@mindstudio-ai/remy 0.1.173 → 0.1.175

package/dist/headless.js

@@ -2655,10 +2655,10 @@ var queryDatabaseTool = {
 
 // src/subagents/common/analyzeImage.ts
 var VISION_MODEL = "claude-4-6-sonnet";
-var VISION_MODEL_OVERRIDE = JSON.stringify({
+var VISION_MODEL_OVERRIDE = {
   model: VISION_MODEL,
   config: { thinkingBudget: "off" }
-});
+};
 async function analyzeImage(params) {
   const { prompt, imageUrl, timeout = 2e5, onLog } = params;
   return runCli(
@@ -3540,7 +3540,7 @@ var browserAutomationTool = {
   clearable: true,
   definition: {
     name: "runAutomatedBrowserTest",
-    description: "Run an automated browser test against the live preview. Describe what to test \u2014 the agent figures out how. Use after meaningful changes to frontend code, to reproduce user-reported issues, or to test end-to-end flows. Reports settled states between steps; it cannot reliably observe transient animation frames. For timing-sensitive bugs, read the source first. Never give it explicit values to use when filling out forms or creating accounts \u2014 it will use its own judgement (often it needs specific values to trigger dev-mode bypasses of things like login verification codes).",
+    description: "Run an automated browser test against the live preview. Describe what to test \u2014 the agent figures out how. Use after meaningful changes to frontend code, to reproduce user-reported issues, or to test end-to-end flows. Never give it explicit values to use when filling out forms or creating accounts \u2014 it will use its own judgement (often it needs specific values to trigger dev-mode bypasses of things like login verification codes).",
     inputSchema: {
       type: "object",
       properties: {
@@ -3616,7 +3616,8 @@ var browserAutomationTool = {
                     imageUrl: s.result.url,
                     prompt: buildScreenshotAnalysisPrompt({
                       styleMap: s.result.styleMap
-                    })
+                    }),
+                    visionModelOverride: VISION_MODEL_OVERRIDE
                   }
                 }));
                 const batchResult = await runCli(
@@ -3667,7 +3668,7 @@ var screenshotTool = {
   clearable: true,
   definition: {
     name: "screenshot",
-    description: "Capture a full-height screenshot of the app preview and get a description of what's on screen. Captures the settled page state \u2014 it cannot reliably catch animations, transitions, or transient state. For timing-sensitive bugs, read the source instead. Optionally provide specific questions about what you're looking for. Use a bulleted list to ask many questions at once. To ask additional questions about a screenshot you have already captured, pass its URL as imageUrl to skip recapture. If the screenshot requires interaction first (logging in, clicking a tab, dismissing a modal), use the instructions param to describe the steps.",
+    description: "Capture a full-height screenshot of the app preview and get a description of what's on screen. Captures the settled page state \u2014 it cannot catch animations, transitions, or transient state. Optionally provide specific questions about what you're looking for. Use a bulleted list to ask many questions at once. To ask additional questions about a screenshot you have already captured, pass its URL as imageUrl to skip recapture. If the screenshot requires interaction first (logging in, clicking a tab, dismissing a modal), use the instructions param to describe the steps.",
     inputSchema: {
       type: "object",
       properties: {
@@ -4102,10 +4103,10 @@ __export(generateImages_exports, {
 
 // src/subagents/designExpert/tools/images/enhancePrompt.ts
 var ENHANCE_MODEL = "claude-4-6-sonnet";
-var MODEL_OVERRIDE = JSON.stringify({
+var MODEL_OVERRIDE = {
   model: ENHANCE_MODEL,
   config: { reasoning: "false" }
-});
+};
 var SYSTEM_PROMPT = readAsset(
   "subagents/designExpert/tools/images/enhance-image-prompt.md"
 );
@@ -6857,7 +6858,14 @@ var HeadlessSession = class {
     if (requestId) {
       payload.requestId = requestId;
     }
-    process.stdout.write(JSON.stringify(payload) + "\n");
+    const line = JSON.stringify(payload) + "\n";
+    if (event === "history") {
+      log14.info("Wrote history event to stdout", {
+        requestId,
+        bytes: line.length
+      });
+    }
+    process.stdout.write(line);
   }
   /**
    * Emit a `completed` event and mark completedEmitted. Includes
@@ -7439,6 +7447,15 @@ var HeadlessSession = class {
         startIndex--;
       }
       const endIndex = before;
+      log14.info("History response", {
+        requestId,
+        startIndex,
+        endIndex,
+        count: endIndex - startIndex,
+        totalMessageCount: total,
+        beforeParam: rawBefore,
+        limitParam: rawLimit
+      });
       this.dispatchSimple(requestId, "history", () => ({
         messages: this.state.messages.slice(startIndex, endIndex),
         startIndex,

package/dist/index.js

@@ -2963,10 +2963,10 @@ var init_analyzeImage = __esm({
     "use strict";
     init_runCli();
     VISION_MODEL = "claude-4-6-sonnet";
-    VISION_MODEL_OVERRIDE = JSON.stringify({
+    VISION_MODEL_OVERRIDE = {
       model: VISION_MODEL,
       config: { thinkingBudget: "off" }
-    });
+    };
   }
 });
 
@@ -3890,13 +3890,14 @@ var init_browserAutomation = __esm({
     init_browserLock();
     init_screenshot();
     init_runCli();
+    init_analyzeImage();
     init_logger();
     log6 = createLogger("browser-automation");
     browserAutomationTool = {
       clearable: true,
       definition: {
         name: "runAutomatedBrowserTest",
-        description: "Run an automated browser test against the live preview. Describe what to test \u2014 the agent figures out how. Use after meaningful changes to frontend code, to reproduce user-reported issues, or to test end-to-end flows. Reports settled states between steps; it cannot reliably observe transient animation frames. For timing-sensitive bugs, read the source first. Never give it explicit values to use when filling out forms or creating accounts \u2014 it will use its own judgement (often it needs specific values to trigger dev-mode bypasses of things like login verification codes).",
+        description: "Run an automated browser test against the live preview. Describe what to test \u2014 the agent figures out how. Use after meaningful changes to frontend code, to reproduce user-reported issues, or to test end-to-end flows. Never give it explicit values to use when filling out forms or creating accounts \u2014 it will use its own judgement (often it needs specific values to trigger dev-mode bypasses of things like login verification codes).",
         inputSchema: {
           type: "object",
           properties: {
@@ -3972,7 +3973,8 @@ var init_browserAutomation = __esm({
                         imageUrl: s.result.url,
                         prompt: buildScreenshotAnalysisPrompt({
                           styleMap: s.result.styleMap
-                        })
+                        }),
+                        visionModelOverride: VISION_MODEL_OVERRIDE
                       }
                     }));
                     const batchResult = await runCli(
@@ -4032,7 +4034,7 @@ var init_screenshot2 = __esm({
       clearable: true,
       definition: {
         name: "screenshot",
-        description: "Capture a full-height screenshot of the app preview and get a description of what's on screen. Captures the settled page state \u2014 it cannot reliably catch animations, transitions, or transient state. For timing-sensitive bugs, read the source instead. Optionally provide specific questions about what you're looking for. Use a bulleted list to ask many questions at once. To ask additional questions about a screenshot you have already captured, pass its URL as imageUrl to skip recapture. If the screenshot requires interaction first (logging in, clicking a tab, dismissing a modal), use the instructions param to describe the steps.",
+        description: "Capture a full-height screenshot of the app preview and get a description of what's on screen. Captures the settled page state \u2014 it cannot catch animations, transitions, or transient state. Optionally provide specific questions about what you're looking for. Use a bulleted list to ask many questions at once. To ask additional questions about a screenshot you have already captured, pass its URL as imageUrl to skip recapture. If the screenshot requires interaction first (logging in, clicking a tab, dismissing a modal), use the instructions param to describe the steps.",
         inputSchema: {
           type: "object",
           properties: {
@@ -4549,10 +4551,10 @@ var init_enhancePrompt = __esm({
     init_runCli();
     init_assets();
     ENHANCE_MODEL = "claude-4-6-sonnet";
-    MODEL_OVERRIDE = JSON.stringify({
+    MODEL_OVERRIDE = {
       model: ENHANCE_MODEL,
       config: { reasoning: "false" }
-    });
+    };
     SYSTEM_PROMPT = readAsset(
       "subagents/designExpert/tools/images/enhance-image-prompt.md"
     );
@@ -7611,7 +7613,14 @@ var init_headless = __esm({
         if (requestId) {
           payload.requestId = requestId;
         }
-        process.stdout.write(JSON.stringify(payload) + "\n");
+        const line = JSON.stringify(payload) + "\n";
+        if (event === "history") {
+          log14.info("Wrote history event to stdout", {
+            requestId,
+            bytes: line.length
+          });
+        }
+        process.stdout.write(line);
       }
       /**
        * Emit a `completed` event and mark completedEmitted. Includes
@@ -8193,6 +8202,15 @@ var init_headless = __esm({
             startIndex--;
           }
           const endIndex = before;
+          log14.info("History response", {
+            requestId,
+            startIndex,
+            endIndex,
+            count: endIndex - startIndex,
+            totalMessageCount: total,
+            beforeParam: rawBefore,
+            limitParam: rawLimit
+          });
           this.dispatchSimple(requestId, "history", () => ({
             messages: this.state.messages.slice(startIndex, endIndex),
             startIndex,

package/dist/prompt/static/coding.md

@@ -7,16 +7,14 @@
 - Match the scope of changes to what was asked. Solve the current problem with the minimum code required. A bug fix is just a bug fix, not an opportunity to refactor the surrounding code. A new feature is just that feature, not a reason to introduce abstractions for hypothetical future needs. Prefer repeating a few lines of straightforward code over creating a helper that's only used once.
 
 ### Verification
-Run `lspDiagnostics` after every turn where you have edited code in any meaningful way. You don't need to run it for things like changing copy or CSS colors, but you should run it after any structural changes to code. It catches syntax errors, broken imports, and type mismatches instantly. After a big build or significant changes, also do a lightweight runtime check to catch the things static analysis misses (schema mismatches, missing imports, bad queries):
-
-- Spot-check methods with `runMethod`. The dev database is a disposable snapshot that will have been seeded with scenario data, so don't worry about being destructive.
-- For frontend work, take a single `screenshot` to confirm the main view renders correctly or look at the browser log for any console errors in the user's preview.
-- Use `runAutomatedBrowserTest` to verify an interactive flow that you can't confirm from a screenshot, when the user reports something broken that you can't identify from code alone, or whenever the verification involves driving the app through multiple interactions.
-- If the browser is unavailable, skip the visual check and verify through methods, logs, and code instead. Browser unavailability is an infrastructure issue, not a code problem — don't try to diagnose or fix it.
+Run `lspDiagnostics` after every turn where you have edited code in any meaningful way. You don't need to run it for things like changing copy or CSS colors, but you should run it after any structural changes to code. It catches syntax errors, broken imports, and type mismatches instantly. After a big build or significant changes, also do a lightweight runtime check to catch the things static analysis misses (schema mismatches, missing imports, bad queries). Your runtime check can include:
+- Spot-checking methods with `runMethod`. The dev database is a disposable snapshot that will have been seeded with scenario data, so don't worry about being destructive.
+- For frontend work, taking a `screenshot` to confirm the main view renders correctly or look at the browser log for any console errors in the user's preview.
+- Using `runAutomatedBrowserTest` to verify an interactive flow that you can't confirm from a screenshot, when the user reports something broken that you can't identify from code alone, or whenever the verification involves driving the app through multiple interactions.
 
 Aim for confidence that the core happy paths work. If the 80% case is solid, the remaining edge cases are likely fine and the user can surface them in chat. Don't screenshot every page, test every permutation, or verify every secondary flow. One or two runtime checks that confirm the app loads and data flows through is enough.
 
-When making mechanical edits as part of iterating with the user (e.g., moving elements, changing labels, small redesigns and refactors), don't re-screenshot to confirm, simply trust your code. Re-screenshot only when changes are structural enough that the visual outcome is genuinely uncertain (new layout, new component composition, new route), or when the user reports something visible that you can't see in the code.
+When making mechanical edits as part of iterating with the user (e.g., moving elements, changing labels, small redesigns and refactors), don't screenshot to confirm, simply trust your code. Re-screenshot only when changes are structural enough that the visual outcome is genuinely uncertain (new layout, new component composition, new route), or when the user reports something visible that you can't see in the code. The screenshot tool captures static/settled state - don't try to hack it with different instructions to capture transient states or animations or things like that. If what you need is not avaialble via screenshot, fall back to static analysis by tracing code.
 
 ### Process Logs
 

package/dist/subagents/browserAutomation/prompt.md

@@ -31,6 +31,8 @@ paragraph "No results found"
 
 Each interactive element has a `[ref=eN]` you can use to target it.
 
+Note: the snapshot concatenates inline text and strips whitespace. If you need to verify spacing or pixel-level layout, use evaluate to run a script to get values. The snapshot is intended to help you understand page layout and target elements for interaction, not as an authoritative report of how the page renders.
+
 ### Commands
 
 - `snapshot`: Get the current page state. Always do this first and after action batches to verify results. Waits for network requests to settle.
@@ -149,7 +151,7 @@ You can use the `screenshotFullPage` tool to take a full-height screenshot of th
   - evaluate auto-returns simple expressions. `"script": "document.title"` works directly. For multi-statement scripts, use explicit return.
   - The snapshot in the response is always the most current page state. Even if a wait times out, check the snapshot field; the content you were waiting for may have appeared by then.
   - Execution stops on first error. If step 2 of 5 fails, steps 3-5 don't run. The response will contain results for steps 0-2 (with step 2 having an error field) plus the current snapshot. Adjust and retry from the failed step.
-  - If something fails, bail early. Do not attempt to diagnose why; do not do things like attempt different inputs to try to work around an error - just report the failure and early return.
+  - If something fails, bail early. Do not attempt to diagnose why; do not do things like attempt different inputs to try to work around an error - just report the failure and early return. If something is not visible or obvious (e.g., animations, transient states, etc), do not attempt to hack the browser commands into figuring it out - simply accept that the browser automation is limited in this regard and can not reliably reproduce the state required to test.
 </rules>
 
 <voice>

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@mindstudio-ai/remy",
-  "version": "0.1.173",
+  "version": "0.1.175",
   "description": "MindStudio coding agent",
   "repository": {
     "type": "git",