@koda-sl/baker-cli 0.352.0-dev.09c54d90c → 0.352.0-dev.fabc18041

package/README.md

@@ -5487,7 +5487,7 @@ Turn a source/inspiration image into a **runnable, self-validated static-ad canv
 2. **element selection** (`~google/gemini-flash-latest`) — picks the **main, identity-critical** elements (the brand logo, a showcased product, a trust badge) **plus any foreground/hero person or animal** — the emotional focal point — even a generic one, because a free-generated face/muzzle reads as AI and grows artifacts; the emotional hero always gets a real-reference slot. When the advertiser's logo appears in **more than one lockup** (a square/icon **mark** and a horizontal **wordmark**), each is emitted as its **own** element (e.g. `LOGO_MARK`, `LOGO_WORDMARK`) so you drop the right file in each slot instead of stretching one logo to cover both. The describe pass also records the ad's **typography** under a `fonts` block (each typeface's classification, a best-guess family, and its weight/case) so you know exactly what to drop at the brand-font slot. Background extras are dropped. Each element is stamped back onto its blueprint entry as a `reference_image` label so the JSON self-documents which slot grounds which subject.
 3. **global layout** (`~google/gemini-flash-latest`) — produces a structured `layout` block in `prompt.json`: the column/row grid, each region's `x_pct`/`y_pct` bounds, panel splits, background/shape, and every text block's relative size/weight/case/alignment. This is what gives the generator a precise composition to rebuild.
 
-It then scaffolds a canvas that ingests `prompt.json`, wires **one `[TODO]` ingest slot per detected element** (plus an optional brand-font → type-specimen) into `image_generate`, and wires the original image in for composition only — left out of the render when it shows a person, because the generator copies that face whatever the prompt says (the blueprint's layout carries the composition instead). Each **animal hero** is additionally fused into a generated **multi-view reference sheet** (`image_reference_sheet`, a turnaround built from the one dropped photo) that the render grounds on instead of the lone flat snapshot — the same identity lock the video scaffold uses, so the muzzle stays consistent and artifact-free from a single reference. Pass `--skip-actor-sheets` to ground straight on the dropped photo. A **person hero**'s slot takes a company Avatar's identity sheet (`baker avatars`) and the render keeps that face, so it is never a found photo of a real person. The canvas is validated before it's written. stdout returns `{ ok, canvas_path, prompt_path, models, aspect_ratio, placements, output_formats, layout_regions, stats, checklist }` — the **checklist** lists every real asset to drop in (and which heroes get a sheet).
+It then scaffolds a canvas that ingests `prompt.json`, wires **one `[TODO]` ingest slot per detected element** (plus an optional brand-font → type-specimen) into `image_generate`, and wires the original image in for composition only. Each **person/animal hero** is additionally fused into a generated **multi-view reference sheet** (`image_reference_sheet`, a turnaround built from the one dropped photo) that the render grounds on instead of the lone flat snapshot — the same identity lock the video scaffold uses, so the face/muzzle stays consistent and artifact-free from a single reference. Pass `--skip-actor-sheets` to ground straight on the dropped photo. The canvas is validated before it's written. stdout returns `{ ok, canvas_path, prompt_path, models, aspect_ratio, placements, output_formats, layout_regions, stats, checklist }` — the **checklist** lists every real asset to drop in (and which heroes get a sheet).
 
 **Automatic placement fan-out.** By default the canvas doesn't stop at one ratio: the hero renders once at the platform's primary ratio, then an `image_aspect_adapt` node recomposes it into that platform's full placement set, and the canvas `output` points at the adapted images. The placement set is derived from `--platform` (`meta` → 4:5 + 9:16, `google` → 1:1 + 4:5 + 16:9, `linkedin` → 1:1, `tiktok` → 9:16, `youtube` → 16:9); override it with `--placements <preset>` or disable it with `--placements none` for a single base ad. The hero ratio passes through the adapt node **free**; each other ratio is a billed recomposition — and because adaptation re-generates, every ratio should be eyeballed for text and brand-color fidelity before publish (`ad-formats.md` in the scaffold skill covers this).
 
@@ -5516,7 +5516,7 @@ baker canvas run ./static-ad.canvas.json
 | `--placements <preset\|none>` | derived from `--platform` | Override the fan-out set: `meta_feed`\|`meta_feed_stories`\|`meta_all`\|`linkedin`\|`google_demandgen`\|`tiktok`\|`youtube`, or `none` for a single base ad. |
 | `--aspect <ratio>` | inferred from the image, else `9:16` | Force the hero's aspect ratio (clamped into the placement set so it stays a free pass-through). |
 | `--skip-font` | off | Skip the brand-font → type-specimen slot. |
-| `--skip-actor-sheets` | off | Ground each animal hero on its lone dropped photo instead of a generated multi-view reference sheet. |
+| `--skip-actor-sheets` | off | Ground each person/animal on its lone dropped photo instead of a generated multi-view reference sheet. |
 
 Scaffolding runs (and bills) the two vision passes; **running** the result generates a billed image. `baker canvas validate` does not check that the `[TODO]` paths exist — supply the real files before `run`.
 

package/dist/cli.js

@@ -11035,6 +11035,41 @@ function adCopyHints(payload) {
   return tells.slice(0, MAX_HINTS).map(hintFor);
 }
 
+// src/commands/ads/google/copy-list-usage.ts
+var AD_COPY_LIST_USAGE = `repeat the flag once per item, or pass one comma-separated list and wrap an item that contains a comma in double quotes: '"Fast, free quotes",Book today'. A flag given once is read as a list, so a lone item with a comma needs its double quotes too: '"Fast, free quotes"'`;
+var COPY_LIST_FLAG = {
+  headlines: "--headlines",
+  descriptions: "--descriptions",
+  longHeadlines: "--long-headlines"
+};
+function valueAt(root, path48) {
+  let node = root;
+  for (const key of path48) {
+    if (node === null || typeof node !== "object") return void 0;
+    node = node[key];
+  }
+  return node;
+}
+function countRefusal(op, issue) {
+  if (issue.code !== "too_big" && issue.code !== "too_small" || issue.origin !== "array") return void 0;
+  const flag = COPY_LIST_FLAG[String(issue.path.at(-1))];
+  if (!flag) return void 0;
+  const value = valueAt(op, issue.path);
+  const count2 = Array.isArray(value) ? value.length : void 0;
+  return issue.code === "too_big" ? { flag, count: count2, bound: `at most ${issue.maximum}`, tooMany: true } : { flag, count: count2, bound: `at least ${issue.minimum}`, tooMany: false };
+}
+function copyListCountFix(op, issues) {
+  const refusals = issues.map((issue) => countRefusal(op, issue));
+  if (refusals.length === 0 || refusals.some((refusal) => refusal === void 0)) return void 0;
+  const counted = refusals.filter((refusal) => refusal !== void 0);
+  const limits = counted.map(({ flag, count: count2, bound }) => `${flag} has ${count2 ?? "the wrong number of"} items; this ad takes ${bound}.`).join(" ");
+  const commaCause = counted.some((refusal) => refusal.tooMany) ? ` If you meant fewer items, the copy was cut at its commas: a flag given once is read as a comma-separated list, so an item with a comma in it ("Fast, free quotes", a decimal like "4,5/5") becomes two. To keep a comma, ${AD_COPY_LIST_USAGE} \u2014 or put the content in --file as JSON.` : "";
+  return {
+    action: "fix_copy_list_count",
+    explanation: `${limits}${commaCause} ${commaCause ? "Otherwise add" : "Add"} or drop items to fit the limit, then stage again; the field names are right.`
+  };
+}
+
 // src/commands/ads/google/draft-status.ts
 var OPERATION_LABELS = {
   create: "Creating",
@@ -11355,7 +11390,7 @@ async function stageGoogleOp(raw, hints2) {
         code: "VALIDATION_ERROR",
         message: fields.map((f) => f.path ? `${f.path}: ${f.message}` : f.message).join("; "),
         fields,
-        fix: capabilityFix("google-ads")
+        fix: copyListCountFix(raw, preflight.error.issues) ?? capabilityFix("google-ads")
       }
     });
     process.exit(1);
@@ -11391,7 +11426,7 @@ async function stageGoogleOps(rawOps, hints2) {
           code: "VALIDATION_ERROR",
           message: fields.map((f) => `${f.path}: ${f.message}`).join("; "),
           fields,
-          fix: capabilityFix("google-ads")
+          fix: copyListCountFix(raw, preflight.error.issues) ?? capabilityFix("google-ads")
         }
       });
       process.exit(1);
@@ -16798,7 +16833,6 @@ import { defineCommand as defineCommand31 } from "citty";
 
 // src/commands/ads/google/ad-copy-flags.ts
 var AD_COPY_LIST_FLAGS = ["headlines", "descriptions", "long-headlines"];
-var AD_COPY_LIST_USAGE = `repeat the flag once per item, or pass one comma-separated list and wrap an item that contains a comma in double quotes: '"Fast, free quotes",Book today'. A flag given once is read as a list, so a lone item with a comma needs its double quotes too: '"Fast, free quotes"'`;
 function skipBlanks(value, at) {
   let i = at;
   while (value[i] === " " || value[i] === "	") i++;
@@ -41330,6 +41364,7 @@ import { z as z40 } from "zod";
 var GEN_ASPECT_RATIOS = /* @__PURE__ */ new Set(["1:1", "4:5", "9:16", "16:9", "4:3", "3:4", "2:3", "3:2", "21:9"]);
 var DEFAULT_ASPECT_RATIO = "9:16";
 var SHEET_SUBJECT_TYPE2 = {
+  person: "person",
   animal: "character"
 };
 var ACTOR_SHEET_MODEL2 = "google/gemini-3-pro-image-preview";
@@ -41339,8 +41374,7 @@ var ADAPT_IMAGE_SIZE = "2K";
 var ADAPT_GUIDANCE = "Keep the headline, logo, CTA, and hero subject fully visible in every ratio. Reproduce every text string verbatim \u2014 no dropped, added, or altered characters \u2014 and preserve the exact brand-color treatment (e.g. a black\u2192red word pivot), never flattening it.";
 var Blueprint = z40.object({
   meta: z40.object({ estimated_aspect_ratio: z40.string().optional() }).loose().optional(),
-  text_content: z40.array(z40.object({ text: z40.string().optional() }).loose()).optional(),
-  people: z40.array(z40.unknown()).optional()
+  text_content: z40.array(z40.object({ text: z40.string().optional() }).loose()).optional()
 }).loose();
 var ElementLocator = z40.object({
   collection: z40.enum(["subjects", "people", "brands_logos"]),
@@ -41396,7 +41430,6 @@ function roleForType2(type) {
     case "product":
       return "the showcased product; keep this exact product identity. Ignore any caption text printed on this reference.";
     case "person":
-      return "the emotional hero \u2014 the brand's cast person: a company Avatar's identity sheet, a multi-view turnaround of ONE person the brand may show. Keep this exact face, hair and build, and render them once in the pose, gaze, wardrobe and expression the blueprint specifies \u2014 never as a grid. Ignore the cell labels printed on the sheet.";
     case "animal":
       return "the emotional hero \u2014 the face/muzzle/eyes the viewer locks onto and exactly where AI artifacts (malformed mouths, teeth, foam, extra fingers) betray a fake. It MUST be grounded in this real reference, never invented from prose. Keep this identity AND the expression the blueprint specifies. Ignore any caption text printed on this reference.";
     default:
@@ -41407,10 +41440,7 @@ function todoPath2(el, label) {
   const desc = el.description ? ` \u2014 ${el.description}` : "";
   const expr = el.expression ? `, with a ${el.expression} expression` : "";
   const t = el.type.toLowerCase();
-  if (t === "person") {
-    return `[TODO: drop the identity sheet of a company Avatar for ${label}${desc}${expr} \u2014 REQUIRED: this is the emotional hero and the render keeps this face, so use \`baker avatars list\` or cast one with \`baker avatars create\`, then put its sheetUrl here; never a found photo of a real person (web search, the source ad), who never agreed to appear in this ad]`;
-  }
-  const lock = t === "animal" ? " \u2014 REQUIRED: this is the emotional hero; ground it in a real reference (image-library / video-library / Pinterest), do not delete this slot and free-generate" : "";
+  const lock = t === "person" || t === "animal" ? " \u2014 REQUIRED: this is the emotional hero; ground it in a real reference (image-library / video-library / Pinterest), do not delete this slot and free-generate" : "";
   return `[TODO: drop a real image for ${label} (${el.type})${desc}${expr}${lock} \u2014 replace with a path relative to this canvas file, never an absolute path]`;
 }
 function sheetSubjectDescription(el) {
@@ -41450,9 +41480,6 @@ function textFidelityDirective(blueprint, hasFont) {
     "Keep all type crisp, legible, and large enough to read cleanly; no garbled, warped, or duplicated characters, and no unreadable microtext."
   ].join("\n");
 }
-function showsPerson(blueprint, elements) {
-  return (blueprint.people?.length ?? 0) > 0 || elements.some((el) => el.type.toLowerCase() === "person");
-}
 function scaffoldStaticAd(input, elementsInput, opts) {
   const blueprint = Blueprint.parse(input);
   const elements = MainElements.parse(elementsInput);
@@ -41518,18 +41545,15 @@ function scaffoldStaticAd(input, elementsInput, opts) {
       params: { text: specimenText(blueprint), font_size: 72, line_height: 1.2, max_width: 1600 }
     });
   }
-  const originalAsReference = !showsPerson(blueprint, elements);
   const reference = [
     ...elementSlots.map((s) => s.ref),
     ...hasFont ? ["$ref:type_ref.image"] : [],
-    ...originalAsReference ? ["$ref:original.asset"] : []
+    "$ref:original.asset"
   ];
   const legend = [
     ...elementSlots.map((s) => `- ${s.label} \u2014 ${roleForType2(s.type)}`),
     ...hasFont ? ["- TYPE SPECIMEN \u2014 render every text string in this exact typeface; ignore its color and layout."] : [],
-    ...originalAsReference ? [
-      "- ORIGINAL \u2014 use ONLY for composition, layout, and proportions. IGNORE its text, its logo, its brand name, and its colors entirely \u2014 they belong to a different brand."
-    ] : []
+    "- ORIGINAL \u2014 use ONLY for composition, layout, and proportions. IGNORE its text, its logo, its brand name, and its colors entirely \u2014 they belong to a different brand."
   ].join("\n");
   const exclusionsLast = imageProfileFor(opts.genModel)?.constraintPlacement === "last";
   const exclusionsBlock = "Exclusions: no invented badges, icons, or seals; render exactly the subjects the blueprint names and no others; no competitor branding; no watermark.";
@@ -41593,7 +41617,7 @@ function scaffoldStaticAd(input, elementsInput, opts) {
     schema: "baker-canvas/1",
     metadata: {
       name: "static ad reproduction",
-      description: "Scaffolded by `baker canvas scaffold-static-ad`. Edit `prompt.json` (the blueprint generated from the source image) into the ad you want, drop a real asset at each [TODO] ingest path (logo/product/subject/badge + brand font), then `baker canvas run`. The original is wired for composition only, and left out of the render when it shows a person, whose face the generator would copy."
+      description: "Scaffolded by `baker canvas scaffold-static-ad`. Edit `prompt.json` (the blueprint generated from the source image) into the ad you want, drop a real asset at each [TODO] ingest path (logo/product/subject/badge + brand font), then `baker canvas run`. The original is wired for composition only."
     },
     nodes,
     output: fanOut ? { node: "adapt", output: "images" } : { node: "gen", output: "images" }
@@ -42063,7 +42087,7 @@ var scaffoldStaticAdCommand = defineCommand114({
     "skip-font": { type: "boolean", description: "Skip the brand-font \u2192 type-specimen slot" },
     "skip-actor-sheets": {
       type: "boolean",
-      description: "Ground each animal hero on its lone dropped photo instead of a generated multi-view sheet (a person hero always grounds on the company Avatar sheet dropped at its slot)"
+      description: "Ground each person/animal on its lone dropped photo instead of a generated multi-view sheet"
     }
   },
   async run({ args: args5 }) {
@@ -42204,7 +42228,7 @@ var scaffoldStaticAdCommand = defineCommand114({
             edit_prompt: `Edit ${path21.basename(blueprintPath)} \u2014 it is the blueprint generated from your image; rewrite it into the ad you want (palette, copy, claims, subjects). It feeds the generator directly.`,
             assets_to_supply: report.elements,
             font_slot: report.includes_font ? "Drop a brand font at the [TODO] brandfont path (the describe pass recorded the ad's typefaces under `fonts` in prompt.json \u2014 match those). The font is wired into the render as a TYPE SPECIMEN reference so generated text takes the brand letterforms. Delete the brandfont + type_ref nodes to skip it." : "skipped (--skip-font)",
-            actor_sheets: report.actor_sheets.length > 0 ? `Each animal hero (${report.actor_sheets.join(", ")}) is fused into a generated multi-view reference sheet (image_reference_sheet) that the render grounds on \u2014 so drop ONE clean photo at that hero's ingest and the sheet builds the consistent turnaround. Pass --skip-actor-sheets to ground on the lone photo instead. A person hero takes a company Avatar's identity sheet at its slot (baker avatars), never a found photo.` : "none (no animal heroes detected, or --skip-actor-sheets). A person hero takes a company Avatar's identity sheet at its slot (baker avatars), never a found photo.",
+            actor_sheets: report.actor_sheets.length > 0 ? `Each living hero (${report.actor_sheets.join(", ")}) is fused into a generated multi-view reference sheet (image_reference_sheet) that the render grounds on \u2014 so drop ONE clean photo at that hero's ingest and the sheet builds the consistent turnaround. Pass --skip-actor-sheets to ground on the lone photo instead.` : "none (no person/animal heroes detected, or --skip-actor-sheets)",
             placements: report.placements === null ? `Single format (${report.aspect_ratio}). Pass --placements meta_feed_stories|meta_all|linkedin|\u2026 to auto-fan-out to a platform's set.` : `Auto-fans the approved hero out to ${report.placements} \u2192 ${report.output_formats.join(", ")}. The hero ratio (${report.aspect_ratio}) passes through free; each other ratio is a billed render, AI-recomposed (not cropped). Re-read the copy and confirm the brand-color treatment on EVERY ratio before publishing \u2014 adaptation re-generates and can silently mangle text or flatten a color device. Pass --placements none for a single base ad.`,
             note: "Populate as you go: for each [TODO] ingest slot, source its real asset and wire it into the slot right away \u2014 one at a time, not all sourced first then reconciled at the end. For any LOGO slot, use the company's committed brand logo from `src/brand/logos/` (per BRAND.md's Logos map) \u2014 do NOT re-fetch it with `baker images logo <domain>` when a brand logo exists, because the Brandfetch CDN often serves an OUTDATED mark and that is how an old logo ships on a live ad. When every slot is filled, `baker canvas validate` then `baker canvas run`. Running generates a billed image \u2014 it is not free."
           }