-nemotron-ultra-253b":{"name":"Llama 3.1 Nemotron Ultra 253B","image":false},"nvidia/nemotron-voicechat":{"name":"Nemotron VoiceChat","image":false},"nvidia/llama-nemotron-embed-vl-1b-v2":{"name":"Llama Nemotron Embed VL 1B v2","image":true},"nvidia/nemotron-content-safety-reasoning-4b":{"name":"Nemotron Content Safety Reasoning 4B","image":false},"nvidia/nemotron-3-ultra-550b-a55b":{"name":"Nemotron 3 Ultra 550B A55B","image":false},"nvidia/llama-nemotron-rerank-vl-1b-v2":{"name":"Llama Nemotron Rerank VL 1B v2","image":true},"nvidia/nemotron-mini-4b-instruct":{"name":"Nemotron Mini 4B Instruct","image":false},"nvidia/nemotron-nano-9b-v2":{"name":"Nemotron Nano 9B v2","image":false},"nvidia/nemotron-cascade-2-30b-a3b":{"name":"Nemotron Cascade 2 30B A3B","image":false},"nvidia/nemotron-3.5-content-safety":{"name":"Nemotron 3.5 Content Safety","image":true},"google/gemini-2.5-computer-use-preview-10-2025":{"name":"Gemini 2.5 Computer Use Preview","image":true},"google/deep-research-preview-04-2026":{"name":"Gemini Deep Research Preview","image":true},"google/gemini-3.1-flash-tts-preview":{"name":"Gemini 3.1 Flash TTS Preview","image":false},"google/gemini-flash-latest":{"name":"Gemini Flash Latest","image":true},"google/gemini-embedding-2":{"name":"Gemini Embedding 2","image":true},"google/lyria-3-pro-preview":{"name":"Lyria 3 Pro Preview","image":true},"google/gemini-3.5-flash":{"name":"Gemini 3.5 Flash","image":true},"google/gemini-2.5-flash":{"name":"Gemini 2.5 Flash","image":true},"google/gemini-3.5-flash-lite":{"name":"Gemini 3.5 Flash Lite","image":true},"google/lyria-3-clip-preview":{"name":"Lyria 3 Clip Preview","image":true},"google/gemini-omni-flash-preview":{"name":"Gemini Omni Flash Preview","image":true},"google/veo-3.1-generate-preview":{"name":"Veo 3.1 Preview","image":true},"google/gemini-2.5-pro-tts":{"name":"Gemini 2.5 Pro TTS","image":false},"google/deep-research-max-preview-04-2026":{"name":"Deep Research Max Preview","image":true},"google/gemini-3-pro-image-preview":{"name":"Nano Banana Pro","image":true},"google/gemini-3.1-flash-lite-preview":{"name":"Gemini 3.1 Flash Lite Preview","image":true},"google/gemini-2.5-flash-tts":{"name":"Gemini 2.5 Flash TTS","imLine truncated
+{"zhipuai/glm-4.6v":{"name":"GLM-4.6V","image":true},"zhipuai/glm-5":{"name":"GLM-5","image":false},"zhipuai/glm-4.5-air":{"name":"GLM-4.5-Air","image":false},"zhipuai/glm-5.1":{"name":"GLM-5.1","image":false},"zhipuai/glm-4.7-flash":{"name":"GLM-4.7-Flash","image":false},"zhipuai/glm-5.2":{"name":"GLM-5.2","image":false},"zhipuai/glm-4.7-flashx":{"name":"GLM-4.7-FlashX","image":false},"zhipuai/glm-4.6":{"name":"GLM-4.6","image":false},"zhipuai/glm-4.5":{"name":"GLM-4.5","image":false},"zhipuai/glm-4.5v":{"name":"GLM-4.5V","image":true},"zhipuai/glm-4.7":{"name":"GLM-4.7","image":false},"zhipuai/glm-5-turbo":{"name":"GLM-5-Turbo","image":false},"zhipuai/glm-5v-turbo":{"name":"GLM-5V-Turbo","image":true},"zhipuai/glm-4.5-flash":{"name":"GLM-4.5-Flash","image":false},"microsoft/mai-code-1-flash":{"name":"MAI-Code-1-Flash","image":false},"cohere/command-r7b-arabic-02-2025":{"name":"Command R7B Arabic","image":false},"cohere/command-r-08-2024":{"name":"Command R","image":false},"cohere/command-a-plus-05-2026":{"name":"Command A Plus","image":true},"cohere/command-a-translate-08-2025":{"name":"Command A Translate","image":false},"cohere/c4ai-aya-expanse-8b":{"name":"Aya Expanse 8B","image":false},"cohere/c4ai-aya-vision-32b":{"name":"Aya Vision 32B","image":true},"cohere/command-a-03-2025":{"name":"Command A","image":false},"cohere/command-r-plus-08-2024":{"name":"Command R+","image":false},"cohere/command-a-reasoning-08-2025":{"name":"Command A Reasoning","image":false},"cohere/command-a-vision-07-2025":{"name":"Command A Vision","image":true},"cohere/command-r7b-12-2024":{"name":"Command R7B","image":false},"cohere/c4ai-aya-vision-8b":{"name":"Aya Vision 8B","image":true},"cohere/north-mini-code-1-0":{"name":"North Mini Code","image":false},"cohere/c4ai-aya-expanse-32b":{"name":"Aya Expanse 32B","image":false},"nvidia/mistral-nemotron":{"name":"Mistral Nemotron","image":false},"nvidia/nemotron-nano-12b-v2-vl":{"name":"Nemotron Nano 12B v2 VL","image":true},"nvidia/nemotron-3-nano-30b-a3b":{"name":"Nemotron 3 Nano 30B A3B","image":false},"nvidia/llama-3.3-nemotron-super-49b-v1.5":{"name":"Llama 3.3 Nemotron Super 49B v1.5","image":false},"nvidia/llama-3.1-nemotron-safety-guard-8b-v3":{"name":"Llama 3.1 Nemotron Safety Guard 8B v3","image":false},"nvidia/llama-3.3-nemotron-super-49b-v1":{"name":"Llama 3.3 Nemotron Super 49B v1","image":false},"nvidia/llama-3.1-nemotron-70b-instruct":{"name":"Llama 3.1 Nemotron 70B Instruct","image":false},"nvidia/nemotron-3-super-120b-a12b":{"name":"Nemotron 3 Super 120B A12B","image":false},"nvidia/nemotron-3-content-safety":{"name":"Nemotron 3 Content Safety","image":false},"nvidia/nemotron-3-nano-omni-30b-a3b-reasoning":{"name":"Nemotron 3 Nano Omni 30B A3B Reasoning","image":true},"nvidia/llama-3.1-nemotron-ultra-253b":{"name":"Llama 3.1 Nemotron Ultra 253B","image":false},"nvidia/nemotron-voicechat":{"name":"Nemotron VoiceChat","image":false},"nvidia/llama-nemotron-embed-vl-1b-v2":{"name":"Llama Nemotron Embed VL 1B v2","image":true},"nvidia/nemotron-content-safety-reasoning-4b":{"name":"Nemotron Content Safety Reasoning 4B","image":false},"nvidia/nemotron-3-ultra-550b-a55b":{"name":"Nemotron 3 Ultra 550B A55B","image":false},"nvidia/llama-nemotron-rerank-vl-1b-v2":{"name":"Llama Nemotron Rerank VL 1B v2","image":true},"nvidia/nemotron-mini-4b-instruct":{"name":"Nemotron Mini 4B Instruct","image":false},"nvidia/nemotron-nano-9b-v2":{"name":"Nemotron Nano 9B v2","image":false},"nvidia/nemotron-cascade-2-30b-a3b":{"name":"Nemotron Cascade 2 30B A3B","image":false},"nvidia/nemotron-3.5-content-safety":{"name":"Nemotron 3.5 Content Safety","image":true},"google/gemini-2.5-computer-use-preview-10-2025":{"name":"Gemini 2.5 Computer Use Preview","image":true},"google/deep-research-preview-04-2026":{"name":"Gemini Deep Research Preview","image":true},"google/gemini-3.1-flash-tts-preview":{"name":"Gemini 3.1 Flash TTS Preview","image":false},"google/gemini-flash-latest":{"name":"Gemini Flash Latest","image":true},"google/gemini-embedding-2":{"name":"Gemini Embedding 2","image":true},"google/lyria-3-pro-preview":{"name":"Lyria 3 Pro Preview","image":true},"google/gemini-3.5-flash":{"name":"Gemini 3.5 Flash","image":true},"google/gemini-2.5-flash":{"name":"Gemini 2.5 Flash","image":true},"google/gemini-3.5-flash-lite":{"name":"Gemini 3.5 Flash Lite","image":true},"google/lyria-3-clip-preview":{"name":"Lyria 3 Clip Preview","image":true},"google/gemini-omni-flash-preview":{"name":"Gemini Omni Flash Preview","image":true},"google/veo-3.1-generate-preview":{"name":"Veo 3.1 Preview","image":true},"google/gemini-2.5-pro-tts":{"name":"Gemini 2.5 Pro TTS","image":false},"google/deep-research-max-preview-04-2026":{"name":"Deep Research Max Preview","image":true},"google/gemini-3-pro-image-preview":{"name":"Nano Banana Pro","image":true},"google/gemini-3.1-flash-lite-preview":{"name":"Gemini 3.1 Flash Lite Preview","image":true},"google/gemini-2.5-flash-tts":{"name":"Gemini 2.5 Flash TTS","imLine truncated

package.json

@@ -1,6 +1,6 @@
 {
   "name": "kimi-eyes",
-  "version": "1.0.3",
+  "version": "1.0.4",
   "description": "Give non-multimodal models in Kimi Code the ability to analyze images and screenshots via a user-configured VLM (OpenAI-compatible or Anthropic protocol).",
   "keywords": [
     "kimi",