@mcptoolshop/style-dataset-lab 1.1.0

package/scripts/generate-ipadapter.js

@@ -0,0 +1,339 @@
+#!/usr/bin/env node
+
+/**
+ * generate-ipadapter.js — IP-Adapter guided generation for structural subjects.
+ *
+ * Uses a reference image to guide the shape/structure of the output
+ * while the text prompt adds identity-specific details (wear, patching, name).
+ *
+ * Usage:
+ *   node scripts/generate-ipadapter.js --subject corrigan --ref <image> --seeds 4
+ *   node scripts/generate-ipadapter.js --subject corrigan --ref <image> --seeds 4 --dry-run
+ *
+ * Options:
+ *   --subject NAME    Subject name for output file naming
+ *   --ref PATH        Reference image path (the structural truth source)
+ *   --prompt TEXT      Generation prompt
+ *   --negative TEXT    Negative prompt
+ *   --seeds N          Number of seeds (default: 4)
+ *   --weight W         IP-Adapter weight (default: 0.55)
+ *   --start S          IP-Adapter start (default: 0.0)
+ *   --end E            IP-Adapter end (default: 0.8)
+ *   --dry-run          Print workflow without running
+ */
+
+import { readFile, writeFile, mkdir } from "node:fs/promises";
+import { join } from "node:path";
+
+const COMFY_URL = process.env.COMFY_URL || "http://127.0.0.1:8188";
+const REPO_ROOT = new URL("..", import.meta.url).pathname.replace(/^\/([A-Z]:)/, "$1");
+const GAME = process.argv.find((a, i) => process.argv[i - 1] === '--game') || 'star-freight';
+const GAME_ROOT = join(REPO_ROOT, 'games', GAME);
+
+const DEFAULTS = {
+  checkpoint: "dreamshaperXL_v21TurboDPMSDE.safetensors",
+  loras: [{ name: "classipeintxl_v21.safetensors", weight: 1.0 }],
+  ipadapter_model: "ip-adapter-plus_sdxl_vit-h.safetensors",
+  clip_vision_model: "CLIP-ViT-H-14-laion2B-s32B-b79K.safetensors",
+  steps: 10,
+  cfg: 2.5,
+  sampler: "dpmpp_sde",
+  scheduler: "karras",
+  width: 1024,
+  height: 1024,
+  base_seed: 27200,
+};
+
+function parseArgs() {
+  const args = process.argv.slice(2);
+  const get = (flag, def) => {
+    const i = args.indexOf(flag);
+    return i >= 0 && i + 1 < args.length ? args[i + 1] : def;
+  };
+  return {
+    subject: get("--subject", "unknown"),
+    refPath: get("--ref", null),
+    prompt: get("--prompt", null),
+    negative: get("--negative", null),
+    seeds: parseInt(get("--seeds", "4"), 10),
+    weight: parseFloat(get("--weight", "0.55")),
+    start: parseFloat(get("--start", "0.0")),
+    end: parseFloat(get("--end", "0.8")),
+    dryRun: args.includes("--dry-run"),
+  };
+}
+
+function buildIPAdapterWorkflow(prompt, negativePrompt, seed, refImageName, weight, start, end) {
+  const nodes = {};
+  let nextId = 1;
+
+  // Checkpoint
+  const ckptId = String(nextId++);
+  nodes[ckptId] = {
+    class_type: "CheckpointLoaderSimple",
+    inputs: { ckpt_name: DEFAULTS.checkpoint },
+  };
+
+  let modelOut = [ckptId, 0];
+  let clipOut = [ckptId, 1];
+
+  // LoRA
+  for (const lora of DEFAULTS.loras) {
+    const loraId = String(nextId++);
+    nodes[loraId] = {
+      class_type: "LoraLoader",
+      inputs: {
+        model: modelOut,
+        clip: clipOut,
+        lora_name: lora.name,
+        strength_model: lora.weight,
+        strength_clip: lora.weight,
+      },
+    };
+    modelOut = [loraId, 0];
+    clipOut = [loraId, 1];
+  }
+
+  // Load CLIP Vision
+  const clipVisionId = String(nextId++);
+  nodes[clipVisionId] = {
+    class_type: "CLIPVisionLoader",
+    inputs: { clip_name: DEFAULTS.clip_vision_model },
+  };
+
+  // Load IP-Adapter model
+  const ipaLoadId = String(nextId++);
+  nodes[ipaLoadId] = {
+    class_type: "IPAdapterModelLoader",
+    inputs: { ipadapter_file: DEFAULTS.ipadapter_model },
+  };
+
+  // Load reference image
+  const refImgId = String(nextId++);
+  nodes[refImgId] = {
+    class_type: "LoadImage",
+    inputs: { image: refImageName },
+  };
+
+  // Apply IP-Adapter (using IPAdapterAdvanced from ComfyUI_IPAdapter_plus)
+  const ipaApplyId = String(nextId++);
+  nodes[ipaApplyId] = {
+    class_type: "IPAdapterAdvanced",
+    inputs: {
+      model: modelOut,
+      ipadapter: [ipaLoadId, 0],
+      image: [refImgId, 0],
+      clip_vision: [clipVisionId, 0],
+      weight: weight,
+      weight_type: "linear",
+      combine_embeds: "concat",
+      start_at: start,
+      end_at: end,
+      embeds_scaling: "V only",
+    },
+  };
+  modelOut = [ipaApplyId, 0];
+
+  // Positive prompt
+  const posId = String(nextId++);
+  nodes[posId] = {
+    class_type: "CLIPTextEncode",
+    inputs: { text: prompt, clip: clipOut },
+  };
+
+  // Negative prompt
+  const negId = String(nextId++);
+  nodes[negId] = {
+    class_type: "CLIPTextEncode",
+    inputs: { text: negativePrompt, clip: clipOut },
+  };
+
+  // Empty latent
+  const latentId = String(nextId++);
+  nodes[latentId] = {
+    class_type: "EmptyLatentImage",
+    inputs: { width: DEFAULTS.width, height: DEFAULTS.height, batch_size: 1 },
+  };
+
+  // KSampler
+  const samplerId = String(nextId++);
+  nodes[samplerId] = {
+    class_type: "KSampler",
+    inputs: {
+      model: modelOut,
+      positive: [posId, 0],
+      negative: [negId, 0],
+      latent_image: [latentId, 0],
+      seed,
+      steps: DEFAULTS.steps,
+      cfg: DEFAULTS.cfg,
+      sampler_name: DEFAULTS.sampler,
+      scheduler: DEFAULTS.scheduler,
+      denoise: 1.0,
+    },
+  };
+
+  // VAE Decode
+  const vaeId = String(nextId++);
+  nodes[vaeId] = {
+    class_type: "VAEDecode",
+    inputs: { samples: [samplerId, 0], vae: [ckptId, 2] },
+  };
+
+  // Save
+  const saveId = String(nextId++);
+  nodes[saveId] = {
+    class_type: "SaveImage",
+    inputs: { images: [vaeId, 0], filename_prefix: "ipa" },
+  };
+
+  return nodes;
+}
+
+async function comfyHealth() {
+  try {
+    const res = await fetch(`${COMFY_URL}/system_stats`);
+    return res.ok;
+  } catch {
+    return false;
+  }
+}
+
+const MAX_POLL_MS = 600_000; // 10 minutes
+
+async function submitAndWait(workflow) {
+  const clientId = `ipa-${Date.now()}`;
+  const res = await fetch(`${COMFY_URL}/prompt`, {
+    method: "POST",
+    headers: { "Content-Type": "application/json" },
+    body: JSON.stringify({ prompt: workflow, client_id: clientId }),
+  });
+  if (!res.ok) {
+    const text = await res.text();
+    throw new Error(`ComfyUI submit failed: ${res.status} ${text}`);
+  }
+  const { prompt_id: promptId } = await res.json();
+  const start = Date.now();
+  while (true) {
+    if (Date.now() - start > MAX_POLL_MS) throw new Error(`ComfyUI poll timeout after ${MAX_POLL_MS/1000}s for prompt ${promptId}`);
+    await new Promise((r) => setTimeout(r, 1000));
+    const histRes = await fetch(`${COMFY_URL}/history/${promptId}`);
+    if (!histRes.ok) continue;
+    const history = await histRes.json();
+    const entry = history[promptId];
+    if (!entry) continue;
+    if (entry.status?.completed) return entry;
+    if (entry.status?.status_str === "error") {
+      throw new Error(`ComfyUI error: ${JSON.stringify(entry.status)}`);
+    }
+  }
+}
+
+async function downloadImage(filename, subfolder, type = "output") {
+  const url = `${COMFY_URL}/view?filename=${encodeURIComponent(filename)}&subfolder=${encodeURIComponent(subfolder || "")}&type=${type}`;
+  const res = await fetch(url);
+  if (!res.ok) throw new Error(`Download failed: ${res.status}`);
+  return Buffer.from(await res.arrayBuffer());
+}
+
+async function main() {
+  const opts = parseArgs();
+
+  if (!opts.refPath) {
+    console.error("Error: --ref <path> is required");
+    process.exit(1);
+  }
+  if (!opts.prompt) {
+    console.error("Error: --prompt is required");
+    process.exit(1);
+  }
+
+  const negative = opts.negative ||
+    "photorealistic, photograph, 3d render, anime, cartoon, text, watermark, blurry, low quality";
+
+  console.log("\x1b[1mstyle-dataset-lab\x1b[0m IP-Adapter structural discovery");
+  console.log(`  Subject: ${opts.subject}`);
+  console.log(`  Reference: ${opts.refPath}`);
+  console.log(`  IP-Adapter: ${DEFAULTS.ipadapter_model}`);
+  console.log(`  Weight: ${opts.weight}, Range: ${opts.start}-${opts.end}`);
+  console.log(`  Seeds: ${opts.seeds}`);
+  console.log("");
+
+  if (!opts.dryRun) {
+    const online = await comfyHealth();
+    if (!online) {
+      console.error("\x1b[31mError:\x1b[0m ComfyUI not reachable");
+      process.exit(1);
+    }
+    console.log("\x1b[32m+\x1b[0m ComfyUI online");
+
+    // Upload reference image
+    const refData = await readFile(join(GAME_ROOT, opts.refPath));
+    const refFilename = `ipa_ref_${opts.subject}.png`;
+    const formData = new FormData();
+    formData.append("image", new Blob([refData], { type: "image/png" }), refFilename);
+    formData.append("overwrite", "true");
+    const uploadRes = await fetch(`${COMFY_URL}/upload/image`, { method: "POST", body: formData });
+    if (!uploadRes.ok) {
+      console.error(`Reference upload failed: ${uploadRes.status}`);
+      process.exit(1);
+    }
+    const uploadResult = await uploadRes.json();
+    console.log(`\x1b[32m+\x1b[0m Reference uploaded: ${uploadResult.name}`);
+  }
+
+  await mkdir(join(GAME_ROOT, "outputs/candidates"), { recursive: true });
+
+  const refFilename = `ipa_ref_${opts.subject}.png`;
+
+  for (let si = 0; si < opts.seeds; si++) {
+    const seed = DEFAULTS.base_seed + si;
+    const assetId = `${opts.subject}_ipa_s${si}`;
+    const destPath = `outputs/candidates/${assetId}.png`;
+
+    console.log(`  [${si + 1}/${opts.seeds}] ${assetId} (seed: ${seed})`);
+
+    if (opts.dryRun) continue;
+
+    const startMs = Date.now();
+    const workflow = buildIPAdapterWorkflow(
+      opts.prompt, negative, seed,
+      refFilename, opts.weight, opts.start, opts.end
+    );
+
+    try {
+      const result = await submitAndWait(workflow);
+      const elapsed = Date.now() - startMs;
+
+      const outputs = result.outputs || {};
+      let imageFile = null;
+      let imageSubfolder = "";
+      for (const nodeOut of Object.values(outputs)) {
+        if (nodeOut.images?.length > 0) {
+          imageFile = nodeOut.images[0].filename;
+          imageSubfolder = nodeOut.images[0].subfolder || "";
+          break;
+        }
+      }
+
+      if (!imageFile) {
+        console.log("    \x1b[33m!\x1b[0m No output image");
+        continue;
+      }
+
+      const imgData = await downloadImage(imageFile, imageSubfolder);
+      await writeFile(join(GAME_ROOT, destPath), imgData);
+      console.log(`    \x1b[32m+\x1b[0m ${destPath} (${elapsed}ms, ${imgData.length} bytes)`);
+    } catch (err) {
+      console.log(`    \x1b[31mx\x1b[0m ${err.message}`);
+    }
+  }
+
+  console.log("\n\x1b[32m+\x1b[0m IP-Adapter discovery complete");
+  if (opts.dryRun) console.log("  (dry run)");
+}
+
+main().catch((err) => {
+  console.error(err);
+  process.exit(1);
+});

package/scripts/generate.js

@@ -0,0 +1,351 @@
+#!/usr/bin/env node
+
+/**
+ * generate.js — Drive ComfyUI to produce candidate assets from a prompt pack.
+ *
+ * Usage: node scripts/generate.js [--prompt-pack inputs/prompts/rpg-icons-lane1.json]
+ *
+ * For each subject × variation, submits a workflow to ComfyUI HTTP API,
+ * downloads the output PNG, and writes a provenance record to records/.
+ */
+
+import { readFile, writeFile, mkdir } from "node:fs/promises";
+import { join, basename } from "node:path";
+
+const COMFY_URL = process.env.COMFY_URL || "http://127.0.0.1:8188";
+const REPO_ROOT = new URL("..", import.meta.url).pathname.replace(/^\/([A-Z]:)/, "$1");
+const GAME = process.argv.find((a, i) => process.argv[i - 1] === '--game') || 'star-freight';
+const GAME_ROOT = join(REPO_ROOT, 'games', GAME);
+
+// ── ComfyUI API helpers ──
+
+async function comfyHealth() {
+  try {
+    const res = await fetch(`${COMFY_URL}/system_stats`);
+    return res.ok;
+  } catch {
+    return false;
+  }
+}
+
+function buildWorkflow(prompt, negativePrompt, checkpoint, loras, seed, steps, cfg, sampler, scheduler, width, height, ipAdapterConfig) {
+  const nodes = {};
+  let nextId = 1;
+
+  // Checkpoint loader
+  const ckptId = String(nextId++);
+  nodes[ckptId] = {
+    class_type: "CheckpointLoaderSimple",
+    inputs: { ckpt_name: checkpoint },
+  };
+
+  let modelOut = [ckptId, 0];
+  let clipOut = [ckptId, 1];
+
+  // LoRA loaders (chained)
+  for (const lora of loras) {
+    const loraId = String(nextId++);
+    nodes[loraId] = {
+      class_type: "LoraLoader",
+      inputs: {
+        model: modelOut,
+        clip: clipOut,
+        lora_name: lora.name,
+        strength_model: lora.weight,
+        strength_clip: lora.clip_weight ?? lora.weight,
+      },
+    };
+    modelOut = [loraId, 0];
+    clipOut = [loraId, 1];
+  }
+
+  // IP-Adapter for style reference (if configured)
+  if (ipAdapterConfig) {
+    // Load CLIP Vision model
+    const clipVisionId = String(nextId++);
+    nodes[clipVisionId] = {
+      class_type: "CLIPVisionLoader",
+      inputs: { clip_name: ipAdapterConfig.clip_vision_model },
+    };
+
+    // Load IP-Adapter model
+    const ipaLoaderId = String(nextId++);
+    nodes[ipaLoaderId] = {
+      class_type: "IPAdapterModelLoader",
+      inputs: { ipadapter_file: ipAdapterConfig.model },
+    };
+
+    // Load reference image
+    const refImageId = String(nextId++);
+    nodes[refImageId] = {
+      class_type: "LoadImage",
+      inputs: { image: ipAdapterConfig.reference_image },
+    };
+
+    // Apply IP-Adapter
+    const ipaApplyId = String(nextId++);
+    nodes[ipaApplyId] = {
+      class_type: "IPAdapterApply",
+      inputs: {
+        ipadapter: [ipaLoaderId, 0],
+        clip_vision: [clipVisionId, 0],
+        image: [refImageId, 0],
+        model: modelOut,
+        weight: ipAdapterConfig.weight || 0.6,
+        noise: ipAdapterConfig.noise || 0.0,
+        weight_type: ipAdapterConfig.weight_type || "linear",
+        start_at: ipAdapterConfig.start_at || 0.0,
+        end_at: ipAdapterConfig.end_at || 1.0,
+      },
+    };
+
+    modelOut = [ipaApplyId, 0];
+  }
+
+  // CLIP Text Encode — positive
+  const posId = String(nextId++);
+  nodes[posId] = {
+    class_type: "CLIPTextEncode",
+    inputs: { text: prompt, clip: clipOut },
+  };
+
+  // CLIP Text Encode — negative
+  const negId = String(nextId++);
+  nodes[negId] = {
+    class_type: "CLIPTextEncode",
+    inputs: { text: negativePrompt, clip: clipOut },
+  };
+
+  // Empty latent image
+  const latentId = String(nextId++);
+  nodes[latentId] = {
+    class_type: "EmptyLatentImage",
+    inputs: { width, height, batch_size: 1 },
+  };
+
+  // KSampler
+  const samplerId = String(nextId++);
+  nodes[samplerId] = {
+    class_type: "KSampler",
+    inputs: {
+      model: modelOut,
+      positive: [posId, 0],
+      negative: [negId, 0],
+      latent_image: [latentId, 0],
+      seed,
+      steps,
+      cfg,
+      sampler_name: sampler,
+      scheduler,
+      denoise: 1.0,
+    },
+  };
+
+  // VAE Decode
+  const vaeId = String(nextId++);
+  nodes[vaeId] = {
+    class_type: "VAEDecode",
+    inputs: {
+      samples: [samplerId, 0],
+      vae: [ckptId, 2],
+    },
+  };
+
+  // Save Image
+  const saveId = String(nextId++);
+  nodes[saveId] = {
+    class_type: "SaveImage",
+    inputs: {
+      images: [vaeId, 0],
+      filename_prefix: "sdl",
+    },
+  };
+
+  return { nodes, saveNodeId: saveId };
+}
+
+const MAX_POLL_MS = 600_000; // 10 minutes
+
+async function submitAndWait(workflow) {
+  const clientId = `sdl-${Date.now()}`;
+
+  // Submit
+  const res = await fetch(`${COMFY_URL}/prompt`, {
+    method: "POST",
+    headers: { "Content-Type": "application/json" },
+    body: JSON.stringify({ prompt: workflow, client_id: clientId }),
+  });
+
+  if (!res.ok) {
+    const text = await res.text();
+    throw new Error(`ComfyUI submit failed: ${res.status} ${text}`);
+  }
+
+  const { prompt_id: promptId } = await res.json();
+
+  // Poll history until complete
+  const start = Date.now();
+  while (true) {
+    if (Date.now() - start > MAX_POLL_MS) throw new Error(`ComfyUI poll timeout after ${MAX_POLL_MS/1000}s for prompt ${promptId}`);
+    await new Promise((r) => setTimeout(r, 1000));
+    const histRes = await fetch(`${COMFY_URL}/history/${promptId}`);
+    if (!histRes.ok) continue;
+    const history = await histRes.json();
+    const entry = history[promptId];
+    if (!entry) continue;
+    if (entry.status?.completed) {
+      return entry;
+    }
+    if (entry.status?.status_str === "error") {
+      throw new Error(`ComfyUI generation failed: ${JSON.stringify(entry.status)}`);
+    }
+  }
+}
+
+async function downloadImage(filename, subfolder, type = "output") {
+  const url = `${COMFY_URL}/view?filename=${encodeURIComponent(filename)}&subfolder=${encodeURIComponent(subfolder || "")}&type=${type}`;
+  const res = await fetch(url);
+  if (!res.ok) throw new Error(`Download failed: ${res.status}`);
+  return Buffer.from(await res.arrayBuffer());
+}
+
+// ── Main ──
+
+async function main() {
+  const args = process.argv.slice(2);
+  const packPath = args.find((a) => !a.startsWith("--")) || "inputs/prompts/rpg-icons-lane1.json";
+  const dryRun = args.includes("--dry-run");
+
+  const fullPackPath = join(GAME_ROOT, packPath);
+  const pack = JSON.parse(await readFile(fullPackPath, "utf-8"));
+
+  console.log(`\x1b[1mstyle-dataset-lab\x1b[0m generate`);
+  console.log(`  Lane: ${pack.lane}`);
+  console.log(`  Subjects: ${pack.subjects.length}`);
+  console.log(`  Variations: ${pack.variations.length}`);
+  console.log(`  Total candidates: ${pack.subjects.length * pack.variations.length}`);
+  console.log("");
+
+  if (!dryRun) {
+    const online = await comfyHealth();
+    if (!online) {
+      console.error("\x1b[31mError:\x1b[0m ComfyUI not reachable at " + COMFY_URL);
+      console.error("Start ComfyUI: cd F:/AI-Models/ComfyUI-runtime && python main.py --listen 127.0.0.1 --port 8188");
+      process.exit(1);
+    }
+    console.log("\x1b[32m✓\x1b[0m ComfyUI online");
+  }
+
+  await mkdir(join(GAME_ROOT, "outputs/candidates"), { recursive: true });
+  await mkdir(join(GAME_ROOT, "records"), { recursive: true });
+
+  const d = pack.defaults;
+  let generated = 0;
+  const totalExpected = pack.subjects.length * pack.variations.length;
+
+  for (const subject of pack.subjects) {
+    for (const variation of variation_list(pack.variations)) {
+      const assetId = `${subject.id}_${variation.id}`;
+      const seed = (d.base_seed + generated) + (variation.seed_offset || 0);
+      const fullPrompt = `${pack.style_prefix}, ${subject.prompt}`;
+      const loras = variation.loras || [];
+
+      console.log(`  [${generated + 1}/${totalExpected}] ${assetId} (seed: ${seed}, loras: ${loras.length})`);
+
+      if (dryRun) {
+        generated++;
+        continue;
+      }
+
+      const startMs = Date.now();
+      const { nodes } = buildWorkflow(
+        fullPrompt, pack.negative,
+        d.checkpoint, loras, seed,
+        d.steps, d.cfg, d.sampler, d.scheduler,
+        d.width, d.height,
+      );
+
+      try {
+        const result = await submitAndWait(nodes);
+        const elapsed = Date.now() - startMs;
+
+        // Find output image
+        const outputs = result.outputs || {};
+        let imageFile = null;
+        let imageSubfolder = "";
+        for (const nodeOut of Object.values(outputs)) {
+          if (nodeOut.images && nodeOut.images.length > 0) {
+            imageFile = nodeOut.images[0].filename;
+            imageSubfolder = nodeOut.images[0].subfolder || "";
+            break;
+          }
+        }
+
+        if (!imageFile) {
+          console.log(`    \x1b[33m⚠\x1b[0m No output image found`);
+          generated++;
+          continue;
+        }
+
+        // Download and save
+        const imgData = await downloadImage(imageFile, imageSubfolder);
+        const destPath = `outputs/candidates/${assetId}.png`;
+        await writeFile(join(GAME_ROOT, destPath), imgData);
+
+        // Write provenance record
+        const record = {
+          id: assetId,
+          schema_version: "1.0.0",
+          created_at: new Date().toISOString(),
+          asset_path: destPath,
+          image: { format: "png", width: d.width, height: d.height, bytes: imgData.length },
+          provenance: {
+            workflow_id: pack.lane,
+            workflow_version: "1.0.0",
+            checkpoint: d.checkpoint,
+            loras,
+            prompt: fullPrompt,
+            negative_prompt: pack.negative,
+            seed,
+            steps: d.steps,
+            cfg: d.cfg,
+            sampler: d.sampler,
+            scheduler: d.scheduler,
+            width: d.width,
+            height: d.height,
+            generation_time_ms: elapsed,
+            gpu_model: "RTX 5080",
+            batch_index: generated,
+          },
+          judgment: null,
+          canon: null,
+          iteration: null,
+        };
+
+        await writeFile(
+          join(GAME_ROOT, `records/${assetId}.json`),
+          JSON.stringify(record, null, 2),
+        );
+
+        console.log(`    \x1b[32m✓\x1b[0m ${destPath} (${elapsed}ms, ${imgData.length} bytes)`);
+      } catch (err) {
+        console.log(`    \x1b[31m✗\x1b[0m ${err.message}`);
+      }
+
+      generated++;
+    }
+  }
+
+  console.log("");
+  console.log(`\x1b[32m✓\x1b[0m Generated ${generated} candidates`);
+  if (dryRun) console.log("  (dry run — no images produced)");
+}
+
+function variation_list(variations) {
+  return variations;
+}
+
+main().catch((err) => {
+  console.error(err);
+  process.exit(1);
+});