@mindstudio-ai/remy 0.1.43 → 0.1.45

package/dist/headless.js

@@ -429,11 +429,10 @@ async function* streamChat(params) {
   const { baseUrl: baseUrl2, apiKey, signal, requestId, ...body } = params;
   const url = `${baseUrl2}/_internal/v2/agent/remy/chat`;
   const startTime = Date.now();
-  const messagesWithAttachments = body.messages.filter(
-    (m) => m.attachments && m.attachments.length > 0
-  );
+  const subAgentId = body.subAgentId;
   log3.info("API request", {
     requestId,
+    ...subAgentId && { subAgentId },
     model: body.model,
     messageCount: body.messages.length,
     toolCount: body.tools.length
@@ -451,15 +450,27 @@ async function* streamChat(params) {
     });
   } catch (err) {
     if (signal?.aborted) {
-      log3.warn("Request aborted", { requestId });
+      log3.warn("Request aborted", {
+        requestId,
+        ...subAgentId && { subAgentId }
+      });
       throw err;
     }
-    log3.error("Network error", { requestId, error: err.message });
+    log3.error("Network error", {
+      requestId,
+      ...subAgentId && { subAgentId },
+      error: err.message
+    });
     yield { type: "error", error: `Network error: ${err.message}` };
     return;
   }
   const ttfb = Date.now() - startTime;
-  log3.info("API response", { requestId, status: res.status, ttfbMs: ttfb });
+  log3.info("API response", {
+    requestId,
+    ...subAgentId && { subAgentId },
+    status: res.status,
+    ttfbMs: ttfb
+  });
   if (!res.ok) {
     let errorMessage = `HTTP ${res.status}`;
     try {
@@ -474,6 +485,7 @@ async function* streamChat(params) {
     }
     log3.error("API error", {
       requestId,
+      ...subAgentId && { subAgentId },
       status: res.status,
       error: errorMessage
     });
@@ -503,6 +515,7 @@ async function* streamChat(params) {
       await reader.cancel();
       log3.error("Stream stalled", {
         requestId,
+        ...subAgentId && { subAgentId },
         durationMs: Date.now() - startTime
       });
       yield {
@@ -528,6 +541,7 @@ async function* streamChat(params) {
           const elapsed = Date.now() - startTime;
           log3.info("Stream complete", {
             requestId,
+            ...subAgentId && { subAgentId },
             durationMs: elapsed,
             stopReason: event.stopReason,
             inputTokens: event.usage.inputTokens,
@@ -2225,7 +2239,7 @@ function startStatusWatcher(config) {
     inflight = true;
     try {
       const ctx = getContext();
-      if (!ctx.assistantText && !ctx.lastToolName) {
+      if (!ctx.assistantText && !ctx.lastToolName && !ctx.userMessage) {
         return;
       }
       const res = await fetch(url, {
@@ -2368,7 +2382,8 @@ ${partial}` : "[INTERRUPTED] Agent was interrupted before producing output.",
         getContext: () => ({
           assistantText: getPartialText(contentBlocks),
           lastToolName: currentToolNames || void 0,
-          lastToolResult: lastToolResult || void 0
+          lastToolResult: lastToolResult || void 0,
+          userMessage: task
         }),
         onStatus: (label) => emit2({ type: "status", message: label }),
         signal
@@ -2539,9 +2554,18 @@ Current date/time: ${(/* @__PURE__ */ new Date()).toISOString().replace("T", " "
             }
           };
           toolRegistry?.register(entry);
+          const toolStart = Date.now();
           run2(tc.input);
           const r = await resultPromise;
           toolRegistry?.unregister(tc.id);
+          log4.info("Tool completed", {
+            requestId,
+            parentToolId,
+            toolCallId: tc.id,
+            name: tc.name,
+            durationMs: Date.now() - toolStart,
+            isError: r.isError
+          });
           emit2({
             type: "tool_done",
             id: tc.id,
@@ -2597,14 +2621,30 @@ Current date/time: ${(/* @__PURE__ */ new Date()).toISOString().replace("T", " "
     return wrapRun();
   }
   log4.info("Sub-agent backgrounded", { requestId, parentToolId, agentName });
+  toolRegistry?.register({
+    id: parentToolId,
+    name: agentName,
+    input: { task },
+    abortController: bgAbort,
+    startedAt: Date.now(),
+    settle: () => {
+    },
+    rerun: () => {
+    },
+    getPartialResult: () => ""
+  });
   const ack = await generateBackgroundAck({
     apiConfig,
     agentName: subAgentId || "agent",
     task
   });
-  wrapRun().then((finalResult) => onBackgroundComplete?.(finalResult)).catch(
-    (err) => onBackgroundComplete?.({ text: `Error: ${err.message}`, messages: [] })
-  );
+  wrapRun().then((finalResult) => {
+    toolRegistry?.unregister(parentToolId);
+    onBackgroundComplete?.(finalResult);
+  }).catch((err) => {
+    toolRegistry?.unregister(parentToolId);
+    onBackgroundComplete?.({ text: `Error: ${err.message}`, messages: [] });
+  });
   return { text: ack, messages: [], backgrounded: true };
 }
 
@@ -3053,15 +3093,18 @@ async function enhanceImagePrompt(params) {
       "Transparent background: yes \u2014 the background will be removed. Focus on the subject as an isolated element."
     );
   }
-  const message = `<context>
+  const context = `<context>
 ${contextParts.join("\n")}
-</context>
+</context>`;
+  const message = `${SYSTEM_PROMPT}
+
+${context}
 
 <brief>
 ${brief}
 </brief>`;
   const enhanced = await runCli(
-    `mindstudio generate-text --prompt ${JSON.stringify(SYSTEM_PROMPT)} --message ${JSON.stringify(message)} --output-key enhanced --no-meta`,
+    `mindstudio generate-text --message ${JSON.stringify(message)} --output-key content --no-meta`,
     { timeout: 6e4, onLog }
   );
   return enhanced.trim();
@@ -4738,7 +4781,9 @@ async function runTurn(params) {
         toolRegistry?.register(entry);
         run(tc.input);
         const r = await resultPromise;
-        toolRegistry?.unregister(tc.id);
+        if (!tc.input.background) {
+          toolRegistry?.unregister(tc.id);
+        }
         log7.info("Tool completed", {
           requestId,
           toolCallId: tc.id,

package/dist/index.js

@@ -89,11 +89,10 @@ async function* streamChat(params) {
   const { baseUrl: baseUrl2, apiKey, signal, requestId, ...body } = params;
   const url = `${baseUrl2}/_internal/v2/agent/remy/chat`;
   const startTime = Date.now();
-  const messagesWithAttachments = body.messages.filter(
-    (m) => m.attachments && m.attachments.length > 0
-  );
+  const subAgentId = body.subAgentId;
   log.info("API request", {
     requestId,
+    ...subAgentId && { subAgentId },
     model: body.model,
     messageCount: body.messages.length,
     toolCount: body.tools.length
@@ -111,15 +110,27 @@ async function* streamChat(params) {
     });
   } catch (err) {
     if (signal?.aborted) {
-      log.warn("Request aborted", { requestId });
+      log.warn("Request aborted", {
+        requestId,
+        ...subAgentId && { subAgentId }
+      });
       throw err;
     }
-    log.error("Network error", { requestId, error: err.message });
+    log.error("Network error", {
+      requestId,
+      ...subAgentId && { subAgentId },
+      error: err.message
+    });
     yield { type: "error", error: `Network error: ${err.message}` };
     return;
   }
   const ttfb = Date.now() - startTime;
-  log.info("API response", { requestId, status: res.status, ttfbMs: ttfb });
+  log.info("API response", {
+    requestId,
+    ...subAgentId && { subAgentId },
+    status: res.status,
+    ttfbMs: ttfb
+  });
   if (!res.ok) {
     let errorMessage = `HTTP ${res.status}`;
     try {
@@ -134,6 +145,7 @@ async function* streamChat(params) {
     }
     log.error("API error", {
       requestId,
+      ...subAgentId && { subAgentId },
       status: res.status,
       error: errorMessage
     });
@@ -163,6 +175,7 @@ async function* streamChat(params) {
       await reader.cancel();
       log.error("Stream stalled", {
         requestId,
+        ...subAgentId && { subAgentId },
         durationMs: Date.now() - startTime
       });
       yield {
@@ -188,6 +201,7 @@ async function* streamChat(params) {
           const elapsed = Date.now() - startTime;
           log.info("Stream complete", {
             requestId,
+            ...subAgentId && { subAgentId },
             durationMs: elapsed,
             stopReason: event.stopReason,
             inputTokens: event.usage.inputTokens,
@@ -2163,7 +2177,7 @@ function startStatusWatcher(config) {
     inflight = true;
     try {
       const ctx = getContext();
-      if (!ctx.assistantText && !ctx.lastToolName) {
+      if (!ctx.assistantText && !ctx.lastToolName && !ctx.userMessage) {
         return;
       }
       const res = await fetch(url, {
@@ -2315,7 +2329,8 @@ ${partial}` : "[INTERRUPTED] Agent was interrupted before producing output.",
         getContext: () => ({
           assistantText: getPartialText(contentBlocks),
           lastToolName: currentToolNames || void 0,
-          lastToolResult: lastToolResult || void 0
+          lastToolResult: lastToolResult || void 0,
+          userMessage: task
         }),
         onStatus: (label) => emit2({ type: "status", message: label }),
         signal
@@ -2486,9 +2501,18 @@ Current date/time: ${(/* @__PURE__ */ new Date()).toISOString().replace("T", " "
             }
           };
           toolRegistry?.register(entry);
+          const toolStart = Date.now();
           run2(tc.input);
           const r = await resultPromise;
           toolRegistry?.unregister(tc.id);
+          log3.info("Tool completed", {
+            requestId,
+            parentToolId,
+            toolCallId: tc.id,
+            name: tc.name,
+            durationMs: Date.now() - toolStart,
+            isError: r.isError
+          });
           emit2({
             type: "tool_done",
             id: tc.id,
@@ -2544,14 +2568,30 @@ Current date/time: ${(/* @__PURE__ */ new Date()).toISOString().replace("T", " "
     return wrapRun();
   }
   log3.info("Sub-agent backgrounded", { requestId, parentToolId, agentName });
+  toolRegistry?.register({
+    id: parentToolId,
+    name: agentName,
+    input: { task },
+    abortController: bgAbort,
+    startedAt: Date.now(),
+    settle: () => {
+    },
+    rerun: () => {
+    },
+    getPartialResult: () => ""
+  });
   const ack = await generateBackgroundAck({
     apiConfig,
     agentName: subAgentId || "agent",
     task
   });
-  wrapRun().then((finalResult) => onBackgroundComplete?.(finalResult)).catch(
-    (err) => onBackgroundComplete?.({ text: `Error: ${err.message}`, messages: [] })
-  );
+  wrapRun().then((finalResult) => {
+    toolRegistry?.unregister(parentToolId);
+    onBackgroundComplete?.(finalResult);
+  }).catch((err) => {
+    toolRegistry?.unregister(parentToolId);
+    onBackgroundComplete?.({ text: `Error: ${err.message}`, messages: [] });
+  });
   return { text: ack, messages: [], backgrounded: true };
 }
 var log3;
@@ -3106,15 +3146,18 @@ async function enhanceImagePrompt(params) {
       "Transparent background: yes \u2014 the background will be removed. Focus on the subject as an isolated element."
     );
   }
-  const message = `<context>
+  const context = `<context>
 ${contextParts.join("\n")}
-</context>
+</context>`;
+  const message = `${SYSTEM_PROMPT}
+
+${context}
 
 <brief>
 ${brief}
 </brief>`;
   const enhanced = await runCli(
-    `mindstudio generate-text --prompt ${JSON.stringify(SYSTEM_PROMPT)} --message ${JSON.stringify(message)} --output-key enhanced --no-meta`,
+    `mindstudio generate-text --message ${JSON.stringify(message)} --output-key content --no-meta`,
     { timeout: 6e4, onLog }
   );
   return enhanced.trim();
@@ -4980,7 +5023,9 @@ async function runTurn(params) {
         toolRegistry?.register(entry);
         run(tc.input);
         const r = await resultPromise;
-        toolRegistry?.unregister(tc.id);
+        if (!tc.input.background) {
+          toolRegistry?.unregister(tc.id);
+        }
         log6.info("Tool completed", {
           requestId,
           toolCallId: tc.id,

package/dist/prompt/compiled/design.md

@@ -55,6 +55,7 @@ Forms should feel like interactions, not paperwork.
 - Inline validation — show errors as the user types, not after submit. Validation must never introduce layout shift.
 - Loading states after submission. Always indicate that something is happening.
 - Disabled states should be visually distinct but not jarring.
+- Media uploads should optimistically load in a local preview of an image or video and show upload progress
 - Even data entry can be beautiful. Pay attention to alignment, padding, and spacing. Consistency is key.
 
 #### Form Elements

package/dist/prompt/compiled/interfaces.md

@@ -54,9 +54,21 @@ const api = createClient<{
 const { vendorId } = await api.submitVendorRequest({ name: 'Acme' });
 const { vendors } = await api.listVendors();
 
-// File operations
-const { url } = await platform.requestFile({ type: 'image' });
-const cdnUrl = await platform.uploadFile(file);
+// File upload (returns CDN URL)
+const url = await platform.uploadFile(file);
+
+// With progress tracking
+const url = await platform.uploadFile(file, {
+  onProgress: (fraction) => setProgress(fraction), // 0 to 1
+});
+
+// With abort support
+const controller = new AbortController();
+const url = await platform.uploadFile(file, {
+  signal: controller.signal,
+  onProgress: (f) => setProgress(f),
+});
+controller.abort(); // cancels the upload
 
 // Current user (display only)
 auth.userId;

package/dist/prompt/static/coding.md

@@ -16,11 +16,14 @@ After editing code, check your work with `lspDiagnostics` or by reading the file
 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.
 
 ### Process Logs
-Process logs are available at `.logs/` for debugging:
-  - `.logs/tunnel.log`: method execution, schema sync, session lifecycle, platform connection
-  - `.logs/devServer.log`: frontend build errors, HMR, module resolution failures
-  - `.logs/requests.ndjson`: structured NDJSON log of every method and scenario execution with full input, output, errors (including stack traces), console output, and duration. Use `tail -5 .logs/requests.ndjson | jq .` or `grep '"success":false' .logs/requests.ndjson | jq .` to inspect.
-  - `.logs/browser.ndjson`: browser-side events captured from the web preview. Includes console output, uncaught JS errors with stack traces, failed network requests, and user interactions (clicks). Use `grep '"type":"error"' .logs/browser.ndjson | jq .` to find frontend errors.
+
+Process logs are available at .logs/ in NDJSON format (one JSON object per line) for debugging. Each line has at minimum ts (unix millis) and msg fields, plus structured context like level, module, requestId, toolCallId where available. You can use `jq` to examine logs and debug failures. Tools like run method or run scenario execute synchronously, so log data will be available by the time those tools return their results to you, there is no need to `sleep` before querying logfiles.
+  - `.logs/tunnel.ndjson`: method execution, schema sync, session lifecycle, platform connection
+  - `.logs/devServer.ndjson`: frontend build errors, HMR, module resolution failures
+  - `.logs/system.ndjson`: sandbox server logs — agent lifecycle, tool dispatch, file watching, process management
+  - `.logs/agent.ndjson`: coding agent protocol events and errors
+  - `.logs/requests.ndjson`: structured log of every method and scenario execution with full input, output, errors (including stack traces), console output, and duration
+  - `.logs/browser.ndjson`: browser-side events from the web preview — console output, uncaught JS errors with stack traces, failed network requests, user interactions
 
 ### MindStudio SDK
 For any work involving AI models, external actions (web scraping, email, SMS), or third-party API/OAuth connections, prefer the `@mindstudio-ai/agent` SDK. It removes the need to research API methods, configure keys and tokens, or require the user to set up developer accounts.

package/dist/prompt/static/team.md

@@ -10,9 +10,9 @@ Note: when you talk about the team to the user, refer to them by their name or a
 
 Your designer. Consult for any visual decision — choosing a color, picking fonts, proposing a layout, soucing images, reviewing whether something looks good. Not just during intake or big design moments. If you're about to write CSS and you're not sure about a color, ask. If you just built a page and want a gut check, ask the designer to take a quick look. If the user says "I don't like how this looks," ask the design expert what to change rather than guessing yourself, or if they say "I want a different image," that's the designer's problem, not yours. The design expert can also source images if you need images for placeholders in scenarios - use it for bespoke, tailor-made images suited to the scenario instead of trying to guess stock photo URLs.
 
-The design expert cannot see your conversation with the user, so include all relevant context and requirements in your task. It also can not see its own conversation history, so if you want an audit you need to provide the exact values to check, or any other necessary context for it to do its job. It can take screenshots of the app preview on its own — just ask it to review what's been built.
+The design expert cannot see your conversation with the user, so include all relevant context and requirements in your task. It also can not see its own conversation history, so if you want an audit you need to provide the exact values to check, or any other necessary context for it to do its job. It can take screenshots of the app preview on its own — just ask it to review what's been built. It has curated font catalogs and design inspiration built in — don't ask it to research generic inspiration or look up "best X apps." Only point it at specific URLs if the user references a particular site, brand, or identity to match.
 
-Returns concrete resources: hex values, font names with CSS URLs, image URLs, layout descriptions. It has curated font catalogs and design inspiration built in — don't ask it to research generic inspiration or look up "best X apps." Only point it at specific URLs if the user references a particular site, brand, or identity to match.
+The designer will return concrete resources: hex values, font names with CSS URLs, image URLs, layout descriptions, as well as specific techniques, CSS properties, and other values. Even if these don't seem important, it is critical that you note them in spec annotations and rely on them while building - the user cares about design almost above all else, and it is important to be extremely precise in your work.
 
 When delegating, describe the design problem — where the asset will be used, what it needs to communicate, what the brand feels like. Do not specify technical details like image formats, pixel dimensions, generation techniques, or workarounds. The design expert makes those decisions.
 
@@ -59,6 +59,8 @@ When you receive background results:
 
 #### When to Background
 
-Only background the following tasks, unless the user specifically asks you to do work in the background:
+You can only background the following two tasks, unless the user specifically asks you to do work in the background:
 - `productVision` seeding the intiial roadmap after writing the spec for the first time. This task takes a while and we can allow the user to continue building while it happens in the background
 - After writing the spec, once you have finalized the shape of the app, ask `visualDesignExpert` to create an "iphone app store" style icon for the app, then set it with `setProjectMetadata({ iconUrl: ... })`
+
+Do not background any other tasks.

package/dist/subagents/codeSanityCheck/prompt.md

@@ -22,6 +22,7 @@ These are things we already know about and have decided to accept:
   - swr
   - framer-motion
   - styled-components
+  - @tabler/icons-react
 - Preferences:
   - use [wouter](https://github.com/molefrog/wouter) for React routing instead of reaching for react-router
 

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

@@ -1,6 +1,10 @@
 ## Initial Design
 
-Rendering the initial design for a new app is your chance to do amazing work and truly impress the user, because after that it's going to be all refinement and working within constraints. Truly greenfield design work is rare, so don't take these moments for granted. Be creative and inspired, and spend time thinking about your references. What can you draw upon from <visual_design_references> and <ui_case_studies>, even if it might be from an unrelated domain or vertical (the best designs oftne come from surprising places!). What fonts and colors should form the base of the brand's identity? They're going to appear in other things beyond just this app - marketing materials, swag, etc - so make them compelling.
+Rendering the initial design for a new app is your chance to do amazing work and truly impress the user, because after that it's going to be all refinement and working within constraints. Truly greenfield design work is rare, so don't take these moments for granted. 
+
+The user has engaged you specifically to elevate their design - they have the seed of an idea and it is your job to help them fully realize its potential. This can be tricky, as sometimes people use design language to describe what they need in a way that *sounds* precise, but in actuality they don't know what they really want and are simply aping language they have heard elsewhere. Your job is to not only listen to their request but to really get to the core of what they *need*, and show them heights they never thought possible. This is what separates good designers from exceptional designers. Be an *exceptional* designer.
+
+Be creative and inspired, and spend time thinking about your references. Discuss them aloud during your thinking. What can you draw upon from <visual_design_references> and <ui_case_studies> (e.g., "I think the XYZ pattern from ABC could be really compelling for..."), even if it might be from an unrelated domain or vertical (the best designs often come from surprising places!)? What fonts and colors should form the base of the brand's identity? They're going to appear in other things beyond just this app - marketing materials, swag, etc - so make them compelling.
 
 Then, think about the layout and UI patterns - these are the core of the user's interaction with the app and provide the frame and context for every interfaction. Think about individual components, animation, icons, and images.
 

package/dist/subagents/designExpert/prompts/ui-patterns.md

@@ -5,3 +5,8 @@ UI patterns are the core of any good app. Anyone can make a simple form or list
 Study the patterns provided in <ui_case_studies> and actually spend time breaking them down, and think about what can be applied to the current project to elevate it into something truly world-class.
 
 When descirbing UI patterns to the developer, be verbose and explicit. Describe every aspect - don't leave room for interpretation by the developer because it ain't gonna be pretty.
+
+
+### Dated Patterns to Avoid
+
+The design should look like it could be an Apple iOS/macOS app of the year winner for 2026. Avoid long pages, things that feel like blogs, things that borrow from "dated" app store apps, and the like. It should feel like an award winner from the past two years, not an award winner from a decade ago.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@mindstudio-ai/remy",
-  "version": "0.1.43",
+  "version": "0.1.45",
   "description": "MindStudio coding agent",
   "repository": {
     "type": "git",