@mindstudio-ai/remy 0.1.27 → 0.1.29

package/dist/headless.js

@@ -2076,16 +2076,16 @@ var runMethodTool = {
 var SCREENSHOT_ANALYSIS_PROMPT = "Describe everything visible on screen from top to bottom \u2014 every element, its position, its size relative to the viewport, its colors, its content. Be thorough and spatial. After the inventory, note anything that looks visually broken (overlapping elements, clipped text, misaligned components).";
 async function captureAndAnalyzeScreenshot(promptOrOptions) {
   let prompt;
-  let viewportOnly = false;
+  let fullPage = false;
   if (typeof promptOrOptions === "object" && promptOrOptions !== null) {
     prompt = promptOrOptions.prompt;
-    viewportOnly = promptOrOptions.viewportOnly ?? false;
+    fullPage = promptOrOptions.fullPage ?? false;
   } else {
     prompt = promptOrOptions;
   }
   const ssResult = await sidecarRequest(
     "/screenshot",
-    { fullPage: !viewportOnly },
+    { fullPage },
     { timeout: 12e4 }
   );
   log.debug("Screenshot response", { ssResult });
@@ -2109,7 +2109,7 @@ async function captureAndAnalyzeScreenshot(promptOrOptions) {
 var screenshotTool = {
   definition: {
     name: "screenshot",
-    description: "Capture a screenshot of the app preview and get a description of what's on screen. Optionally provide a specific question about what you're looking for. Set viewportOnly to capture just what the user sees on screen.",
+    description: "Capture a screenshot of the app preview and get a description of what's on screen. Optionally provide a specific question about what you're looking for. By default captures the viewport (what the user sees). Set fullPage to capture the entire scrollable page.",
     inputSchema: {
       type: "object",
       properties: {
@@ -2117,9 +2117,9 @@ var screenshotTool = {
           type: "string",
           description: "Optional question about the screenshot. If omitted, returns a general description of what's visible."
         },
-        viewportOnly: {
+        fullPage: {
           type: "boolean",
-          description: "Capture only the visible viewport instead of the full scrollable page. Use when checking above-the-fold layout or viewport-relative sizing like 100vh."
+          description: "Capture the full scrollable page instead of just the viewport. Use when you need to see below-the-fold content."
         }
       }
     }
@@ -2128,7 +2128,7 @@ var screenshotTool = {
     try {
       return await captureAndAnalyzeScreenshot({
         prompt: input.prompt,
-        viewportOnly: input.viewportOnly
+        fullPage: input.fullPage
       });
     } catch (err) {
       return `Error taking screenshot: ${err.message}`;
@@ -2294,7 +2294,7 @@ Current date/time: ${(/* @__PURE__ */ new Date()).toISOString().replace("T", " "
           if (externalTools.has(tc.name) && resolveExternalTool) {
             result = await resolveExternalTool(tc.id, tc.name, tc.input);
           } else {
-            result = await executeTool2(tc.name, tc.input);
+            result = await executeTool2(tc.name, tc.input, tc.id);
           }
           const isError = result.startsWith("Error");
           emit2({
@@ -2629,7 +2629,7 @@ var DESIGN_EXPERT_TOOLS = [
   },
   {
     name: "screenshot",
-    description: "Capture a screenshot of the app preview. Returns a CDN URL with visual analysis. Use to review the current state of the UI being built. Set viewportOnly to capture just what the user sees on screen.",
+    description: "Capture a screenshot of the app preview. Returns a CDN URL with visual analysis. Use to review the current state of the UI being built. By default captures the viewport. Set fullPage to capture the entire scrollable page.",
     inputSchema: {
       type: "object",
       properties: {
@@ -2637,22 +2637,22 @@ var DESIGN_EXPERT_TOOLS = [
           type: "string",
           description: "Optional specific question about the screenshot."
         },
-        viewportOnly: {
+        fullPage: {
           type: "boolean",
-          description: "Capture only the visible viewport instead of the full scrollable page. Use when checking above-the-fold layout or viewport-relative sizing like 100vh."
+          description: "Capture the full scrollable page instead of just the viewport. Use when you need to see below-the-fold content."
         }
       }
     }
   },
   {
     name: "runBrowserTest",
-    description: "Run an automated browser test against the live app preview. Use to verify visual implementation: check computed styles, navigate between pages, take analyzed screenshots. Describe what you want to verify and the browser agent handles the interaction.",
+    description: "Run an automated browser test against the live app preview. Use to verify implementation details via getComputedStyle: font-family names, exact colors, spacing, borders, shadows, font sizes, transforms. Also supports navigation between pages and screenshots. Use this to confirm the right fonts are loaded and CSS values match the spec.",
     inputSchema: {
       type: "object",
       properties: {
         task: {
           type: "string",
-          description: 'What to verify, in natural language. E.g., "Check that the hero section cards have border-radius: 24px and the correct rotation angles" or "Navigate to /about and screenshot it".'
+          description: 'What to verify, in natural language. Focus on measurable properties: "Check the hero cards have border-radius: 24px and box-shadow" or "Verify the background color of the CTA section is #C4FF0D".'
         }
       },
       required: ["task"]
@@ -2684,13 +2684,13 @@ var DESIGN_EXPERT_TOOLS = [
     }
   }
 ];
-async function executeDesignExpertTool(name, input, context) {
+async function executeDesignExpertTool(name, input, context, toolCallId) {
   switch (name) {
     case "screenshot": {
       try {
         return await captureAndAnalyzeScreenshot({
           prompt: input.prompt,
-          viewportOnly: input.viewportOnly
+          fullPage: input.fullPage
         });
       } catch (err) {
         return `Error taking screenshot: ${err.message}`;
@@ -2790,7 +2790,13 @@ ${analysis}`;
       if (!context) {
         return "Error: browser testing requires execution context (only available in headless mode)";
       }
-      return browserAutomationTool.execute({ task: input.task }, context);
+      return browserAutomationTool.execute(
+        { task: input.task },
+        {
+          ...context,
+          toolCallId: toolCallId || context.toolCallId
+        }
+      );
     }
     default:
       return `Error: unknown tool "${name}"`;
@@ -3020,7 +3026,7 @@ var designExpertTool = {
       task: input.task,
       tools: DESIGN_EXPERT_TOOLS,
       externalTools: /* @__PURE__ */ new Set(),
-      executeTool: (name, input2) => executeDesignExpertTool(name, input2, context),
+      executeTool: (name, input2, toolCallId) => executeDesignExpertTool(name, input2, context, toolCallId),
       apiConfig: context.apiConfig,
       model: context.model,
       subAgentId: "visualDesignExpert",

package/dist/index.js

@@ -2017,16 +2017,16 @@ var init_runMethod = __esm({
 // src/tools/_helpers/screenshot.ts
 async function captureAndAnalyzeScreenshot(promptOrOptions) {
   let prompt;
-  let viewportOnly = false;
+  let fullPage = false;
   if (typeof promptOrOptions === "object" && promptOrOptions !== null) {
     prompt = promptOrOptions.prompt;
-    viewportOnly = promptOrOptions.viewportOnly ?? false;
+    fullPage = promptOrOptions.fullPage ?? false;
   } else {
     prompt = promptOrOptions;
   }
   const ssResult = await sidecarRequest(
     "/screenshot",
-    { fullPage: !viewportOnly },
+    { fullPage },
     { timeout: 12e4 }
   );
   log.debug("Screenshot response", { ssResult });
@@ -2065,7 +2065,7 @@ var init_screenshot2 = __esm({
     screenshotTool = {
       definition: {
         name: "screenshot",
-        description: "Capture a screenshot of the app preview and get a description of what's on screen. Optionally provide a specific question about what you're looking for. Set viewportOnly to capture just what the user sees on screen.",
+        description: "Capture a screenshot of the app preview and get a description of what's on screen. Optionally provide a specific question about what you're looking for. By default captures the viewport (what the user sees). Set fullPage to capture the entire scrollable page.",
         inputSchema: {
           type: "object",
           properties: {
@@ -2073,9 +2073,9 @@ var init_screenshot2 = __esm({
               type: "string",
               description: "Optional question about the screenshot. If omitted, returns a general description of what's visible."
             },
-            viewportOnly: {
+            fullPage: {
               type: "boolean",
-              description: "Capture only the visible viewport instead of the full scrollable page. Use when checking above-the-fold layout or viewport-relative sizing like 100vh."
+              description: "Capture the full scrollable page instead of just the viewport. Use when you need to see below-the-fold content."
             }
           }
         }
@@ -2084,7 +2084,7 @@ var init_screenshot2 = __esm({
         try {
           return await captureAndAnalyzeScreenshot({
             prompt: input.prompt,
-            viewportOnly: input.viewportOnly
+            fullPage: input.fullPage
           });
         } catch (err) {
           return `Error taking screenshot: ${err.message}`;
@@ -2257,7 +2257,7 @@ Current date/time: ${(/* @__PURE__ */ new Date()).toISOString().replace("T", " "
           if (externalTools.has(tc.name) && resolveExternalTool) {
             result = await resolveExternalTool(tc.id, tc.name, tc.input);
           } else {
-            result = await executeTool2(tc.name, tc.input);
+            result = await executeTool2(tc.name, tc.input, tc.id);
           }
           const isError = result.startsWith("Error");
           emit2({
@@ -2570,13 +2570,13 @@ function resolvePath(filename) {
   const local4 = path5.join(base2, filename);
   return fs11.existsSync(local4) ? local4 : path5.join(base2, "subagents", "designExpert", filename);
 }
-async function executeDesignExpertTool(name, input, context) {
+async function executeDesignExpertTool(name, input, context, toolCallId) {
   switch (name) {
     case "screenshot": {
       try {
         return await captureAndAnalyzeScreenshot({
           prompt: input.prompt,
-          viewportOnly: input.viewportOnly
+          fullPage: input.fullPage
         });
       } catch (err) {
         return `Error taking screenshot: ${err.message}`;
@@ -2676,7 +2676,13 @@ ${analysis}`;
       if (!context) {
         return "Error: browser testing requires execution context (only available in headless mode)";
       }
-      return browserAutomationTool.execute({ task: input.task }, context);
+      return browserAutomationTool.execute(
+        { task: input.task },
+        {
+          ...context,
+          toolCallId: toolCallId || context.toolCallId
+        }
+      );
     }
     default:
       return `Error: unknown tool "${name}"`;
@@ -2744,7 +2750,7 @@ var init_tools2 = __esm({
       },
       {
         name: "screenshot",
-        description: "Capture a screenshot of the app preview. Returns a CDN URL with visual analysis. Use to review the current state of the UI being built. Set viewportOnly to capture just what the user sees on screen.",
+        description: "Capture a screenshot of the app preview. Returns a CDN URL with visual analysis. Use to review the current state of the UI being built. By default captures the viewport. Set fullPage to capture the entire scrollable page.",
         inputSchema: {
           type: "object",
           properties: {
@@ -2752,22 +2758,22 @@ var init_tools2 = __esm({
               type: "string",
               description: "Optional specific question about the screenshot."
             },
-            viewportOnly: {
+            fullPage: {
               type: "boolean",
-              description: "Capture only the visible viewport instead of the full scrollable page. Use when checking above-the-fold layout or viewport-relative sizing like 100vh."
+              description: "Capture the full scrollable page instead of just the viewport. Use when you need to see below-the-fold content."
             }
           }
         }
       },
       {
         name: "runBrowserTest",
-        description: "Run an automated browser test against the live app preview. Use to verify visual implementation: check computed styles, navigate between pages, take analyzed screenshots. Describe what you want to verify and the browser agent handles the interaction.",
+        description: "Run an automated browser test against the live app preview. Use to verify implementation details via getComputedStyle: font-family names, exact colors, spacing, borders, shadows, font sizes, transforms. Also supports navigation between pages and screenshots. Use this to confirm the right fonts are loaded and CSS values match the spec.",
         inputSchema: {
           type: "object",
           properties: {
             task: {
               type: "string",
-              description: 'What to verify, in natural language. E.g., "Check that the hero section cards have border-radius: 24px and the correct rotation angles" or "Navigate to /about and screenshot it".'
+              description: 'What to verify, in natural language. Focus on measurable properties: "Check the hero cards have border-radius: 24px and box-shadow" or "Verify the background color of the CTA section is #C4FF0D".'
             }
           },
           required: ["task"]
@@ -3042,7 +3048,7 @@ Visual design expert. Describe the situation and what you need \u2014 the agent
           task: input.task,
           tools: DESIGN_EXPERT_TOOLS,
           externalTools: /* @__PURE__ */ new Set(),
-          executeTool: (name, input2) => executeDesignExpertTool(name, input2, context),
+          executeTool: (name, input2, toolCallId) => executeDesignExpertTool(name, input2, context, toolCallId),
           apiConfig: context.apiConfig,
           model: context.model,
           subAgentId: "visualDesignExpert",

package/dist/subagents/codeSanityCheck/prompt.md

@@ -18,6 +18,10 @@ These are things we already know about and have decided to accept:
 
 - Limited browser support for `oklch` gradients using `in <colorspace>` syntax — we accept the compatibility tradeoff for better color quality
 - Limited browser support for CSS scroll-driven animations (`animation-timeline: scroll()` / `view()`)  - we accept this tradeoff
+- Libraries we know are actively maintained, don't bother checking:
+  - swr
+  - framer-motion
+  - styled-components
 
 ### Common pitfalls (always flag these)
 

package/dist/subagents/designExpert/prompts/instructions.md

@@ -1,7 +1,9 @@
 ## Tool usage
 
 - Use `analyzeReferenceImageOrUrl` to analyze any image URL or website URL. Websites are automatically screenshotted. Omit the prompt for a standard design analysis, or provide a custom prompt for specific questions. Do not screenshot font specimen pages, documentation, or other text-heavy pages — use `fetchUrl` for those instead.
-- Use `screenshot` to capture the current state of the app preview. Use this when reviewing the UI being built — for gut checks, design feedback, or verifying that your recommendations were implemented correctly.
+- Use `screenshot` to see the current state of the app preview. This is your primary tool for visual review. Use `fullPage: true` to see the entire page at once. Remember, the screenshot analysis is not overly precise - for example, it cannot reliably identify specific fonts by name — it can only describe what letterforms look like.
+- Use `runBrowserTest` only to verify specific computed values that a screenshot can't tell you: exact font-family names, hex colors, pixel measurements, border-radius, box-shadow, transforms. Keep requests focused — ask it to check specific properties on specific elements, not to take screenshots or scroll around.
+- **screenshot vs runBrowserTest**: Screenshot to *see* the page. Browser test to *measure* specific CSS values. Never use the browser agent for tasks your screenshot tool can handle, or vice versa.
 - Use `searchGoogle` and `fetchUrl` only when the user references something specific: a particular website to match, a brand to look up, a company whose identity you need to research. You already have curated fonts, inspiration references, and strong internal knowledge — don't search the web for generic inspiration or "best X apps." The web is for specific lookups, not creative direction.
 - When proposing multiple options, make them genuinely different directions (dark + bold vs. light + editorial) rather than minor variations.
 - When multiple tool calls are independent, make them all in a single turn. Searching for three different products, or fetching two reference sites: batch them instead of doing one per turn.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@mindstudio-ai/remy",
-  "version": "0.1.27",
+  "version": "0.1.29",
   "description": "MindStudio coding agent",
   "repository": {
     "type": "git",