itsmebasil

---
name: Pattern Jewellery Automation
description: Automates jewellery product marketing using Google Vertex AI (Gemini and Imagen) and Google Drive.
---

# Pattern Jewellery Automation Skill

## Overview
This skill automates the creation of high-end marketing content for Pattern Jewellery products. It orchestrates a sophisticated multi-agent pipeline: securely ingesting raw product photos, generating lifestyle and studio images via Imagen 3, writing SEO-optimized copy via Gemini 1.5 Pro, and systematically organizing the final assets in Google Drive. 

---

## 📥 Input Schema
The skill expects a trigger payload with the following fields:
- `product_image` (String): URL or base64 string of the raw product photograph.
- `product_details` (Object):
  - `name`: Product title (e.g. "Diamond Blue Sapphire Ring")
  - `sku`: Unique identifier (e.g. "R4389")
  - `category`: Organization category (e.g. "rings")
  - `material`: Composition (e.g. "18K white gold, 0.32ct diamond")
  - `price_now`: Current retail price (e.g. 4455)
  - `description`: Core design breakdown.

## 📤 Output Schema
- `model_image_url`: Link to the generated lifestyle model image.
- `product_image_url`: Link to the generated product-only image.
- `caption`: Formatted Instagram caption highlighting the luxury aesthetic.
- `hashtags`: Array of 20 optimized tags.
- `drive_link`: Public/Internal Google Drive folder URL hosting all generated assets.

---

## ⚙️ Workflow Execution Steps

### 1. Vision & Prompt Generation (Gemini 1.5 Pro)
The system visually analyzes the `product_image` alongside the `product_details` to determine design intricacy, materials, and aesthetic quality. It then outputs two strictly constrained prompts:
* **Model Prompt (Max 120 Tokens):** A lifestyle photograph prompt targeting the Gulf luxury market. It details an elegant model wearing the piece in an upscale interior (e.g., modern Dubai), with specific studio lighting and bokeh settings.
* **Product Prompt (Max 120 Tokens):** A premium product-only photography prompt placing the piece on luxury backgrounds (e.g., white Carrara marble, deep navy velvet) equipped with three-point studio lighting and macro lens specs.

### 2. Parallel Image Generation (Imagen 3)
Using Google Vertex AI, this step dispatches parallel requests:
* Generates the `model_image` (1 sample, ultra quality, 4:5 aspect ratio, adult generation allowed).
* Generates the `product_image` (1 sample, ultra quality, 1:1 aspect ratio, tack-sharp).

### 3. Parallel Content Generation (Gemini 1.5 Pro)
Concurrently with the image rendering, the LLM drafts an engaging Instagram caption matching Pattern Jewellery's aspirational and traditional-modern fusion tone. It seamlessly integrates the price point and structural details, returning the copy alongside a 20-tag hashtag package.

### 4. Storage & Compilation (Google Drive)
All final image bytes (`.jpg`) and text output (`.txt`) are piped into the Google Drive API. They are systematically uploaded into a structured directory constraint: `/Pattern_Jewellery/{category}/{sku}/`. 

---

## 🧠 Memory Rules & State Management
This skill utilizes a persistent, 4-field memory map to iteratively improve generation over time based on user feedback. The core keys are:
- `style`: Default is "editorial"
- `tone`: Default is "aspirational-luxury"
- `background_preference`: Default is "white-marble"
- `top_performing_caption`: Cached high-performing copy for tone-matching.
*These variables dynamically inject into the prompt generation templates (Step 1).*

## ⚡ Caching Protocol
To minimize unnecessary GPU execution costs:
1. Incoming images are hashed (SHA-256 or pHash).
2. Lookups occur against a Redis/Local cache mapping.
3. If the exact same image and metadata payload are received within the TTL window (30 days), the pipeline bypasses Gemini/Imagen entirely and immediately returns the cached Google Drive URL.

## 📂 Bundled Files
- `jewellery_openclaw_skill.json`: The core JSON pipeline graph mapped to OpenCLAW UI.
- `jewellery_openclaw_skill.py`: Background FastAPI worker capable of executing the pipeline outside of OpenCLAW. 
- `pattern_jewellery_openclaw_system.html`: Front-end architectural diagram and design blueprint.

FILE:jewellery_openclaw_skill.py
import os
import io
import asyncio
import hashlib
from typing import Dict, Any, List, Optional
from fastapi import FastAPI, HTTPException, BackgroundTasks
from pydantic import BaseModel
from tenacity import retry, stop_after_attempt, wait_exponential
import httpx

# AI and Cloud providers
from anthropic import AsyncAnthropic
import vertexai
from vertexai.preview.vision_models import ImageGenerationModel
from googleapiclient.discovery import build
from google.oauth2 import service_account
from googleapiclient.http import MediaIoBaseUpload
import redis.asyncio as redis

app = FastAPI(title="Jewellery Content Engine (OpenCLAW Skill)")

# -------------------------------------------------------------------
# Configuration & Initialization
# -------------------------------------------------------------------
# Environment variables expected:
# ANTHROPIC_API_KEY
# GOOGLE_APPLICATION_CREDENTIALS (path to service account json)
# REDIS_URL (e.g. redis://localhost:6379)
# GCP_PROJECT_ID
# GCP_LOCATION (e.g., us-central1)

aclient = AsyncAnthropic(api_key=os.getenv("ANTHROPIC_API_KEY"))

try:
    vertexai.init(
        project=os.getenv("GCP_PROJECT_ID"), 
        location=os.getenv("GCP_LOCATION", "us-central1")
    )
except Exception as e:
    print(f"Vertex AI Init Warning: {e}")

try:
    redis_client = redis.from_url(os.getenv("REDIS_URL", "redis://localhost:6379"), decode_responses=True)
except Exception as e:
    print(f"Redis Init Warning: {e}")

try:
    creds = service_account.Credentials.from_service_account_file(
        os.getenv("GOOGLE_APPLICATION_CREDENTIALS", "gcp-service-account.json"), 
        scopes=['https://www.googleapis.com/auth/drive.file']
    )
    drive_service = build('drive', 'v3', credentials=creds)
except Exception as e:
    print(f"Drive Init Warning: {e}")

# -------------------------------------------------------------------
# Data Models
# -------------------------------------------------------------------
class WebhookPayload(BaseModel):
    image_url: str
    product_name: str
    metadata: Optional[Dict[str, Any]] = {}

# -------------------------------------------------------------------
# Core Functions
# -------------------------------------------------------------------
async def download_image(url: str) -> bytes:
    async with httpx.AsyncClient() as client:
        response = await client.get(url)
        response.raise_for_status()
        return response.content

def get_image_hash(image_bytes: bytes) -> str:
    return hashlib.sha256(image_bytes).hexdigest()

@retry(stop=stop_after_attempt(3), wait=wait_exponential(multiplier=1, min=2, max=10))
async def generate_prompts(image_url: str, metadata: dict) -> dict:
    prompt = f"Analyze this jewellery piece context: {metadata}. Provide ONLY a JSON with two keys: 'model_prompt' (describing an AI model wearing it naturally) and 'product_prompt' (studio lighting for the product alone)."
    
    response = await aclient.messages.create(
        model="claude-3-5-sonnet-20241022",
        max_tokens=500,
        system="You are an expert jewellery art director.",
        messages=[{
            "role": "user",
            "content": [
                {"type": "image", "source": {"type": "url", "url": image_url}},
                {"type": "text", "text": prompt}
            ]
        }]
    )
    # Simplified JSON extraction
    import json
    return json.loads(response.content[0].text)

@retry(stop=stop_after_attempt(3), wait=wait_exponential(multiplier=1, min=2, max=10))
async def generate_content(product_name: str, metadata: dict) -> str:
    prompt = f"Write a catchy Instagram caption and a short SEO product description for {product_name}. Details: {metadata}"
    response = await aclient.messages.create(
        model="claude-3-5-sonnet-20241022",
        max_tokens=600,
        system="You are a luxury jewellery copywriter.",
        messages=[{"role": "user", "content": prompt}]
    )
    return response.content[0].text

@retry(stop=stop_after_attempt(3), wait=wait_exponential(multiplier=2, min=4, max=15))
async def generate_imagen_asset(prompt: str) -> bytes:
    loop = asyncio.get_event_loop()
    model = ImageGenerationModel.from_pretrained("imagen-3.0-generate-001")
    def _gen():
        return model.generate_images(
            prompt=prompt,
            number_of_images=1,
            aspect_ratio="1:1",
            person_generation="ALLOW_ADULT"
        )
    response = await loop.run_in_executor(None, _gen)
    return response.images[0]._image_bytes

@retry(stop=stop_after_attempt(3), wait=wait_exponential(multiplier=1, min=2, max=10))
def upload_to_drive(file_bytes_or_text, file_name: str, folder_id: str, mime_type: str):
    if isinstance(file_bytes_or_text, str):
        file_bytes_or_text = file_bytes_or_text.encode('utf-8')
        
    file_stream = io.BytesIO(file_bytes_or_text)
    media = MediaIoBaseUpload(file_stream, mimetype=mime_type, resumable=True)
    
    file_metadata = {'name': file_name, 'parents': [folder_id]}
    file = drive_service.files().create(body=file_metadata, media_body=media, fields='id, webViewLink').execute()
    return file.get('webViewLink')

def create_drive_folder(folder_name: str, parent_id: str = None) -> str:
    file_metadata = {
        'name': folder_name,
        'mimeType': 'application/vnd.google-apps.folder'
    }
    if parent_id:
        file_metadata['parents'] = [parent_id]
        
    file = drive_service.files().create(body=file_metadata, fields='id').execute()
    return file.get('id')

# -------------------------------------------------------------------
# Pipeline Execution Definition
# -------------------------------------------------------------------
async def run_pipeline(payload: WebhookPayload, folder_id_root: str):
    try:
        # Step 1: Input & Hash Check
        image_bytes = await download_image(payload.image_url)
        img_hash = get_image_hash(image_bytes)
        
        cahed_result = await redis_client.get(f"jewellery_pipeline:{img_hash}")
        if cahed_result:
            return {"status": "success", "cached": True, "links": cahed_result}

        # Step 2: Prompt Generation
        prompts = await generate_prompts(payload.image_url, payload.metadata)

        # Step 3: Parallel Generation
        model_img_task = asyncio.create_task(generate_imagen_asset(prompts['model_prompt']))
        prod_img_task = asyncio.create_task(generate_imagen_asset(prompts['product_prompt']))
        content_task = asyncio.create_task(generate_content(payload.product_name, payload.metadata))

        model_img_bytes, prod_img_bytes, text_content = await asyncio.gather(
            model_img_task, prod_img_task, content_task
        )

        # Step 4: Storage
        loop = asyncio.get_event_loop()
        folder_id = await loop.run_in_executor(None, create_drive_folder, payload.product_name, folder_id_root)
        
        links = {}
        links['model'] = await loop.run_in_executor(None, upload_to_drive, model_img_bytes, "model_image.jpg", folder_id, "image/jpeg")
        links['product'] = await loop.run_in_executor(None, upload_to_drive, prod_img_bytes, "product_image.jpg", folder_id, "image/jpeg")
        links['content'] = await loop.run_in_executor(None, upload_to_drive, text_content, "content.txt", folder_id, "text/plain")

        # Step 5: Save to Cache
        import json
        await redis_client.setex(f"jewellery_pipeline:{img_hash}", 2592000, json.dumps(links))

        return {"status": "success", "links": links}

    except Exception as e:
        return {"status": "error", "message": str(e)}

# -------------------------------------------------------------------
# OpenCLAW Webhook Trigger
# -------------------------------------------------------------------
@app.post("/api/v1/jewellery/process")
async def process_jewellery_content(payload: WebhookPayload):
    # This acts as the synchronous webhook return, processing in background if needed
    # Wait for completion to return real results for integration ease.
    import os
    parent_folder = os.getenv("DRIVE_FOLDER_ID", "root") 
    result = await run_pipeline(payload, parent_folder)
    
    if result["status"] == "error":
        raise HTTPException(status_code=500, detail=result["message"])
        
    return result

if __name__ == "__main__":
    import uvicorn
    uvicorn.run(app, host="0.0.0.0", port=8000)

FILE:pattern_jewellery_openclaw_system.html
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>Pattern Jewellery — OpenClaw Skill System</title>
<link href="https://fonts.googleapis.com/css2?family=Cormorant+Garamond:ital,wght@0,300;0,400;0,600;1,300;1,400&family=DM+Mono:wght@300;400;500&display=swap" rel="stylesheet">
<style>
  :root {
    --gold: #b8963e;
    --gold-light: #d4b06a;
    --gold-pale: #f5ecd7;
    --navy: #0e1628;
    --navy-mid: #1a2540;
    --slate: #2c3a55;
    --ink: #111827;
    --mist: #f8f6f2;
    --border: rgba(184,150,62,0.25);
    --tag-bg: rgba(184,150,62,0.1);
  }

  *, *::before, *::after { box-sizing: border-box; margin: 0; padding: 0; }

  body {
    font-family: 'DM Mono', monospace;
    background: var(--navy);
    color: #e8e0d4;
    min-height: 100vh;
    line-height: 1.6;
  }

  /* ── HEADER ── */
  header {
    background: linear-gradient(135deg, var(--navy) 0%, var(--navy-mid) 60%, #1c2e50 100%);
    border-bottom: 1px solid var(--border);
    padding: 3rem 4rem 2.5rem;
    position: relative;
    overflow: hidden;
  }
  header::before {
    content: '';
    position: absolute;
    top: -60px; right: -60px;
    width: 280px; height: 280px;
    border-radius: 50%;
    border: 1px solid rgba(184,150,62,0.12);
  }
  header::after {
    content: '';
    position: absolute;
    top: -30px; right: -30px;
    width: 180px; height: 180px;
    border-radius: 50%;
    border: 1px solid rgba(184,150,62,0.08);
  }
  .header-eyebrow {
    font-family: 'DM Mono', monospace;
    font-size: 0.65rem;
    letter-spacing: 0.25em;
    text-transform: uppercase;
    color: var(--gold);
    margin-bottom: 0.75rem;
  }
  header h1 {
    font-family: 'Cormorant Garamond', serif;
    font-weight: 300;
    font-size: 2.6rem;
    letter-spacing: 0.02em;
    color: #fff;
    line-height: 1.1;
    margin-bottom: 0.5rem;
  }
  header h1 span { color: var(--gold); font-style: italic; }
  .header-sub {
    font-size: 0.72rem;
    color: rgba(232,224,212,0.5);
    letter-spacing: 0.1em;
  }
  .product-badge {
    display: inline-flex;
    align-items: center;
    gap: 0.5rem;
    margin-top: 1.5rem;
    padding: 0.5rem 1rem;
    background: var(--tag-bg);
    border: 1px solid var(--border);
    border-radius: 2px;
    font-size: 0.68rem;
    color: var(--gold-light);
    letter-spacing: 0.12em;
  }
  .dot { width: 6px; height: 6px; background: var(--gold); border-radius: 50%; }

  /* ── LAYOUT ── */
  main { max-width: 1200px; margin: 0 auto; padding: 3rem 4rem; }

  /* ── SECTION ── */
  .component {
    margin-bottom: 4rem;
    border: 1px solid var(--border);
    border-radius: 4px;
    overflow: hidden;
  }
  .component-header {
    display: flex;
    align-items: center;
    gap: 1rem;
    padding: 1rem 1.5rem;
    background: rgba(184,150,62,0.07);
    border-bottom: 1px solid var(--border);
  }
  .component-num {
    font-family: 'Cormorant Garamond', serif;
    font-size: 1.4rem;
    font-weight: 300;
    color: var(--gold);
    font-style: italic;
    min-width: 2rem;
  }
  .component-title {
    font-size: 0.7rem;
    letter-spacing: 0.2em;
    text-transform: uppercase;
    color: #c8bfb0;
  }
  .component-tag {
    margin-left: auto;
    padding: 0.2rem 0.6rem;
    background: var(--tag-bg);
    border: 1px solid var(--border);
    border-radius: 2px;
    font-size: 0.6rem;
    color: var(--gold);
    letter-spacing: 0.1em;
  }

  /* ── CODE BLOCK ── */
  .code-wrap { padding: 1.5rem; background: #080e1a; overflow-x: auto; }
  pre {
    font-family: 'DM Mono', monospace;
    font-size: 0.72rem;
    line-height: 1.7;
    color: #c8d0e0;
    white-space: pre;
  }
  .tok-key   { color: #7eb8f7; }
  .tok-str   { color: #a8d9a0; }
  .tok-num   { color: #f0c060; }
  .tok-bool  { color: #e08080; }
  .tok-cmt   { color: #5a6a8a; font-style: italic; }
  .tok-gold  { color: var(--gold-light); }

  /* ── PROSE SECTION ── */
  .prose-wrap { padding: 1.5rem 2rem; }
  .prose-wrap p { font-size: 0.78rem; color: #b8b0a4; line-height: 1.8; margin-bottom: 0.8rem; }

  /* ── TEMPLATES ── */
  .template-block {
    padding: 1.5rem 2rem;
    border-bottom: 1px solid rgba(184,150,62,0.1);
  }
  .template-block:last-child { border-bottom: none; }
  .template-label {
    font-size: 0.6rem;
    letter-spacing: 0.2em;
    text-transform: uppercase;
    color: var(--gold);
    margin-bottom: 0.75rem;
  }
  .template-text {
    font-family: 'DM Mono', monospace;
    font-size: 0.72rem;
    line-height: 1.8;
    color: #d0c8be;
    background: #080e1a;
    padding: 1rem 1.25rem;
    border-left: 2px solid var(--gold);
    white-space: pre-wrap;
  }
  .token-pill {
    display: inline-block;
    margin-top: 0.6rem;
    padding: 0.2rem 0.7rem;
    background: rgba(184,150,62,0.12);
    border: 1px solid rgba(184,150,62,0.3);
    border-radius: 20px;
    font-size: 0.6rem;
    color: var(--gold-light);
    letter-spacing: 0.1em;
  }
  .var-pill {
    color: #7eb8f7;
    font-style: italic;
  }

  /* ── WORKFLOW DIAGRAM ── */
  .workflow-wrap { padding: 2rem; }
  .flow-row {
    display: flex;
    align-items: stretch;
    gap: 0;
    margin-bottom: 0;
  }
  .flow-row.parallel { align-items: flex-start; margin-left: 3rem; }
  .agent-box {
    flex: 1;
    border: 1px solid var(--border);
    border-radius: 3px;
    padding: 1rem 1.25rem;
    background: #0d1525;
    position: relative;
  }
  .agent-box .a-label {
    font-size: 0.58rem;
    letter-spacing: 0.18em;
    text-transform: uppercase;
    color: var(--gold);
    margin-bottom: 0.3rem;
  }
  .agent-box .a-name {
    font-family: 'Cormorant Garamond', serif;
    font-size: 1rem;
    color: #fff;
    margin-bottom: 0.5rem;
  }
  .agent-box .a-io {
    font-size: 0.65rem;
    color: #7a8aa0;
    line-height: 1.6;
  }
  .arrow {
    display: flex;
    align-items: center;
    padding: 0 0.5rem;
    color: var(--gold);
    font-size: 1.2rem;
    flex-shrink: 0;
  }
  .arrow-down {
    text-align: center;
    color: var(--gold);
    font-size: 1rem;
    padding: 0.4rem 0;
    margin-left: 2rem;
  }
  .parallel-label {
    font-size: 0.58rem;
    letter-spacing: 0.15em;
    text-transform: uppercase;
    color: rgba(184,150,62,0.5);
    margin-bottom: 0.5rem;
    margin-left: 3rem;
  }
  .flow-join {
    margin-left: 3rem;
    font-size: 0.65rem;
    color: #5a6a8a;
    padding: 0.3rem 0;
  }
  .data-tag {
    display: inline-block;
    padding: 0.15rem 0.4rem;
    background: rgba(126,184,247,0.08);
    border: 1px solid rgba(126,184,247,0.2);
    border-radius: 2px;
    font-size: 0.6rem;
    color: #7eb8f7;
    margin: 0.1rem 0;
    font-style: italic;
  }

  /* ── MEMORY / CACHE ── */
  .rule-list { padding: 1.5rem 2rem; }
  .rule-item {
    display: flex;
    gap: 0.75rem;
    margin-bottom: 0.7rem;
    align-items: flex-start;
  }
  .rule-icon { color: var(--gold); flex-shrink: 0; font-size: 0.7rem; padding-top: 0.05rem; }
  .rule-text { font-size: 0.73rem; color: #b0a898; line-height: 1.6; }
  .rule-text strong { color: #d4ccc4; font-weight: 500; }

  /* ── TOKEN BUDGET ── */
  .budget-grid {
    display: grid;
    grid-template-columns: repeat(auto-fit, minmax(200px, 1fr));
    gap: 1px;
    background: var(--border);
    border: 1px solid var(--border);
    margin: 1.5rem 2rem;
  }
  .budget-cell {
    background: #0d1525;
    padding: 1rem 1.25rem;
  }
  .budget-agent { font-size: 0.6rem; letter-spacing: 0.15em; text-transform: uppercase; color: #5a7a9a; margin-bottom: 0.3rem; }
  .budget-tokens {
    font-family: 'Cormorant Garamond', serif;
    font-size: 2rem;
    font-weight: 300;
    color: var(--gold);
    line-height: 1;
  }
  .budget-label { font-size: 0.6rem; color: #5a6a7a; margin-top: 0.2rem; }
  .budget-total { background: rgba(184,150,62,0.08); }
  .budget-total .budget-tokens { color: #fff; }

  /* ── FOOTER ── */
  footer {
    text-align: center;
    padding: 2rem;
    border-top: 1px solid var(--border);
    font-size: 0.62rem;
    color: #3a4a60;
    letter-spacing: 0.12em;
    text-transform: uppercase;
  }
</style>
</head>
<body>

<header>
  <div class="header-eyebrow">OpenClaw Agent System · Production-Ready</div>
  <h1>Pattern Jewellery <span>Automation</span></h1>
  <div class="header-sub">skill_schema · tool_definitions · prompt_templates · workflow · memory · cache · token_optimization</div>
  <div class="product-badge">
    <span class="dot"></span>
    R4389 · Diamond &amp; Blue Sapphire Ring · 18K White Gold · AED 4,455
  </div>
</header>

<main>

<!-- ═══════════════════════════════════════════════════
     COMPONENT 1: SKILL SCHEMA
═══════════════════════════════════════════════════ -->
<section class="component">
  <div class="component-header">
    <span class="component-num">01</span>
    <span class="component-title">Skill Schema</span>
    <span class="component-tag">JSON</span>
  </div>
  <div class="code-wrap"><pre>{
  <span class="tok-key">"skill_name"</span>: <span class="tok-str">"pattern_jewellery_automation"</span>,
  <span class="tok-key">"version"</span>: <span class="tok-str">"1.0.0"</span>,
  <span class="tok-key">"description"</span>: <span class="tok-str">"Automates jewellery product marketing: enhances images, generates lifestyle and product visuals, creates SEO content, and stores outputs in Google Drive for Pattern Jewellery."</span>,

  <span class="tok-key">"pattern_jewellery_specs"</span>: {
    <span class="tok-key">"intricate_design_focus"</span>: <span class="tok-bool">true</span>,
    <span class="tok-key">"high_detail_textures"</span>: <span class="tok-bool">true</span>,
    <span class="tok-key">"fusion_aesthetic"</span>: <span class="tok-str">"traditional-modern"</span>,
    <span class="tok-key">"target_market"</span>: <span class="tok-str">"Gulf luxury"</span>,
    <span class="tok-key">"hallmark"</span>: <span class="tok-str">"750"</span>
  },

  <span class="tok-key">"input_schema"</span>: {
    <span class="tok-key">"product_image"</span>: {
      <span class="tok-key">"type"</span>: <span class="tok-str">"string"</span>,
      <span class="tok-key">"format"</span>: <span class="tok-str">"url_or_base64"</span>,
      <span class="tok-key">"required"</span>: <span class="tok-bool">true</span>,
      <span class="tok-key">"description"</span>: <span class="tok-str">"Raw product photograph from studio shoot"</span>
    },
    <span class="tok-key">"product_details"</span>: {
      <span class="tok-key">"type"</span>: <span class="tok-str">"object"</span>,
      <span class="tok-key">"required"</span>: <span class="tok-bool">true</span>,
      <span class="tok-key">"properties"</span>: {
        <span class="tok-key">"name"</span>:        { <span class="tok-key">"type"</span>: <span class="tok-str">"string"</span>, <span class="tok-key">"example"</span>: <span class="tok-str">"Diamond Blue Sapphire Ring"</span> },
        <span class="tok-key">"sku"</span>:         { <span class="tok-key">"type"</span>: <span class="tok-str">"string"</span>, <span class="tok-key">"example"</span>: <span class="tok-str">"R4389"</span> },
        <span class="tok-key">"category"</span>:    { <span class="tok-key">"type"</span>: <span class="tok-str">"string"</span>, <span class="tok-key">"example"</span>: <span class="tok-str">"rings"</span> },
        <span class="tok-key">"material"</span>:    { <span class="tok-key">"type"</span>: <span class="tok-str">"string"</span>, <span class="tok-key">"example"</span>: <span class="tok-str">"18K white gold 2.24gm, natural diamond 0.32ct, blue sapphire 0.78ct"</span> },
        <span class="tok-key">"price_now"</span>:   { <span class="tok-key">"type"</span>: <span class="tok-str">"number"</span>, <span class="tok-key">"example"</span>: <span class="tok-num">4455</span> },
        <span class="tok-key">"price_was"</span>:   { <span class="tok-key">"type"</span>: <span class="tok-str">"number"</span>, <span class="tok-key">"example"</span>: <span class="tok-num">8990</span> },
        <span class="tok-key">"description"</span>: { <span class="tok-key">"type"</span>: <span class="tok-str">"string"</span>, <span class="tok-key">"example"</span>: <span class="tok-str">"Oval blue sapphire centrepiece with double diamond halo, prong-set in 18K white gold, hallmarked 750"</span> }
      }
    }
  },

  <span class="tok-key">"output_schema"</span>: {
    <span class="tok-key">"model_image_url"</span>:   { <span class="tok-key">"type"</span>: <span class="tok-str">"string"</span>, <span class="tok-key">"description"</span>: <span class="tok-str">"URL of generated lifestyle model image"</span> },
    <span class="tok-key">"product_image_url"</span>: { <span class="tok-key">"type"</span>: <span class="tok-str">"string"</span>, <span class="tok-key">"description"</span>: <span class="tok-str">"URL of generated premium product-only image"</span> },
    <span class="tok-key">"caption"</span>:           { <span class="tok-key">"type"</span>: <span class="tok-str">"string"</span>, <span class="tok-key">"description"</span>: <span class="tok-str">"SEO-optimized social media caption"</span> },
    <span class="tok-key">"hashtags"</span>:          { <span class="tok-key">"type"</span>: <span class="tok-str">"array"</span>, <span class="tok-key">"items"</span>: { <span class="tok-key">"type"</span>: <span class="tok-str">"string"</span> } },
    <span class="tok-key">"drive_link"</span>:        { <span class="tok-key">"type"</span>: <span class="tok-str">"string"</span>, <span class="tok-key">"description"</span>: <span class="tok-str">"Google Drive folder URL for all assets"</span> }
  },

  <span class="tok-key">"steps"</span>: [
    {
      <span class="tok-key">"step"</span>: <span class="tok-num">1</span>,
      <span class="tok-key">"id"</span>: <span class="tok-str">"prompt_generation"</span>,
      <span class="tok-key">"agent"</span>: <span class="tok-str">"creative_agent"</span>,
      <span class="tok-key">"depends_on"</span>: <span class="tok-str">"vision_agent"</span>,
      <span class="tok-key">"description"</span>: <span class="tok-str">"Generate model and product image prompts from visual analysis and memory"</span>
    },
    {
      <span class="tok-key">"step"</span>: <span class="tok-num">2</span>,
      <span class="tok-key">"id"</span>: <span class="tok-str">"image_generation"</span>,
      <span class="tok-key">"agent"</span>: <span class="tok-str">"vision_agent"</span>,
      <span class="tok-key">"tool"</span>: <span class="tok-str">"imagen_api"</span>,
      <span class="tok-key">"parallel"</span>: <span class="tok-bool">true</span>,
      <span class="tok-key">"description"</span>: <span class="tok-str">"Generate model lifestyle image and premium product image concurrently"</span>
    },
    {
      <span class="tok-key">"step"</span>: <span class="tok-num">3</span>,
      <span class="tok-key">"id"</span>: <span class="tok-str">"content_generation"</span>,
      <span class="tok-key">"agent"</span>: <span class="tok-str">"content_agent"</span>,
      <span class="tok-key">"tool"</span>: <span class="tok-str">"claude_llm"</span>,
      <span class="tok-key">"parallel"</span>: <span class="tok-bool">true</span>,
      <span class="tok-key">"description"</span>: <span class="tok-str">"Generate SEO caption and hashtags concurrently with image generation"</span>
    },
    {
      <span class="tok-key">"step"</span>: <span class="tok-num">4</span>,
      <span class="tok-key">"id"</span>: <span class="tok-str">"storage"</span>,
      <span class="tok-key">"agent"</span>: <span class="tok-str">"storage_agent"</span>,
      <span class="tok-key">"tool"</span>: <span class="tok-str">"drive_api"</span>,
      <span class="tok-key">"depends_on"</span>: [<span class="tok-str">"image_generation"</span>, <span class="tok-str">"content_generation"</span>],
      <span class="tok-key">"description"</span>: <span class="tok-str">"Upload all assets to organized Google Drive folder"</span>
    }
  ]
}</pre></div>
</section>

<!-- ═══════════════════════════════════════════════════
     COMPONENT 2: TOOL DEFINITIONS
═══════════════════════════════════════════════════ -->
<section class="component">
  <div class="component-header">
    <span class="component-num">02</span>
    <span class="component-title">Tool Definitions</span>
    <span class="component-tag">JSON</span>
  </div>
  <div class="code-wrap"><pre>[
  {
    <span class="tok-key">"tool_name"</span>: <span class="tok-str">"claude_llm"</span>,
    <span class="tok-key">"description"</span>: <span class="tok-str">"Anthropic Claude for prompt generation and social content creation"</span>,
    <span class="tok-key">"api_endpoint"</span>: <span class="tok-str">"https://api.anthropic.com/v1/messages"</span>,
    <span class="tok-key">"auth_method"</span>: <span class="tok-str">"bearer_token"</span>,
    <span class="tok-key">"auth_header"</span>: <span class="tok-str">"x-api-key"</span>,
    <span class="tok-key">"model"</span>: <span class="tok-str">"claude-3-5-haiku-20241022"</span>,
    <span class="tok-key">"functions"</span>: [
      {
        <span class="tok-key">"name"</span>: <span class="tok-str">"generate_image_prompt"</span>,
        <span class="tok-key">"description"</span>: <span class="tok-str">"Generate optimized Imagen prompts for model and product visuals"</span>,
        <span class="tok-key">"max_tokens"</span>: <span class="tok-num">300</span>,
        <span class="tok-key">"input_schema"</span>: {
          <span class="tok-key">"image_analysis"</span>: {
            <span class="tok-key">"type"</span>: <span class="tok-str">"object"</span>,
            <span class="tok-key">"fields"</span>: [<span class="tok-str">"design_type"</span>, <span class="tok-str">"material_detected"</span>, <span class="tok-str">"quality_score"</span>]
          },
          <span class="tok-key">"product_details"</span>: {
            <span class="tok-key">"type"</span>: <span class="tok-str">"object"</span>,
            <span class="tok-key">"fields"</span>: [<span class="tok-str">"name"</span>, <span class="tok-str">"material"</span>, <span class="tok-str">"description"</span>]
          },
          <span class="tok-key">"memory"</span>: {
            <span class="tok-key">"type"</span>: <span class="tok-str">"object"</span>,
            <span class="tok-key">"fields"</span>: [<span class="tok-str">"style"</span>, <span class="tok-str">"background_preference"</span>]
          }
        },
        <span class="tok-key">"output_schema"</span>: {
          <span class="tok-key">"model_prompt"</span>: { <span class="tok-key">"type"</span>: <span class="tok-str">"string"</span>, <span class="tok-key">"max_tokens"</span>: <span class="tok-num">120</span> },
          <span class="tok-key">"product_prompt"</span>: { <span class="tok-key">"type"</span>: <span class="tok-str">"string"</span>, <span class="tok-key">"max_tokens"</span>: <span class="tok-num">120</span> }
        },
        <span class="tok-key">"response_format"</span>: <span class="tok-str">"json"</span>
      },
      {
        <span class="tok-key">"name"</span>: <span class="tok-str">"generate_social_content"</span>,
        <span class="tok-key">"description"</span>: <span class="tok-str">"Generate SEO-optimized Instagram caption and hashtag array"</span>,
        <span class="tok-key">"max_tokens"</span>: <span class="tok-num">220</span>,
        <span class="tok-key">"input_schema"</span>: {
          <span class="tok-key">"product_details"</span>: {
            <span class="tok-key">"type"</span>: <span class="tok-str">"object"</span>,
            <span class="tok-key">"fields"</span>: [<span class="tok-str">"name"</span>, <span class="tok-str">"price_now"</span>, <span class="tok-str">"price_was"</span>, <span class="tok-str">"material"</span>, <span class="tok-str">"description"</span>]
          },
          <span class="tok-key">"memory"</span>: {
            <span class="tok-key">"type"</span>: <span class="tok-str">"object"</span>,
            <span class="tok-key">"fields"</span>: [<span class="tok-str">"tone"</span>, <span class="tok-str">"top_performing_caption"</span>]
          }
        },
        <span class="tok-key">"output_schema"</span>: {
          <span class="tok-key">"caption"</span>:  { <span class="tok-key">"type"</span>: <span class="tok-str">"string"</span>, <span class="tok-key">"max_chars"</span>: <span class="tok-num">220</span> },
          <span class="tok-key">"hashtags"</span>: { <span class="tok-key">"type"</span>: <span class="tok-str">"array"</span>, <span class="tok-key">"max_items"</span>: <span class="tok-num">20</span> }
        },
        <span class="tok-key">"response_format"</span>: <span class="tok-str">"json"</span>
      }
    ]
  },

  {
    <span class="tok-key">"tool_name"</span>: <span class="tok-str">"imagen_api"</span>,
    <span class="tok-key">"description"</span>: <span class="tok-str">"Google Imagen 3 for high-fidelity jewellery image generation"</span>,
    <span class="tok-key">"api_endpoint"</span>: <span class="tok-str">"https://us-central1-aiplatform.googleapis.com/v1/projects/{project_id}/locations/us-central1/publishers/google/models/imagen-3.0-generate-002:predict"</span>,
    <span class="tok-key">"auth_method"</span>: <span class="tok-str">"oauth2_service_account"</span>,
    <span class="tok-key">"model_version"</span>: <span class="tok-str">"imagen-3.0-generate-002"</span>,
    <span class="tok-key">"functions"</span>: [
      {
        <span class="tok-key">"name"</span>: <span class="tok-str">"generate_model_image"</span>,
        <span class="tok-key">"description"</span>: <span class="tok-str">"Generate lifestyle shot of model wearing jewellery"</span>,
        <span class="tok-key">"input_schema"</span>: {
          <span class="tok-key">"prompt"</span>:       { <span class="tok-key">"type"</span>: <span class="tok-str">"string"</span>, <span class="tok-key">"required"</span>: <span class="tok-bool">true</span> },
          <span class="tok-key">"aspect_ratio"</span>: { <span class="tok-key">"type"</span>: <span class="tok-str">"string"</span>, <span class="tok-key">"default"</span>: <span class="tok-str">"4:5"</span>, <span class="tok-key">"enum"</span>: [<span class="tok-str">"1:1"</span>, <span class="tok-str">"4:5"</span>, <span class="tok-str">"9:16"</span>] },
          <span class="tok-key">"quality"</span>:      { <span class="tok-key">"type"</span>: <span class="tok-str">"string"</span>, <span class="tok-key">"default"</span>: <span class="tok-str">"ultra"</span> },
          <span class="tok-key">"sample_count"</span>: { <span class="tok-key">"type"</span>: <span class="tok-str">"integer"</span>, <span class="tok-key">"default"</span>: <span class="tok-num">1</span> }
        },
        <span class="tok-key">"output_schema"</span>: {
          <span class="tok-key">"image_url"</span>: { <span class="tok-key">"type"</span>: <span class="tok-str">"string"</span>, <span class="tok-key">"format"</span>: <span class="tok-str">"gcs_uri"</span> }
        }
      },
      {
        <span class="tok-key">"name"</span>: <span class="tok-str">"generate_product_image"</span>,
        <span class="tok-key">"description"</span>: <span class="tok-str">"Generate premium product-only image on luxury background"</span>,
        <span class="tok-key">"input_schema"</span>: {
          <span class="tok-key">"prompt"</span>:       { <span class="tok-key">"type"</span>: <span class="tok-str">"string"</span>, <span class="tok-key">"required"</span>: <span class="tok-bool">true</span> },
          <span class="tok-key">"aspect_ratio"</span>: { <span class="tok-key">"type"</span>: <span class="tok-str">"string"</span>, <span class="tok-key">"default"</span>: <span class="tok-str">"1:1"</span>, <span class="tok-key">"enum"</span>: [<span class="tok-str">"1:1"</span>, <span class="tok-str">"4:5"</span>] },
          <span class="tok-key">"quality"</span>:      { <span class="tok-key">"type"</span>: <span class="tok-str">"string"</span>, <span class="tok-key">"default"</span>: <span class="tok-str">"ultra"</span> },
          <span class="tok-key">"sample_count"</span>: { <span class="tok-key">"type"</span>: <span class="tok-str">"integer"</span>, <span class="tok-key">"default"</span>: <span class="tok-num">1</span> }
        },
        <span class="tok-key">"output_schema"</span>: {
          <span class="tok-key">"image_url"</span>: { <span class="tok-key">"type"</span>: <span class="tok-str">"string"</span>, <span class="tok-key">"format"</span>: <span class="tok-str">"gcs_uri"</span> }
        }
      }
    ]
  },

  {
    <span class="tok-key">"tool_name"</span>: <span class="tok-str">"drive_api"</span>,
    <span class="tok-key">"description"</span>: <span class="tok-str">"Google Drive for organized asset storage"</span>,
    <span class="tok-key">"api_endpoint"</span>: <span class="tok-str">"https://www.googleapis.com/upload/drive/v3/files"</span>,
    <span class="tok-key">"auth_method"</span>: <span class="tok-str">"oauth2_service_account"</span>,
    <span class="tok-key">"scopes"</span>: [<span class="tok-str">"https://www.googleapis.com/auth/drive.file"</span>],
    <span class="tok-key">"folder_structure"</span>: <span class="tok-str">"/Pattern_Jewellery/{product_category}/{YYYY-MM-DD}/{sku}/"</span>,
    <span class="tok-key">"functions"</span>: [
      {
        <span class="tok-key">"name"</span>: <span class="tok-str">"upload_to_drive"</span>,
        <span class="tok-key">"description"</span>: <span class="tok-str">"Upload file to structured Drive folder, return shareable link"</span>,
        <span class="tok-key">"input_schema"</span>: {
          <span class="tok-key">"file_data"</span>:   { <span class="tok-key">"type"</span>: <span class="tok-str">"string"</span>, <span class="tok-key">"format"</span>: <span class="tok-str">"binary_or_gcs_uri"</span> },
          <span class="tok-key">"folder_path"</span>: { <span class="tok-key">"type"</span>: <span class="tok-str">"string"</span>, <span class="tok-key">"example"</span>: <span class="tok-str">"/Pattern_Jewellery/rings/2025-01-15/R4389/"</span> },
          <span class="tok-key">"file_name"</span>:   { <span class="tok-key">"type"</span>: <span class="tok-str">"string"</span>, <span class="tok-key">"example"</span>: <span class="tok-str">"R4389_model_lifestyle.jpg"</span> },
          <span class="tok-key">"mime_type"</span>:   { <span class="tok-key">"type"</span>: <span class="tok-str">"string"</span>, <span class="tok-key">"default"</span>: <span class="tok-str">"image/jpeg"</span> }
        },
        <span class="tok-key">"output_schema"</span>: {
          <span class="tok-key">"drive_link"</span>: { <span class="tok-key">"type"</span>: <span class="tok-str">"string"</span>, <span class="tok-key">"format"</span>: <span class="tok-str">"https://drive.google.com/drive/folders/{folder_id}"</span> },
          <span class="tok-key">"file_id"</span>:    { <span class="tok-key">"type"</span>: <span class="tok-str">"string"</span> }
        }
      }
    ]
  }
]</pre></div>
</section>

<!-- ═══════════════════════════════════════════════════
     COMPONENT 3: PROMPT TEMPLATES
═══════════════════════════════════════════════════ -->
<section class="component">
  <div class="component-header">
    <span class="component-num">03</span>
    <span class="component-title">Prompt Templates</span>
    <span class="component-tag">3 Templates</span>
  </div>

  <div class="template-block">
    <div class="template-label">Template 1 · Model Lifestyle Image Prompt · Max 120 tokens</div>
    <div class="template-text">Professional luxury jewellery lifestyle photograph, Gulf market. Elegant Middle Eastern woman, refined pose, manicured hand foregrounded. Wearing <span class="var-pill">{product_name}</span> crafted in <span class="var-pill">{material}</span>. Macro detail: <span class="var-pill">{design_details}</span>. Setting: modern Dubai interior, soft bokeh, neutral silk backdrop. Studio lighting with rim highlight on stone facets and prong setting. Editorial fashion quality. 8K, photorealistic, no text.</div>
    <div class="token-pill">≈ 78 tokens · within 120-token limit ✓</div>
  </div>

  <div class="template-block">
    <div class="template-label">Template 2 · Premium Product-Only Image Prompt · Max 120 tokens</div>
    <div class="template-text">Luxury jewellery product photography, no model. <span class="var-pill">{product_name}</span> in <span class="var-pill">{material}</span>. Key features: <span class="var-pill">{key_features}</span>. Surface: white Carrara marble with deep navy velvet accent. Macro lens, tack-sharp stone facets and prong detail. Three-point studio lighting: soft key, hard fill catching diamond brilliance, warm backlight. Floating shadow. 8K, photorealistic, commercial retail grade, no text.</div>
    <div class="token-pill">≈ 80 tokens · within 120-token limit ✓</div>
  </div>

  <div class="template-block">
    <div class="template-label">Template 3 · Social Media Content Prompt · Max 180 tokens</div>
    <div class="template-text">Generate Instagram content for Pattern Jewellery, Dubai. Luxury tone, traditional-modern fusion voice.

Product: <span class="var-pill">{product_name}</span>
Price: AED <span class="var-pill">{price_now}</span> (was AED <span class="var-pill">{price_was}</span>)
USP: <span class="var-pill">{unique_selling_points}</span>

Return JSON only:
{
  "caption": "2–3 sentence luxury caption, highlight offer, end with CTA (link in bio / DM to order)",
  "hashtags": ["array", "of", "20", "mixed", "broad+niche", "tags", "no", "spaces"]
}

Brand rules: aspirational, never salesy. Reference craftsmanship. Include price drop naturally.</div>
    <div class="token-pill">≈ 102 tokens · within 180-token limit ✓</div>
  </div>
</section>

<!-- ═══════════════════════════════════════════════════
     COMPONENT 4: MULTI-AGENT WORKFLOW DIAGRAM
═══════════════════════════════════════════════════ -->
<section class="component">
  <div class="component-header">
    <span class="component-num">04</span>
    <span class="component-title">Multi-Agent Workflow Design</span>
    <span class="component-tag">4 Agents</span>
  </div>

  <div class="workflow-wrap">

    <!-- AGENT 1: VISION -->
    <div class="flow-row">
      <div class="agent-box" style="max-width:380px;">
        <div class="a-label">Step 0 · Sequential</div>
        <div class="a-name">Vision Agent</div>
        <div class="a-io">
          IN → <span class="data-tag">product_image (url)</span><br>
          TOOL → Internal vision model<br>
          OUT → <span class="data-tag">design_type</span> <span class="data-tag">material_detected</span> <span class="data-tag">quality_score</span>
        </div>
      </div>
    </div>

    <div class="arrow-down">↓ image_analysis (3 fields only)</div>

    <!-- AGENT 2: CREATIVE -->
    <div class="flow-row">
      <div class="agent-box" style="max-width:380px;">
        <div class="a-label">Step 1 · Sequential</div>
        <div class="a-name">Creative Agent</div>
        <div class="a-io">
          IN → <span class="data-tag">image_analysis</span> <span class="data-tag">product_details</span> <span class="data-tag">memory.style</span> <span class="data-tag">memory.background_preference</span><br>
          TOOL → claude_llm :: generate_image_prompt<br>
          OUT → <span class="data-tag">model_prompt</span> <span class="data-tag">product_prompt</span>
        </div>
      </div>
    </div>

    <div class="arrow-down">↓ prompts dispatched in parallel to Steps 2 &amp; 3</div>

    <!-- PARALLEL: AGENTS 3 + CONTENT -->
    <div class="parallel-label">⟳ Parallel Execution (Steps 2 &amp; 3)</div>
    <div class="flow-row" style="gap:1rem; max-width:820px;">
      <div class="agent-box">
        <div class="a-label">Step 2 · Parallel</div>
        <div class="a-name">Image Agent (Vision)</div>
        <div class="a-io">
          IN → <span class="data-tag">model_prompt</span> <span class="data-tag">product_prompt</span><br>
          TOOL → imagen_api :: generate_model_image<br>
          &nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; imagen_api :: generate_product_image<br>
          OUT → <span class="data-tag">model_image_url</span> <span class="data-tag">product_image_url</span>
        </div>
      </div>
      <div class="arrow" style="color:#3a4a60;">|</div>
      <div class="agent-box">
        <div class="a-label">Step 3 · Parallel</div>
        <div class="a-name">Content Agent</div>
        <div class="a-io">
          IN → <span class="data-tag">product_details</span> <span class="data-tag">memory.tone</span> <span class="data-tag">memory.top_performing_caption</span><br>
          TOOL → claude_llm :: generate_social_content<br>
          OUT → <span class="data-tag">caption</span> <span class="data-tag">hashtags[]</span>
        </div>
      </div>
    </div>

    <div class="flow-join">→ await both parallel steps complete</div>
    <div class="arrow-down">↓ all 4 outputs collected</div>

    <!-- AGENT 4: STORAGE -->
    <div class="flow-row">
      <div class="agent-box" style="max-width:500px;">
        <div class="a-label">Step 4 · Sequential</div>
        <div class="a-name">Storage Agent</div>
        <div class="a-io">
          IN → <span class="data-tag">model_image_url</span> <span class="data-tag">product_image_url</span> <span class="data-tag">caption</span> <span class="data-tag">hashtags[]</span> <span class="data-tag">product_details.sku</span> <span class="data-tag">product_details.category</span><br>
          TOOL → drive_api :: upload_to_drive (×3: model img, product img, metadata.json)<br>
          FOLDER → /Pattern_Jewellery/rings/2025-01-15/R4389/<br>
          OUT → <span class="data-tag">drive_link (folder URL)</span>
        </div>
      </div>
    </div>

    <div class="arrow-down">↓ final skill output returned to user</div>

  </div>
</section>

<!-- ═══════════════════════════════════════════════════
     COMPONENT 5: MEMORY SYSTEM
═══════════════════════════════════════════════════ -->
<section class="component">
  <div class="component-header">
    <span class="component-num">05</span>
    <span class="component-title">Memory System Structure</span>
    <span class="component-tag">JSON + Rules</span>
  </div>
  <div class="code-wrap"><pre><span class="tok-cmt">// memory.json — persistent, lightweight, 4 fields max</span>
{
  <span class="tok-key">"style"</span>:                  <span class="tok-str">"editorial"</span>,
  <span class="tok-key">"tone"</span>:                   <span class="tok-str">"aspirational-luxury"</span>,
  <span class="tok-key">"background_preference"</span>:  <span class="tok-str">"white-marble"</span>,
  <span class="tok-key">"top_performing_caption"</span>: <span class="tok-str">"Timeless elegance, crafted for you. Now at AED 4,455 — link in bio."</span>
}

<span class="tok-cmt">// ── UPDATE RULES ──────────────────────────────────────────</span>
<span class="tok-cmt">// WHEN:  Only on explicit user correction or positive feedback signal</span>
<span class="tok-cmt">// HOW:   Replace field value atomically; no nested objects added</span>
<span class="tok-cmt">// WHERE: /system/memory/pattern_jewellery_memory.json (persistent)</span>
<span class="tok-cmt">// LIMIT: 4 fields hard cap — no expansion permitted</span>

<span class="tok-cmt">// ── INJECTION POINTS ──────────────────────────────────────</span>
<span class="tok-cmt">// Creative Agent receives: style + background_preference</span>
<span class="tok-cmt">// Content Agent receives:  tone + top_performing_caption</span>

<span class="tok-cmt">// ── BEFORE / AFTER EXAMPLE ────────────────────────────────</span>

<span class="tok-cmt">// BEFORE (default state):</span>
{
  <span class="tok-key">"style"</span>:                  <span class="tok-str">"editorial"</span>,
  <span class="tok-key">"tone"</span>:                   <span class="tok-str">"aspirational-luxury"</span>,
  <span class="tok-key">"background_preference"</span>:  <span class="tok-str">"white-marble"</span>,
  <span class="tok-key">"top_performing_caption"</span>: <span class="tok-str">""</span>
}

<span class="tok-cmt">// USER FEEDBACK: "I prefer dark navy velvet backgrounds, they performed better last week"</span>

<span class="tok-cmt">// AFTER (single field updated):</span>
{
  <span class="tok-key">"style"</span>:                  <span class="tok-str">"editorial"</span>,
  <span class="tok-key">"tone"</span>:                   <span class="tok-str">"aspirational-luxury"</span>,
  <span class="tok-key">"background_preference"</span>:  <span class="tok-str">"deep-navy-velvet"</span>,
  <span class="tok-key">"top_performing_caption"</span>: <span class="tok-str">""</span>
}
<span class="tok-cmt">// Creative Agent now injects "deep-navy-velvet" into Template 2 at {background_preference}</span></pre></div>
</section>

<!-- ═══════════════════════════════════════════════════
     COMPONENT 6: CACHING LOGIC
═══════════════════════════════════════════════════ -->
<section class="component">
  <div class="component-header">
    <span class="component-num">06</span>
    <span class="component-title">Caching Logic</span>
    <span class="component-tag">JSON + Flow</span>
  </div>
  <div class="code-wrap"><pre><span class="tok-cmt">// cache.json — Redis or local JSON, 30-day TTL</span>
{
  <span class="tok-key">"image_hashes"</span>: {
    <span class="tok-key">"a3f2c1b9_diamond-blue-sapphire-ring-r4389"</span>: {
      <span class="tok-key">"hash"</span>:       <span class="tok-str">"a3f2c1b9"</span>,
      <span class="tok-key">"created_at"</span>: <span class="tok-str">"2025-01-15T08:30:00Z"</span>,
      <span class="tok-key">"expires_at"</span>: <span class="tok-str">"2025-02-14T08:30:00Z"</span>
    }
  },
  <span class="tok-key">"enhanced_prompts"</span>: {
    <span class="tok-key">"a3f2c1b9_diamond-blue-sapphire-ring-r4389"</span>: {
      <span class="tok-key">"model_prompt"</span>:   <span class="tok-str">"Professional luxury jewellery lifestyle photograph..."</span>,
      <span class="tok-key">"product_prompt"</span>: <span class="tok-str">"Luxury jewellery product photography, no model..."</span>
    }
  },
  <span class="tok-key">"final_outputs"</span>: {
    <span class="tok-key">"a3f2c1b9_diamond-blue-sapphire-ring-r4389"</span>: {
      <span class="tok-key">"model_image_url"</span>:   <span class="tok-str">"https://drive.google.com/file/d/abc123"</span>,
      <span class="tok-key">"product_image_url"</span>: <span class="tok-str">"https://drive.google.com/file/d/def456"</span>,
      <span class="tok-key">"caption"</span>:           <span class="tok-str">"Timeless elegance, crafted for you..."</span>,
      <span class="tok-key">"hashtags"</span>:          [<span class="tok-str">"#PatternJewellery"</span>, <span class="tok-str">"#DubaiJewellery"</span>],
      <span class="tok-key">"drive_link"</span>:        <span class="tok-str">"https://drive.google.com/drive/folders/xyz789"</span>
    }
  }
}

<span class="tok-cmt">// ── CACHE KEY FORMAT ──────────────────────────────────────</span>
<span class="tok-cmt">// "{pHash_8char}_{product_name_slug}"</span>
<span class="tok-cmt">// Example: "a3f2c1b9_diamond-blue-sapphire-ring-r4389"</span>

<span class="tok-cmt">// ── DETECTION ALGORITHM ───────────────────────────────────</span>
<span class="tok-cmt">// Algorithm: pHash (perceptual hash) — tolerates JPEG compression artifacts</span>
<span class="tok-cmt">// Match threshold: Hamming distance ≤ 10 (identical/near-identical images)</span>
<span class="tok-cmt">// product_details change detection: SHA1 hash of JSON.stringify(product_details)</span>

<span class="tok-cmt">// ── LOGIC FLOW ────────────────────────────────────────────</span>
<span class="tok-cmt">// 1. Receive product_image + product_details</span>
<span class="tok-cmt">// 2. Compute pHash(product_image) → hash_key</span>
<span class="tok-cmt">// 3. Compute SHA1(product_details) → detail_fingerprint</span>
<span class="tok-cmt">// 4. cache_key = hash_key + "_" + slug(product_details.name + product_details.sku)</span>
<span class="tok-cmt">// 5. IF cache_key in final_outputs AND NOT expired AND detail_fingerprint unchanged:</span>
<span class="tok-cmt">//      RETURN cached final_outputs[cache_key]  ← skip all agents</span>
<span class="tok-cmt">// 6. ELSE IF cache_key in enhanced_prompts (image same, details changed):</span>
<span class="tok-cmt">//      SKIP vision_agent + creative_agent</span>
<span class="tok-cmt">//      RE-RUN content_agent + storage_agent only</span>
<span class="tok-cmt">// 7. ELSE (full miss):</span>
<span class="tok-cmt">//      EXECUTE full 4-step workflow</span>
<span class="tok-cmt">//      STORE all layers in cache</span>
<span class="tok-cmt">// 8. Expiration: TTL = 30 days. Manual invalidation: cache.delete(cache_key)</span></pre></div>
</section>

<!-- ═══════════════════════════════════════════════════
     COMPONENT 7: TOKEN OPTIMIZATION
═══════════════════════════════════════════════════ -->
<section class="component">
  <div class="component-header">
    <span class="component-num">07</span>
    <span class="component-title">Token Optimization Strategy</span>
    <span class="component-tag">Budget + Rules</span>
  </div>

  <div class="rule-list">
    <div class="rule-item"><span class="rule-icon">→</span><span class="rule-text"><strong>Short atomic prompts:</strong> Each agent call handles one task only. No multi-instruction prompts. Template variables replace repeated context text.</span></div>
    <div class="rule-item"><span class="rule-icon">→</span><span class="rule-text"><strong>Cached system prompts:</strong> System prompts for claude_llm are cached server-side using Anthropic prompt caching (cache_control: ephemeral). Not resent per call.</span></div>
    <div class="rule-item"><span class="rule-icon">→</span><span class="rule-text"><strong>JSON-only responses:</strong> All LLM outputs use response_format: json. Eliminates preamble, filler, and markdown tokens.</span></div>
    <div class="rule-item"><span class="rule-icon">→</span><span class="rule-text"><strong>Minimal image_analysis payload:</strong> Vision Agent outputs exactly 3 fields (design_type, material_detected, quality_score). No descriptions, no arrays, no nested objects.</span></div>
    <div class="rule-item"><span class="rule-icon">→</span><span class="rule-text"><strong>Selective memory injection:</strong> Creative Agent receives only style + background_preference. Content Agent receives only tone + top_performing_caption. No agent receives the full memory object.</span></div>
    <div class="rule-item"><span class="rule-icon">→</span><span class="rule-text"><strong>Parallel Step 2 + 3:</strong> image_generation and content_generation run concurrently. Eliminates sequential latency and avoids a round-trip token overhead for orchestration.</span></div>
    <div class="rule-item"><span class="rule-icon">→</span><span class="rule-text"><strong>max_tokens hard caps:</strong> claude_llm calls capped at 300 tokens (generate_image_prompt) and 220 tokens (generate_social_content). Model is prevented from over-generating.</span></div>
    <div class="rule-item"><span class="rule-icon">→</span><span class="rule-text"><strong>Cache-first on repeat images:</strong> Full cache hit skips all 4 agents (0 LLM tokens consumed). Partial hit (details changed) skips 2 agents, saving ~170 tokens.</span></div>
    <div class="rule-item"><span class="rule-icon">→</span><span class="rule-text"><strong>Product details passed as minimal subset:</strong> Each agent receives only the fields it requires — never the full product_details object.</span></div>
    <div class="rule-item"><span class="rule-icon">→</span><span class="rule-text"><strong>Memory limited to 4 string fields:</strong> No nested objects, arrays, or metadata stored. Keeps injection payload under 40 tokens per agent call.</span></div>
  </div>

  <div class="budget-grid">
    <div class="budget-cell">
      <div class="budget-agent">Vision Agent</div>
      <div class="budget-tokens">≤ 50</div>
      <div class="budget-label">tokens output (3-field JSON)</div>
    </div>
    <div class="budget-cell">
      <div class="budget-agent">Creative Agent</div>
      <div class="budget-tokens">≤ 240</div>
      <div class="budget-label">tokens output (120 × 2 prompts)</div>
    </div>
    <div class="budget-cell">
      <div class="budget-agent">Content Agent</div>
      <div class="budget-tokens">≤ 180</div>
      <div class="budget-label">tokens output (caption + hashtags)</div>
    </div>
    <div class="budget-cell">
      <div class="budget-agent">Storage Agent</div>
      <div class="budget-tokens">≤ 30</div>
      <div class="budget-label">tokens output (drive_link + file_ids)</div>
    </div>
    <div class="budget-cell budget-total">
      <div class="budget-agent">Total Per Execution</div>
      <div class="budget-tokens">≤ 500</div>
      <div class="budget-label">tokens (well under 600 budget)</div>
    </div>
    <div class="budget-cell">
      <div class="budget-agent">Cache Hit (full)</div>
      <div class="budget-tokens">0</div>
      <div class="budget-label">LLM tokens consumed</div>
    </div>
  </div>
</section>

</main>

<footer>
  Pattern Jewellery Automation · OpenClaw Skill System v1.0.0 · SKU R4389 · Diamond &amp; Blue Sapphire Ring · 18K White Gold
</footer>

</body>
</html>

FILE:jewellery_openclaw_skill.json
{
  "skill_name": "pattern_jewellery_vertex_automation",
  "version": "1.2.0",
  "description": "Automates jewellery product marketing using Google Vertex AI (Gemini for text, Imagen for visuals) and stores final assets in Google Drive. Enhances images, generates lifestyle/product visuals, creates SEO content, and organizes returning final unified assets.",
  
  "pattern_jewellery_specs": {
    "intricate_design_focus": true,
    "high_detail_textures": true,
    "fusion_aesthetic": "traditional-modern",
    "target_market": "Gulf luxury"
  },

  "input_schema": {
    "product_image": {
      "type": "string",
      "format": "url_or_base64",
      "required": true,
      "description": "Raw product photograph from studio shoot"
    },
    "product_details": {
      "type": "object",
      "required": true,
      "properties": {
        "name":        { "type": "string", "example": "Diamond Blue Sapphire Ring" },
        "sku":         { "type": "string", "example": "R4389" },
        "category":    { "type": "string", "example": "rings" },
        "material":    { "type": "string", "example": "18K white gold" },
        "price_now":   { "type": "number", "example": 4455 },
        "description": { "type": "string", "example": "Oval blue sapphire centrepiece with double diamond halo" }
      }
    }
  },

  "output_schema": {
    "model_image_url":   { "type": "string", "description": "URL of generated lifestyle model image" },
    "product_image_url": { "type": "string", "description": "URL of generated premium product-only image" },
    "caption":           { "type": "string", "description": "SEO-optimized social media caption from Gemini" },
    "hashtags":          { "type": "array", "items": { "type": "string" } },
    "drive_link":        { "type": "string", "description": "Google Drive folder URL containing all generated assets" }
  },

  "tools": [
    {
      "tool_name": "gemini_llm",
      "description": "Google Vertex AI Gemini 1.5 Pro for prompt generation and social content creation",
      "api_endpoint": "https://us-central1-aiplatform.googleapis.com/v1/projects/{project_id}/locations/us-central1/publishers/google/models/gemini-1.5-pro:generateContent",
      "auth_method": "bearer_token",
      "auth_header": "Authorization",
      "model": "gemini-1.5-pro",
      "functions": [
        {
          "name": "generate_image_prompt",
          "description": "Generate optimized Imagen prompts for model and product visuals",
          "max_tokens": 300,
          "input_schema": {
            "image_url": {"type": "string"},
            "product_details": {
              "type": "object",
              "fields": ["name", "material", "description"]
            }
          },
          "output_schema": {
            "model_prompt": { "type": "string", "max_tokens": 120 },
            "product_prompt": { "type": "string", "max_tokens": 120 }
          },
          "response_format": "json"
        },
        {
          "name": "generate_social_content",
          "description": "Generate SEO-optimized Instagram caption and hashtag array",
          "max_tokens": 220,
          "input_schema": {
            "product_details": {
              "type": "object",
              "fields": ["name", "price_now", "material", "description"]
            }
          },
          "output_schema": {
            "caption":  { "type": "string", "max_chars": 220 },
            "hashtags": { "type": "array", "max_items": 20 }
          },
          "response_format": "json"
        }
      ]
    },
    {
      "tool_name": "imagen_api",
      "description": "Google Vertex AI Imagen 3 for high-fidelity jewellery image generation",
      "api_endpoint": "https://us-central1-aiplatform.googleapis.com/v1/projects/{project_id}/locations/us-central1/publishers/google/models/imagen-3.0-generate-002:predict",
      "auth_method": "bearer_token",
      "auth_header": "Authorization",
      "model_version": "imagen-3.0-generate-002",
      "functions": [
        {
          "name": "generate_model_image",
          "description": "Generate lifestyle shot of model wearing jewellery",
          "input_schema": {
            "prompt":       { "type": "string", "required": true },
            "aspect_ratio": { "type": "string", "default": "4:5", "enum": ["1:1", "4:5", "9:16"] },
            "person_generation": { "type": "string", "default": "ALLOW_ADULT" }
          },
          "output_schema": {
            "image_bytes": { "type": "string", "format": "base64" }
          }
        },
        {
          "name": "generate_product_image",
          "description": "Generate premium product-only image on luxury background",
          "input_schema": {
            "prompt":       { "type": "string", "required": true },
            "aspect_ratio": { "type": "string", "default": "1:1", "enum": ["1:1", "4:5"] }
          },
          "output_schema": {
            "image_bytes": { "type": "string", "format": "base64" }
          }
        }
      ]
    },
    {
      "tool_name": "drive_api",
      "description": "Google Drive for organized asset storage",
      "api_endpoint": "https://www.googleapis.com/upload/drive/v3/files",
      "auth_method": "oauth2_service_account",
      "scopes": ["https://www.googleapis.com/auth/drive.file"],
      "folder_structure": "/Pattern_Jewellery/{product_category}/{sku}/",
      "functions": [
        {
          "name": "upload_to_drive",
          "description": "Upload file to structured Drive folder, return shareable link",
          "input_schema": {
            "file_data":   { "type": "string", "format": "binary_or_base64" },
            "folder_path": { "type": "string", "example": "/Pattern_Jewellery/rings/R4389/" },
            "file_name":   { "type": "string", "example": "R4389_model_lifestyle.jpg" },
            "mime_type":   { "type": "string", "default": "image/jpeg" }
          },
          "output_schema": {
            "drive_link": { "type": "string", "format": "https://drive.google.com/drive/folders/{folder_id}" },
            "file_id":    { "type": "string" }
          }
        }
      ]
    }
  ],

  "steps": [
    {
      "step": 1,
      "id": "prompt_generation",
      "agent": "creative_agent",
      "tool": "gemini_llm",
      "function": "generate_image_prompt",
      "description": "Analyze the product image and details using Gemini 1.5 Pro to generate tailored model and product generation prompts."
    },
    {
      "step": 2,
      "id": "image_generation",
      "agent": "vision_agent",
      "tool": "imagen_api",
      "parallel": true,
      "description": "Dispatch parallel requests to Imagen 3 to generate the lifestyle model shot and the standalone product shot based on Gemini's prompts."
    },
    {
      "step": 3,
      "id": "content_generation",
      "agent": "content_agent",
      "tool": "gemini_llm",
      "function": "generate_social_content",
      "parallel": true,
      "description": "Use Gemini 1.5 Pro to draft an engaging Instagram caption and SEO hashtag package concurrently with image generation."
    },
    {
      "step": 4,
      "id": "storage",
      "agent": "storage_agent",
      "tool": "drive_api",
      "depends_on": ["image_generation", "content_generation"],
      "description": "Take outputs from the parallel generation steps and upload all assets (images + text) into a dedicated Google Drive folder."
    },
    {
      "step": 5,
      "id": "response_compilation",
      "agent": "system_agent",
      "depends_on": ["storage"],
      "description": "Return the final unified JSON payload containing the Google Drive link and captions."
    }
  ]
}