@orangecatai/adgen-canvas 0.0.29 → 0.0.31

package/dist/dev/index.js

@@ -66,10 +66,10 @@ import {
   serializeAsJSON,
   serializeLibraryAsJSON,
   strokeRectWithRotation_simple
-} from "./chunk-MKNEZM6D.js";
+} from "./chunk-N472V5Q6.js";
 import {
   define_import_meta_env_default
-} from "./chunk-FOD3MMRJ.js";
+} from "./chunk-3SKNYE32.js";
 import {
   en_default
 } from "./chunk-LIELFD5T.js";
@@ -9796,7 +9796,7 @@ var exportCanvas = async (type, elements, appState, files, {
     let blob = canvasToBlob(tempCanvas);
     if (appState.exportEmbedScene) {
       blob = blob.then(
-        (blob2) => import("./data/image-GKX5OVTE.js").then(
+        (blob2) => import("./data/image-C5LHCG7R.js").then(
           ({ encodePngMetadata: encodePngMetadata2 }) => encodePngMetadata2({
             blob: blob2,
             metadata: serializeAsJSON(elements, appState, files, "local")
@@ -24225,6 +24225,7 @@ async function runReviewerAgent(opts) {
     // imagePlacement retained in opts type for backward compatibility but unused
     // in the image-first flow (the ad is a single full-frame generated image).
     userBrief,
+    originalImageDataUrl,
     brandContext,
     chatUrl,
     reviewerModel = "google/gemini-2.5-flash",
@@ -24244,6 +24245,7 @@ async function runReviewerAgent(opts) {
   const fontPreviewIsBase64 = typeof fontPreviewDataUrl === "string" && fontPreviewDataUrl.startsWith("data:");
   const bodyFontPreviewDataUrl = brandContext?.bodyFontPreviewDataUrl;
   const bodyFontPreviewIsBase64 = typeof bodyFontPreviewDataUrl === "string" && bodyFontPreviewDataUrl.startsWith("data:");
+  const originalImageIsBase64 = typeof originalImageDataUrl === "string" && originalImageDataUrl.startsWith("data:");
   const contextLines = [
     `## Ad Brief (what the user requested)`,
     userBrief,
@@ -24251,6 +24253,13 @@ async function runReviewerAgent(opts) {
     `## Frame dimensions`,
     `Width: ${frame.width}px, Height: ${frame.height}px`
   ];
+  if (originalImageIsBase64) {
+    contextLines.push(
+      ``,
+      `## Original reference image (what the user uploaded)`,
+      `The image the user originally attached is included below (immediately after the ad screenshot). Judge fidelity against it: the generated ad should preserve the subject, composition, and style of this reference unless the brief explicitly asked for a different treatment. Flag it as an issue if the ad has drifted from this reference in ways the brief did not request (wrong subject, different color scheme, altered composition).`
+    );
+  }
   if (brandContext) {
     const brandSummary = buildReviewerBrandSummary(brandContext);
     if (brandSummary) {
@@ -24261,7 +24270,7 @@ async function runReviewerAgent(opts) {
     contextLines.push(
       ``,
       `## Brand logo (reference)`,
-      `The brand logo is attached as the second image below (after the ad screenshot). Use it to evaluate whether the logo in the ad matches the brand's actual logo in colour, shape, and typographic treatment.`
+      `The brand logo is attached below (after the ad screenshot${originalImageIsBase64 ? " and the original reference image" : ""}). Use it to evaluate whether the logo in the ad matches the brand's actual logo in colour, shape, and typographic treatment.`
     );
   } else if (logoIsExternalUrl) {
     contextLines.push(
@@ -24292,6 +24301,12 @@ async function runReviewerAgent(opts) {
       image_url: { url: screenshotDataUrl }
     }
   ];
+  if (originalImageIsBase64 && originalImageDataUrl) {
+    userContent.push({
+      type: "image_url",
+      image_url: { url: originalImageDataUrl }
+    });
+  }
   if (logoIsBase64 && logoDataUrl) {
     userContent.push({
       type: "image_url",
@@ -26049,7 +26064,7 @@ var SIZE_SCALE_MAP = {
 };
 function buildBrandContextMessage(ctx) {
   const lines = [
-    "## Brand Identity \u2014 apply these guidelines to all ads you create"
+    "## Brand Identity \u2014 you MUST apply every guideline below to all ads you create, by default"
   ];
   if (ctx.colors) {
     const { primary, secondary, accent, background } = ctx.colors;
@@ -26124,6 +26139,10 @@ function buildBrandContextMessage(ctx) {
       lines.push(`**Imagery direction**: ${parts.join(", ")}`);
     }
   }
+  lines.push(
+    "",
+    "**Precedence**: these are defaults, not absolutes. If the user's request explicitly states or clearly implies something different from a guideline above \u2014 a different color, mood, style, subject treatment, or tone \u2014 follow the user's explicit instruction for that ad instead of the brand guideline it conflicts with. Do not silently override what the user asked for in favor of brand consistency; only apply a guideline where the user's request is silent on that point."
+  );
   return lines.join("\n");
 }
 function buildImageGenSafePrompt(prompt, brandContext, dimensions) {
@@ -26153,7 +26172,9 @@ function buildImageGenSafePrompt(prompt, brandContext, dimensions) {
   }
   const brandImageHint = brandImageInstructions.length > 0 ? ` ${brandImageInstructions.join(" ")}` : "";
   const brandNameHint = brandContext.name ? ` This ad is for the brand "${brandContext.name}" \u2014 all visual elements, copy, and style must reflect this brand exclusively.` : "";
-  return `${prompt}${brandNameHint}${fontHint}${brandImageHint} Render all specified copy as crisp, correctly-spelled, legible text within the image. Maintain generous padding and whitespace \u2014 at minimum 4\u20135% of the frame width as margin on all sides; copy blocks, the logo, and the CTA must each have clear breathing room and must not feel cramped or crowded. Professional advertising quality.`;
+  return `${prompt}${brandNameHint}${fontHint}${brandImageHint} Render all specified copy as crisp, correctly-spelled, legible text within the image. Maintain generous padding and whitespace \u2014 at minimum 4\u20135% of the frame width as margin on all sides; copy blocks, the logo, and the CTA must each have clear breathing room and must not feel cramped or crowded. Professional advertising quality.
+
+${PRECEDENCE_LINE}`;
 }
 var clean = (value) => {
   const trimmed = value?.trim();
@@ -56705,6 +56726,39 @@ function execAddText(args, ctx) {
     statusMessage: `Added text "${args.text.slice(0, 40)}${args.text.length > 40 ? "\u2026" : ""}"`
   };
 }
+var frameGenerationMemory = /* @__PURE__ */ new Map();
+var chatCreatedFrames = [];
+function getChatCreatedFramesContext() {
+  if (chatCreatedFrames.length === 0)
+    return "";
+  const lines = chatCreatedFrames.map(
+    (f, i) => `${i + 1}. frameId="${f.frameId}"${i === chatCreatedFrames.length - 1 ? " (most recent)" : ""} \u2014 ${f.summary}`
+  );
+  return [
+    `## Frames generated in this conversation (not all frames on the canvas \u2014 the canvas may contain other unrelated images/content)`,
+    ...lines,
+    ``,
+    `When the user refers to an image ambiguously \u2014 "the previous one," "that image," "the above ad," "fix it" \u2014 without an @mention or a distinguishing description, they mean the MOST RECENT frame in this list. Do NOT call list_frames() to search the whole canvas for it; use the frameId given here directly. Only search more broadly (list_frames) if the user's wording clearly points at something NOT in this list, or if this list has more than one plausible match and the user's description (e.g. "the dog one," "the second one") disambiguates between entries here.`
+  ].join("\n");
+}
+function resetAndSeedFrameGenerationMemory(messages) {
+  frameGenerationMemory.clear();
+  chatCreatedFrames.length = 0;
+  for (const msg of messages) {
+    const g = msg.generatedFrame;
+    if (!g || frameGenerationMemory.has(g.frameId))
+      continue;
+    frameGenerationMemory.set(g.frameId, {
+      originalPrompt: g.prompt,
+      referenceImageIndex: g.referenceImageIndex,
+      userBrief: g.userBrief
+    });
+    chatCreatedFrames.push({
+      frameId: g.frameId,
+      summary: g.prompt.slice(0, 80) + (g.prompt.length > 80 ? "\u2026" : "")
+    });
+  }
+}
 async function execGenerateImage(args, ctx) {
   const frameId = args.frameId;
   const prompt = args.prompt;
@@ -56713,7 +56767,19 @@ async function execGenerateImage(args, ctx) {
   const width = args.width;
   const height = args.height;
   const referenceImageIndex = args.referenceImageIndex;
-  const referenceImageDataUrl = referenceImageIndex !== void 0 ? ctx.fileAttachments?.[referenceImageIndex]?.dataUrl : void 0;
+  const priorGeneration = frameGenerationMemory.get(frameId);
+  const effectiveReferenceImageIndex = referenceImageIndex !== void 0 ? referenceImageIndex : priorGeneration ? priorGeneration.referenceImageIndex : ctx.fileAttachments && ctx.fileAttachments.length > 0 ? 0 : void 0;
+  const referenceImageDataUrl = effectiveReferenceImageIndex !== void 0 ? ctx.fileAttachments?.[effectiveReferenceImageIndex]?.dataUrl : void 0;
+  const isRegeneration = !!priorGeneration && priorGeneration.originalPrompt !== prompt;
+  const originalPromptAnchor = isRegeneration ? `
+
+ORIGINAL AD BRIEF (this is a revision, not a new ad \u2014 keep the same subject, composition, and color palette as this original brief unless a fix above explicitly changes it):
+"${priorGeneration.originalPrompt}"` : "";
+  const userBrief = priorGeneration?.userBrief ?? ctx.originalUserBrief;
+  const userBriefAnchor = userBrief ? `
+
+USER'S EXACT REQUEST (verbatim, do not paraphrase, simplify, or substitute the specific subject, action, pose, expression, or scene described here \u2014 depict it faithfully even if the description above worded it differently):
+"${userBrief}"` : "";
   const fontNames = [ctx.brandHeadlineFontName, ctx.brandBodyFontName].filter(
     (n, i, arr) => Boolean(n) && arr.indexOf(n) === i
   );
@@ -56748,7 +56814,7 @@ LAYOUT & SPACING RULES \u2014 these are non-negotiable:
 6. NO CLUTTER: The overall composition must feel open and airy \u2014 fewer elements with more space between them is always better than many elements squeezed together.
 7. ASPECT RATIO: Do NOT squish, stretch, or distort any element. Maintain the natural proportions of all visuals, logos, and text at all times.
 
-Render all text as crisp, correctly-spelled, legible copy. Professional advertising quality \u2014 strong visual hierarchy, brand-consistent, generous whitespace throughout.`;
+Render all text as crisp, correctly-spelled, legible copy. Professional advertising quality \u2014 strong visual hierarchy, brand-consistent, generous whitespace throughout.${originalPromptAnchor}${userBriefAnchor}`;
   if (!frameId || !prompt) {
     return {
       success: false,
@@ -56892,6 +56958,19 @@ Render all text as crisp, correctly-spelled, legible copy. Professional advertis
         lastRetrieved: Date.now()
       }
     ]);
+    if (!priorGeneration) {
+      frameGenerationMemory.set(frameId, {
+        originalPrompt: prompt,
+        referenceImageIndex: effectiveReferenceImageIndex,
+        userBrief
+      });
+      chatCreatedFrames.push({
+        frameId,
+        summary: prompt.slice(0, 80) + (prompt.length > 80 ? "\u2026" : "")
+      });
+    } else if (effectiveReferenceImageIndex !== void 0) {
+      priorGeneration.referenceImageIndex = effectiveReferenceImageIndex;
+    }
     return {
       success: true,
       data: {
@@ -56901,7 +56980,8 @@ Render all text as crisp, correctly-spelled, legible copy. Professional advertis
         x: Math.round(absX),
         y: Math.round(absY),
         width: Math.round(width),
-        height: Math.round(height)
+        height: Math.round(height),
+        ...effectiveReferenceImageIndex !== void 0 ? { referenceImageIndex: effectiveReferenceImageIndex } : {}
       },
       statusMessage: `Generated image from prompt "${prompt.slice(0, 50)}${prompt.length > 50 ? "\u2026" : ""}"`
     };
@@ -58374,6 +58454,18 @@ function isStreamingResponse(response) {
 }
 
 // components/ai-chat/agentLoop.ts
+function isRasterImageDataUrl(dataUrl) {
+  return /^data:image\/(png|jpe?g|gif|webp);base64,/i.test(dataUrl);
+}
+function dataUrlByteSize(dataUrl) {
+  const commaIdx = dataUrl.indexOf(",");
+  const b64 = commaIdx >= 0 ? dataUrl.slice(commaIdx + 1) : dataUrl;
+  return Math.floor(b64.length * 3 / 4);
+}
+var MAX_REQUEST_IMAGE_BYTES = 8 * 1024 * 1024;
+var BRAND_CONTEXT_IMAGE_BUDGET = MAX_REQUEST_IMAGE_BYTES / 2;
+var PER_BRAND_IMAGE_BYTE_CAP = 600 * 1024;
+var MAX_TOTAL_IMAGES = 18;
 var SYSTEM_PROMPT = `You are an expert advertising art director embedded inside a canvas editor. You create finished advertisement creatives by generating a single complete image \u2014 visuals AND all copy baked directly into the image \u2014 using an AI image model. You do NOT assemble ads from separate text/shape layers.
 
 ## Core principle \u2014 ONE image IS the ad
@@ -58389,7 +58481,7 @@ Check: is the user asking to change something in an ad that already exists on th
 Triggers: "change the CTA", "update the headline", "change the button text", "make the headline say X", "use a different background", "change the colors", "different photo", etc.
 
 **For a from-scratch (image-generated) ad, the copy is baked into the image \u2014 it CANNOT be patched element-by-element.** To change anything (text or visuals):
-1. **Find the frame** \u2014 use the frameId from an @mention or your own tool-call history. If unsure, call \`list_frames()\`.
+1. **Find the frame** \u2014 use the frameId from an @mention if present. Otherwise, for a vague reference like "the previous one", "that image", "the above ad", or "fix it", use the most recent frameId from the "Frames generated in this conversation" list provided to you \u2014 do NOT call \`list_frames()\` for this, the canvas may contain other unrelated frames that list excludes on purpose. Only call \`list_frames()\` when that list is empty or genuinely doesn't cover what the user means.
 2. **Duplicate the frame to a fresh, non-overlapping position** (you keep the original intact).
 3. Call \`generate_image\` on the duplicate with the SAME prompt as the original, modified with the requested change (new copy text, new visual direction, etc.).
 4. Call \`finalize_ad\` on the new frame.
@@ -58431,11 +58523,13 @@ The \`generate_image\` prompt is the ONLY creative call \u2014 it must produce t
    - Sign-off / brand name: exact words.
    Spell every word exactly as given; do not paraphrase on-image text.
 
-2. **BRAND COLORS** (use exact hex values from the Brand Identity context when present): which color is the background, the headline, the CTA button, accents.
+2. **BRAND COLORS** (use exact hex values from the Brand Identity context when present, by default): which color is the background, the headline, the CTA button, accents.
 
-3. **TYPOGRAPHY STYLE** (described in words; the image model cannot load font files): describe the feel \u2014 e.g. "bold condensed sans-serif headline, clean body text", matching the brand typography from the Brand Identity context.
+3. **TYPOGRAPHY STYLE** (described in words; the image model cannot load font files): describe the feel \u2014 e.g. "bold condensed sans-serif headline, clean body text", matching the brand typography from the Brand Identity context, by default.
 
-4. **VISUAL DIRECTION:** composition and layout (where the subject sits, where copy sits), lighting, mood, and emotional tone from the brief and \`imagery\` / \`visualStyle\`.
+**Brand guidelines vs. the user's request \u2014 precedence:** Brand Identity guidelines (colors, typography, visual style, tone) are defaults that apply where the user hasn't said otherwise. If the user's message explicitly asks for something that conflicts with a guideline \u2014 a different color, a different mood or style, a different tone \u2014 follow the user's explicit instruction for that ad. Do not silently revert to brand colors/style/tone over what the user actually asked for. This does not apply to the brand logo or brand font files (item 6 below) \u2014 those are always reproduced faithfully regardless of the rest of the prompt, since omitting or altering the logo isn't a stylistic choice the user is likely to have intended just by not mentioning it.
+
+4. **VISUAL DIRECTION:** composition and layout (where the subject sits, where copy sits), lighting, mood, and emotional tone from the brief and \`imagery\` / \`visualStyle\`. If the user described a specific subject, action, pose, expression, object, or animal (e.g. "boy playing guitar, dog watching with one ear up, amused"), you MUST depict that exact scene \u2014 do not invent, substitute, or simplify it into something else (e.g. do not turn "playing guitar" into "walking"). Preserve every specific detail the user gave you; only add the brand/style/spacing layers on top of it.
 
 5. **DIMENSIONS / ASPECT:** describe the format so the composition fits (e.g. "square 1080\xD71080 social post", "landscape 1200\xD7628 banner", "portrait 1080\xD71920 story").
 
@@ -58611,8 +58705,12 @@ async function runAgentLoop(opts) {
       allImageAttachments.push(img);
     }
   }
+  const fullUserBrief = userMessages.filter(
+    (m) => m.role === "user" && typeof m.content === "string" && m.content.trim().length > 0
+  ).map((m) => m.content).join("\n\n");
   const augmentedToolCtx = {
     ...toolCtx,
+    ...fullUserBrief ? { originalUserBrief: fullUserBrief } : {},
     // Lets long-running tools (video gen) push interim progress lines into the
     // same status indicator the loop itself writes to.
     onStatus: (message) => onUpdate({ type: "status", message }),
@@ -58645,6 +58743,12 @@ async function runAgentLoop(opts) {
 ${elementContext}`
     });
   }
+  const chatFramesContext = getChatCreatedFramesContext();
+  if (chatFramesContext) {
+    messages.push({ role: "system", content: chatFramesContext });
+  }
+  let brandContextBytesUsed = 0;
+  let brandContextImagesUsed = 0;
   {
     const lastUserText = [...userMessages].reverse().find((m) => m.role === "user")?.content ?? "";
     if (lastUserText && (toolCtx.onSearchBrandAssets || toolCtx.onSearchAdCreatives || toolCtx.onListBrandTemplates)) {
@@ -58658,9 +58762,11 @@ ${elementContext}`
         toolCtx.onSearchAdCreatives ? toolCtx.onSearchAdCreatives(lastUserText) : Promise.resolve([]),
         toolCtx.onListBrandTemplates ? toolCtx.onListBrandTemplates(keywords) : Promise.resolve([])
       ]);
-      const top2Assets = (assetRes.status === "fulfilled" ? assetRes.value : []).filter((a) => a.previewDataUrl).slice(0, 2);
-      const top2Creatives = (creativeRes.status === "fulfilled" ? creativeRes.value : []).filter((c) => c.previewDataUrl).slice(0, 2);
-      const top2Templates = (templateRes.status === "fulfilled" ? templateRes.value : []).filter((t2) => t2.thumbnailDataUrl).slice(0, 2);
+      const top2Assets = (assetRes.status === "fulfilled" ? assetRes.value : []).filter((a) => a.previewDataUrl && isRasterImageDataUrl(a.previewDataUrl)).slice(0, 2);
+      const top2Creatives = (creativeRes.status === "fulfilled" ? creativeRes.value : []).filter((c) => c.previewDataUrl && isRasterImageDataUrl(c.previewDataUrl)).slice(0, 2);
+      const top2Templates = (templateRes.status === "fulfilled" ? templateRes.value : []).filter(
+        (t2) => t2.thumbnailDataUrl && isRasterImageDataUrl(t2.thumbnailDataUrl)
+      ).slice(0, 2);
       console.log(
         `[agent:pre-fetch] brand assets=${top2Assets.length}/${assetRes.status === "fulfilled" ? assetRes.value.length : "err"} ad-creatives=${top2Creatives.length}/${creativeRes.status === "fulfilled" ? creativeRes.value.length : "err"} templates=${top2Templates.length}/${templateRes.status === "fulfilled" ? templateRes.value.length : "err"}`
       );
@@ -58678,8 +58784,27 @@ ${elementContext}`
           dataUrl: t2.thumbnailDataUrl
         }))
       ];
-      if (previews.length > 0) {
-        previews.forEach(
+      const budgetedPreviews = [];
+      for (const p of previews) {
+        const size = dataUrlByteSize(p.dataUrl);
+        if (size > PER_BRAND_IMAGE_BYTE_CAP) {
+          console.warn(
+            `[agent:pre-fetch] dropping "${p.label}" \u2014 ${Math.round(size / 1024)}KB exceeds the per-image cap (${Math.round(PER_BRAND_IMAGE_BYTE_CAP / 1024)}KB)`
+          );
+          continue;
+        }
+        if (brandContextBytesUsed + size > BRAND_CONTEXT_IMAGE_BUDGET || brandContextImagesUsed >= MAX_TOTAL_IMAGES) {
+          console.warn(
+            `[agent:pre-fetch] dropping "${p.label}" \u2014 brand-context image budget exhausted`
+          );
+          continue;
+        }
+        brandContextBytesUsed += size;
+        brandContextImagesUsed += 1;
+        budgetedPreviews.push(p);
+      }
+      if (budgetedPreviews.length > 0) {
+        budgetedPreviews.forEach(
           (p) => console.log(
             `[agent:pre-fetch] injecting "${p.label}" as base64 image \u2014 ${Math.round(p.dataUrl.length / 1024)}KB (${p.dataUrl.slice(
               0,
@@ -58690,7 +58815,7 @@ ${elementContext}`
         messages.push({
           role: "system",
           content: `BRAND REFERENCES (auto-fetched, up to 2 per category):
-${previews.map((p) => `\u2022 ${p.label}`).join(
+${budgetedPreviews.map((p) => `\u2022 ${p.label}`).join(
             "\n"
           )}
 
@@ -58701,9 +58826,9 @@ Visual previews follow in the next message. Use them as context: match their sty
           content: [
             {
               type: "text",
-              text: `Brand reference previews (${previews.length} total):`
+              text: `Brand reference previews (${budgetedPreviews.length} total):`
             },
-            ...previews.map((p) => ({
+            ...budgetedPreviews.map((p) => ({
               type: "image_url",
               image_url: { url: p.dataUrl }
             }))
@@ -58711,17 +58836,57 @@ Visual previews follow in the next message. Use them as context: match their sty
         });
         messages.push({
           role: "assistant",
-          content: `Got it \u2014 I can see ${previews.length} brand reference(s): ${previews.map((p) => p.label).join(", ")}. I'll use these as visual context for the ad.`
+          content: `Got it \u2014 I can see ${budgetedPreviews.length} brand reference(s): ${budgetedPreviews.map((p) => p.label).join(", ")}. I'll use these as visual context for the ad.`
         });
       }
     }
   }
+  const frameScreenshotBytes = frameScreenshot ? dataUrlByteSize(frameScreenshot) : 0;
+  const currentSendImagesForBudget = fileAttachments?.filter((a) => a.type === "image" && a.dataUrl) ?? [];
+  const currentSendImageBytes = currentSendImagesForBudget.reduce(
+    (sum2, a) => sum2 + dataUrlByteSize(a.dataUrl),
+    0
+  );
+  const currentSendImageCount = currentSendImagesForBudget.length + (frameScreenshot ? 1 : 0);
+  const remainingBudgetBytes = MAX_REQUEST_IMAGE_BYTES - brandContextBytesUsed;
+  const remainingBudgetImages = MAX_TOTAL_IMAGES - brandContextImagesUsed;
+  if (currentSendImageBytes + frameScreenshotBytes > remainingBudgetBytes || currentSendImageCount > remainingBudgetImages) {
+    const message = "This message is too large to process \u2014 try attaching fewer or smaller images.";
+    console.warn(
+      `[agent:budget] aborting \u2014 current-send content alone (${Math.round(
+        (currentSendImageBytes + frameScreenshotBytes) / 1024
+      )}KB, ${currentSendImageCount} image(s)) exceeds the remaining request budget`
+    );
+    onUpdate({ type: "final", message });
+    return { reply: message, toolActions: [] };
+  }
+  let historicalBudgetBytes = remainingBudgetBytes - currentSendImageBytes - frameScreenshotBytes;
+  let historicalBudgetImages = remainingBudgetImages - currentSendImageCount;
+  const droppedHistoricalMessageIndices = /* @__PURE__ */ new Set();
+  for (let i = userMessages.length - 1; i >= 0; i--) {
+    const imgs = userMessages[i].imageAttachments;
+    if (!imgs || imgs.length === 0)
+      continue;
+    const bytes = imgs.reduce(
+      (sum2, img) => sum2 + dataUrlByteSize(img.dataUrl),
+      0
+    );
+    if (bytes <= historicalBudgetBytes && imgs.length <= historicalBudgetImages) {
+      historicalBudgetBytes -= bytes;
+      historicalBudgetImages -= imgs.length;
+    } else {
+      droppedHistoricalMessageIndices.add(i);
+      console.warn(
+        `[agent:budget] dropping ${imgs.length} historical image(s) from an earlier turn (${Math.round(bytes / 1024)}KB) \u2014 request budget exceeded`
+      );
+    }
+  }
   for (let i = 0; i < userMessages.length; i++) {
     const msg = userMessages[i];
     const isLast = i === userMessages.length - 1;
     const lastMsgImages = isLast && fileAttachments ? fileAttachments.filter((a) => a.type === "image" && a.dataUrl) : [];
     const lastMsgPdfs = isLast && fileAttachments ? fileAttachments.filter((a) => a.type === "pdf" && a.dataUrl) : [];
-    const historicalImages2 = msg.imageAttachments ?? [];
+    const historicalImages2 = droppedHistoricalMessageIndices.has(i) ? [] : msg.imageAttachments ?? [];
     const hasImages = historicalImages2.length > 0 || lastMsgImages.length > 0;
     const hasFrameScreenshot = isLast && !!frameScreenshot;
     const hasTextFiles = isLast && fileAttachments && fileAttachments.some((a) => a.type === "text" || a.type === "docx");
@@ -58807,7 +58972,16 @@ ${f.textContent}`).join("\n\n")}`;
         if (lastCreatedFrameId) {
           adGenShimmerStore.stopGenerating(lastCreatedFrameId);
         }
-        return { reply, toolActions };
+        return {
+          reply,
+          toolActions,
+          frameGeneration: imageGenData ? {
+            frameId: imageGenData.frameId,
+            prompt: imageGenData.prompt,
+            referenceImageIndex: imageGenData.referenceImageIndex,
+            userBrief: fullUserBrief
+          } : void 0
+        };
       }
       messages.push({
         role: "assistant",
@@ -58936,7 +59110,9 @@ ${f.textContent}`).join("\n\n")}`;
             x: result.data.x,
             y: result.data.y,
             width: result.data.width,
-            height: result.data.height
+            height: result.data.height,
+            referenceImageIndex: result.data.referenceImageIndex,
+            prompt: parsedArgs.prompt ?? ""
           };
           if (onAfterImageGen) {
             onAfterImageGen();
@@ -58998,25 +59174,45 @@ Use the frameId="${imageGenData.frameId}" for all elements. Keep all elements wi
         ]);
         const assetSearchResult = result;
         if (searchToolNames.has(name) && result.success && assetSearchResult.previewDataUrls?.length) {
-          const previews = assetSearchResult.previewDataUrls;
-          console.log(
-            `[agent:mid-loop] ${name} injecting ${previews.length} preview image(s) as base64 vision content: ${previews.map(
-              (p) => `"${p.name}" (${Math.round(p.dataUrl.length / 1024)}KB)`
-            ).join(", ")}`
-          );
-          messages.push({
-            role: "user",
-            content: [
-              {
-                type: "text",
-                text: `Visual previews for the ${previews.length} result(s) from ${name}:`
-              },
-              ...previews.map((p) => ({
-                type: "image_url",
-                image_url: { url: p.dataUrl }
-              }))
-            ]
-          });
+          const budgetedPreviews = [];
+          for (const p of assetSearchResult.previewDataUrls) {
+            const size = dataUrlByteSize(p.dataUrl);
+            if (size > PER_BRAND_IMAGE_BYTE_CAP) {
+              console.warn(
+                `[agent:mid-loop] dropping "${p.name}" \u2014 ${Math.round(size / 1024)}KB exceeds the per-image cap`
+              );
+              continue;
+            }
+            if (brandContextBytesUsed + size > BRAND_CONTEXT_IMAGE_BUDGET || brandContextImagesUsed >= MAX_TOTAL_IMAGES) {
+              console.warn(
+                `[agent:mid-loop] dropping "${p.name}" \u2014 brand-context image budget exhausted`
+              );
+              continue;
+            }
+            brandContextBytesUsed += size;
+            brandContextImagesUsed += 1;
+            budgetedPreviews.push(p);
+          }
+          if (budgetedPreviews.length > 0) {
+            console.log(
+              `[agent:mid-loop] ${name} injecting ${budgetedPreviews.length} preview image(s) as base64 vision content: ${budgetedPreviews.map(
+                (p) => `"${p.name}" (${Math.round(p.dataUrl.length / 1024)}KB)`
+              ).join(", ")}`
+            );
+            messages.push({
+              role: "user",
+              content: [
+                {
+                  type: "text",
+                  text: `Visual previews for the ${budgetedPreviews.length} result(s) from ${name}:`
+                },
+                ...budgetedPreviews.map((p) => ({
+                  type: "image_url",
+                  image_url: { url: p.dataUrl }
+                }))
+              ]
+            });
+          }
         }
         if (name === "generate_image" && result.success && imageGenData) {
           const autoFinalizeId = `auto_finalize_${Date.now()}`;
@@ -59074,6 +59270,7 @@ Use the frameId="${imageGenData.frameId}" for all elements. Keep all elements wi
             let feedback;
             try {
               const frameEl = augmentedToolCtx.excalidrawAPI.getSceneElements().find((e) => e.id === imageGenData.frameId);
+              const originalImageDataUrl = imageGenData.referenceImageIndex !== void 0 ? allImageAttachments[imageGenData.referenceImageIndex]?.dataUrl : void 0;
               feedback = await runReviewerAgent({
                 screenshotDataUrl: bannerScreenshot,
                 htmlSource: "",
@@ -59088,7 +59285,12 @@ Use the frameId="${imageGenData.frameId}" for all elements. Keep all elements wi
                   width: imageGenData.width,
                   height: imageGenData.height
                 },
-                userBrief: typeof userMessages[0]?.content === "string" ? userMessages[0].content : "",
+                // Full conversation's user turns, not just userMessages[0] —
+                // the actual creative brief is frequently NOT the first
+                // message (e.g. "hi" -> clarifying questions -> the message
+                // that actually describes the ad). See fullUserBrief above.
+                userBrief: fullUserBrief,
+                originalImageDataUrl,
                 brandContext,
                 chatUrl,
                 reviewerModel: reviewerModel ?? chatModel,
@@ -59123,7 +59325,7 @@ Use the frameId="${imageGenData.frameId}" for all elements. Keep all elements wi
                 "",
                 `Overall verdict: ${feedback.summary}`,
                 "",
-                `Apply ALL fixes by calling generate_image AGAIN on frameId="${imageGenData.frameId}" with a revised prompt that incorporates every correction above (re-state the full ad \u2014 visuals AND all copy \u2014 with every fix folded in). Prioritise critical issues first. Do NOT call generate_html_banner or add separate text/shape layers \u2014 the ad is a single generated image.`
+                `Apply ALL fixes by calling generate_image AGAIN on frameId="${imageGenData.frameId}" with a revised prompt. This is a REVISION of the existing ad, not a new one: start from the exact prompt you used last time and change ONLY what the fixes above require \u2014 keep the same subject, composition, color palette, and any reference image untouched otherwise. Do NOT invent a new color scheme, layout, or brand style. If a fix mentions a specific position (e.g. "top-right", "bottom-left"), use that exact position. Prioritise critical issues first. Do NOT call generate_html_banner or add separate text/shape layers \u2014 the ad is a single generated image.`
               ].join("\n");
               messages.push({
                 role: "user",
@@ -59188,7 +59390,16 @@ Describe the final ad to the user and stop.`
     if (lastCreatedFrameId) {
       adGenShimmerStore.stopGenerating(lastCreatedFrameId);
     }
-    return { reply: fallbackReply, toolActions };
+    return {
+      reply: fallbackReply,
+      toolActions,
+      frameGeneration: imageGenData ? {
+        frameId: imageGenData.frameId,
+        prompt: imageGenData.prompt,
+        referenceImageIndex: imageGenData.referenceImageIndex,
+        userBrief: fullUserBrief
+      } : void 0
+    };
   } catch (err) {
     if (lastCreatedFrameId) {
       adGenShimmerStore.stopGenerating(lastCreatedFrameId);
@@ -59663,9 +59874,11 @@ var AIChatPanel = React69.forwardRef(
     const [prompt, setPrompt] = useState60("");
     const [isLoading, setIsLoading] = useState60(false);
     const [statusText, setStatusText] = useState60("");
-    const [messages, setMessages] = useState60(
-      initialMessages ?? []
-    );
+    const [messages, setMessages] = useState60(() => {
+      const initial = initialMessages ?? [];
+      resetAndSeedFrameGenerationMemory(initial);
+      return initial;
+    });
     const [sessions, setSessions] = useState60(
       initialSessions ?? []
     );
@@ -59771,6 +59984,7 @@ var AIChatPanel = React69.forwardRef(
       saveCurrentSession();
       const newSessionId = genId();
       setMessages([]);
+      resetAndSeedFrameGenerationMemory([]);
       setCurrentSessionId(newSessionId);
       setIsLoadingSession(false);
       setHistoryOpen(false);
@@ -59788,9 +60002,11 @@ var AIChatPanel = React69.forwardRef(
         setError(null);
         if (session.messages.length > 0) {
           setMessages(session.messages);
+          resetAndSeedFrameGenerationMemory(session.messages);
           setIsLoadingSession(false);
         } else {
           setMessages([]);
+          resetAndSeedFrameGenerationMemory([]);
           setIsLoadingSession(true);
           if (onSessionSwitch) {
             onSessionSwitch(session.id);
@@ -60389,7 +60605,8 @@ var AIChatPanel = React69.forwardRef(
             content: result.reply,
             timestamp: /* @__PURE__ */ new Date(),
             toolActions: result.toolActions.length > 0 ? result.toolActions : void 0,
-            reviewFeedback: lastReviewFeedbackRef.current ?? void 0
+            reviewFeedback: lastReviewFeedbackRef.current ?? void 0,
+            generatedFrame: result.frameGeneration
           };
           const updatedMessagesAgent = [...nextMessages, assistantMsgAgent];
           setMessages(updatedMessagesAgent);
@@ -60463,6 +60680,7 @@ var AIChatPanel = React69.forwardRef(
           });
           if (targetSessionId === currentSessionIdRef.current) {
             setMessages(msgs);
+            resetAndSeedFrameGenerationMemory(msgs);
             setIsLoadingSession(false);
           }
         },