Update app.py

app.py

@@ -8,6 +8,7 @@ import matplotlib.pyplot as plt
 import re
 from huggingface_hub import hf_hub_download
 import tempfile
+import numpy as np
 
 # --- 1. CONFIGURATION & SECRETS ---
 GOOGLE_API_KEY = os.environ.get("GOOGLE_API_KEY")
@@ -37,75 +38,72 @@ def load_gardiner_database():
 gardiner_data = load_gardiner_database()
 gardiner_map = {k: v.get("Description", k) for k, v in gardiner_data.items()}
 
-# --- 3. HTML GENERATORS ---
+# --- 3. CORE LOGIC FUNCTIONS (Reusable) ---
 
-CATEGORIES = {
-    'A': "Men & Monarchs", 'B': "Women & Human Activities", 'C': "Deities",
-    'D': "Parts of Human Body", 'E': "Mammals", 'F': "Parts of Mammals",
-    'G': "Birds", 'H': "Parts of Birds", 'I': "Reptiles & Amphibians",
-    'K': "Fishes", 'L': "Invertebrates", 'M': "Trees & Plants",
-    'N': "Sky, Earth, Water", 'O': "Buildings", 'P': "Ships",
-    'Q': "Furniture", 'R': "Temple Furniture", 'S': "Crowns & Dress",
-    'T': "Warfare & Hunting", 'U': "Agriculture & Crafts", 'V': "Rope & Baskets",
-    'W': "Vessels", 'X': "Loaves & Cakes", 'Y': "Writings & Games",
-    'Z': "Strokes & Figures", 'Aa': "Unclassified"
-}
-
-def generate_gardiner_html():
-    if not gardiner_data:
-        return "<tr><td colspan='4'>No data loaded. Please upload gardiner_codes.json.</td></tr>"
-
-    html_rows = ""
-    grouped = {}
-    for key, data in gardiner_data.items():
-        match = re.match(r"([A-Za-z]+)", data.get("Code", key))
-        prefix = match.group(1) if match else "Unk"
-        if prefix not in grouped: grouped[prefix] = []
-        grouped[prefix].append(data)
+def core_detect(image, conf_threshold):
+    """Core YOLO detection logic."""
+    if image is None or model is None:
+        return None, [], []
     
-    sorted_prefixes = sorted(grouped.keys(), key=lambda x: (len(x), x))
+    results = model.predict(source=image, conf=conf_threshold, iou=0.45, imgsz=640, verbose=False, device='cpu', max_det=300)
+    annotated_array = results[0].plot() 
+    annotated_image = Image.fromarray(annotated_array[..., ::-1])
     
-    for prefix in sorted_prefixes:
-        cat_name = CATEGORIES.get(prefix, f"Category {prefix}")
-        html_rows += f"<tr><td colspan='4' class='category-header'>{cat_name}</td></tr>"
-        items = sorted(grouped[prefix], key=lambda x: int(re.search(r'\d+', x.get("Code", "0")).group()) if re.search(r'\d+', x.get("Code", "0")) else 0)
+    detections = []
+    crops = []
+    
+    for box in results[0].boxes:
+        if box.cls.numel() > 0:
+            cls_id = int(box.cls[0])
+            if 0 <= cls_id < len(model.names):
+                code = model.names[cls_id]
+                conf = float(box.conf[0])
+                xyxy = box.xyxy[0].tolist()
+                
+                # Create detection object
+                detection = {
+                    "code": code,
+                    "description": gardiner_map.get(code, "Unknown"),
+                    "confidence": round(conf, 2),
+                    "box": xyxy
+                }
+                detections.append(detection)
+                
+                # Create crop
+                crop_img = image.crop((xyxy[0], xyxy[1], xyxy[2], xyxy[3]))
+                crops.append((crop_img, f"{code}\n({int(conf*100)}%)"))
+                
+    return annotated_image, detections, crops
+
+def core_translate(keywords_list):
+    """Core Gemini translation logic."""
+    if not GOOGLE_API_KEY:
+        return "Error: API Key Missing", "Error: API Key Missing"
+    if not keywords_list:
+        return "No symbols detected", "No symbols detected"
+
+    keywords_str = ", ".join(keywords_list)
+    prompt = f"""
+    You are an expert Egyptologist AI. I have detected these symbols: [{keywords_str}].
+    Please provide 2 distinct outputs separated by "|||SEPARATOR|||".
+    1. A Mystical Story: Highly atmospheric, sounding like an ancient prophecy. 
+       Do NOT number this section. Use HTML tags <b> for bolding keywords and <br> for new lines.
+    2. An Academic Translation: Direct, linguistic, focusing on grammar. Use standard text.
+    """
+    try:
+        client = genai.Client(api_key=GOOGLE_API_KEY)
+        response = client.models.generate_content(model="gemini-2.5-flash", contents=prompt)
+        parts = response.text.split("|||SEPARATOR|||")
         
-        for item in items:
-            html_rows += f"""
-            <tr>
-                <td style="font-weight:bold; color: #fff;">{item.get("Code", "?")}</td>
-                <td>{item.get("Description", "-")}</td>
-                <td style="font-family:serif; font-size:1.1em;">{item.get("Transliteration", "-")}</td>
-                <td><span class="type-badge">{item.get("Type", "-")}</span></td>
-            </tr>
-            """
-    return html_rows
-
-GARDINER_TABLE_CONTENT = generate_gardiner_html()
-
-# --- 4. LOAD MODEL ---
-
-print("System: Initializing Rosetta Decoder Core...")
-try:
-    model_path = hf_hub_download(
-        repo_id=MODEL_REPO,
-        filename=MODEL_FILENAME,
-        token=HF_TOKEN
-    )
-    model = YOLO(model_path)
-    print("System: Model loaded successfully.")
-except Exception as e:
-    print(f"Error loading model: {e}")
-    model = None
-
-# --- 5. CORE LOGIC (Helpers) ---
-
-def clean_ai_text(text):
-    text = re.sub(r'^\d+[\.\)]\s*', '', text)
-    text = text.replace("**", "")
-    return text
+        def clean(t): return t.replace("**", "").strip()
+        
+        if len(parts) < 2: return clean(response.text), "Could not parse academic style."
+        return clean(parts[0]), clean(parts[1])
+    except Exception as e:
+        return f"Error: {str(e)}", f"Error: {str(e)}"
 
-def generate_analytics_plots(detections, img_width, img_height):
+def core_analytics(detections, img_w, img_h):
+    """Core Matplotlib logic."""
     if not detections: return None
     
     codes = [d['code'] for d in detections]
@@ -116,6 +114,7 @@ def generate_analytics_plots(detections, img_width, img_height):
     fig = plt.figure(figsize=(10, 15))
     fig.patch.set_facecolor('#0f0f23') 
     
+    # 1. Frequency
     ax1 = plt.subplot(3, 1, 1)
     unique_codes = list(set(codes))
     counts = [codes.count(c) for c in unique_codes]
@@ -125,6 +124,7 @@ def generate_analytics_plots(detections, img_width, img_height):
     ax1.set_facecolor('none')
     for spine in ax1.spines.values(): spine.set_color('#d4af37')
 
+    # 2. Confidence
     ax2 = plt.subplot(3, 1, 2)
     ax2.scatter(range(len(confs)), confs, color='#d4af37', alpha=0.7, s=50)
     ax2.set_title('AI Confidence Levels', color='white', fontsize=12, pad=10)
@@ -133,11 +133,12 @@ def generate_analytics_plots(detections, img_width, img_height):
     ax2.set_facecolor('none')
     for spine in ax2.spines.values(): spine.set_color('#d4af37')
 
+    # 3. Heatmap
     ax3 = plt.subplot(3, 1, 3)
-    h = ax3.hist2d(x_centers, y_centers, bins=[20, 20], range=[[0, img_width], [0, img_height]], cmap='inferno')
+    h = ax3.hist2d(x_centers, y_centers, bins=[20, 20], range=[[0, img_w], [0, img_h]], cmap='inferno')
     ax3.set_title('Glyph Spatial Heatmap', color='white', fontsize=12, pad=10)
-    ax3.set_xlim(0, img_width)
-    ax3.set_ylim(img_height, 0)
+    ax3.set_xlim(0, img_w)
+    ax3.set_ylim(img_h, 0)
     ax3.tick_params(colors='white')
     cbar = plt.colorbar(h[3], ax=ax3)
     cbar.ax.yaxis.set_tick_params(color='white')
@@ -146,121 +147,130 @@ def generate_analytics_plots(detections, img_width, img_height):
     plt.tight_layout(pad=4.0)
     return fig
 
-# --- 6. MCP EXPOSED FUNCTIONS ---
-
-@gr.tool
-def detect_hieroglyphs_mcp(image_path: str, conf_threshold: float = 0.25):
-    """Detects hieroglyphs in an image file. Returns JSON data."""
-    if model is None: return {"error": "Model not loaded"}
-    try:
-        image = Image.open(image_path)
-        img_w, img_h = image.size
-        results = model.predict(source=image, conf=conf_threshold, iou=0.45, imgsz=640, verbose=False, device='cpu', max_det=300)
-        detections = []
-        unique_codes = set()
-        for box in results[0].boxes:
-            if box.cls.numel() > 0:
-                cls_id = int(box.cls[0])
-                if 0 <= cls_id < len(model.names):
-                    code = model.names[cls_id]
-                    conf = float(box.conf[0])
-                    unique_codes.add(code)
-                    xyxy = box.xyxy[0].tolist()
-                    meaning = gardiner_map.get(code, "Unknown")
-                    detections.append({"code": code, "meaning": meaning, "confidence": round(conf, 2), "box": xyxy})
-        return {"status": "success", "image_size": [img_w, img_h], "total_detected": len(detections), "unique_codes": list(unique_codes), "detections": detections}
-    except Exception as e: return {"error": str(e)}
-
-@gr.tool
-def translate_hieroglyphs_mcp(keywords: list[str], style: str = "Academic Literal"):
-    """Translates a list of meanings using Gemini."""
-    if not GOOGLE_API_KEY: return "Error: Google API Key not configured."
-    keywords_str = ", ".join(keywords)
-    prompts = {
-        "Mystical Story": f"Create a coherent, mystical, and atmospheric short story using these ancient Egyptian concepts: [{keywords_str}]. Make it sound like a prophecy.",
-        "Academic Literal": f"Provide a direct, grammatical translation of these concepts: [{keywords_str}]. Focus on linguistic structure and sentence formation used in Egyptology.",
-        "Modern Interpretation": f"Translate the essence of these symbols: [{keywords_str}] into a modern, relatable piece of advice or horoscopic reading."
-    }
-    prompt = prompts.get(style, prompts["Academic Literal"])
-    try:
-        client = genai.Client(api_key=GOOGLE_API_KEY)
-        response = client.models.generate_content(model="gemini-2.5-flash", contents=f"You are an expert Egyptologist AI. {prompt}")
-        return response.text
-    except Exception as e: return f"Translation Error: {str(e)}"
-
-@gr.tool
-def get_gardiner_info_mcp(code: str):
-    """Retrieves detailed info about a specific code."""
-    info = gardiner_data.get(code)
-    if info: return info
-    return {"error": f"Code {code} not found in database."}
+# --- 4. LOAD MODEL ---
 
-@gr.tool
-def get_supported_codes_mcp():
-    """Returns list of supported codes."""
-    return list(gardiner_data.keys())
+print("System: Initializing Rosetta Decoder Core...")
+try:
+    model_path = hf_hub_download(
+        repo_id=MODEL_REPO,
+        filename=MODEL_FILENAME,
+        token=HF_TOKEN
+    )
+    model = YOLO(model_path)
+    print("System: Model loaded successfully.")
+except Exception as e:
+    print(f"Error loading model: {e}")
+    model = None
 
-# --- 7. UI PROCESSING PIPELINE ---
+# --- 5. MAIN UI PIPELINE (Orchestrator) ---
 
 def process_pipeline(image, conf_threshold):
+    """
+    Main function used by the Web UI. 
+    Chains detection -> analytics -> translation.
+    """
     if image is None: return None, "", "", None, "", None, []
     if model is None: return None, "Error: Model not loaded.", "", None, "", None, []
 
     try:
-        results = model.predict(source=image, conf=conf_threshold, iou=0.45, imgsz=640, verbose=False, device='cpu', max_det=300)
-        annotated_array = results[0].plot() 
-        annotated_image = Image.fromarray(annotated_array[..., ::-1])
+        # 1. Detect
         img_w, img_h = image.size
+        annotated_img, detections, crops = core_detect(image, conf_threshold)
         
-        detections = []
-        unique_codes = set()
-        crops = []
-        
-        for box in results[0].boxes:
-            if box.cls.numel() > 0:
-                cls_id = int(box.cls[0])
-                if 0 <= cls_id < len(model.names):
-                    code = model.names[cls_id]
-                    conf = float(box.conf[0])
-                    unique_codes.add(code)
-                    xyxy = box.xyxy[0].tolist()
-                    detections.append({"code": code, "confidence": round(conf, 2), "box": xyxy})
-                    crop_img = image.crop((xyxy[0], xyxy[1], xyxy[2], xyxy[3]))
-                    crops.append((crop_img, f"{code}\n({int(conf*100)}%)"))
-
+        # 2. Extract Keywords
+        unique_codes = list(set([d['code'] for d in detections]))
         mapped_words = [gardiner_map.get(code, f"[{code}]") for code in unique_codes]
         
-        # Translation
-        if not GOOGLE_API_KEY:
-            mystical, academic = "API Key Missing", "API Key Missing"
-        else:
-            keywords_str = ", ".join(mapped_words)
-            prompt = f"""
-            You are an expert Egyptologist AI. I have detected these symbols: [{keywords_str}].
-            Please provide 2 distinct outputs separated by "|||SEPARATOR|||".
-            1. A Mystical Story: Highly atmospheric, sounding like an ancient prophecy. 
-               Do NOT number this section. Use HTML tags <b> for bolding keywords and <br> for new lines.
-            2. An Academic Translation: Direct, linguistic, focusing on grammar. Use standard text.
-            """
-            try:
-                client = genai.Client(api_key=GOOGLE_API_KEY)
-                response = client.models.generate_content(model="gemini-2.5-flash", contents=prompt)
-                parts = response.text.split("|||SEPARATOR|||")
-                if len(parts) < 2: mystical, academic = clean_ai_text(response.text), "Could not parse academic style."
-                else: mystical, academic = clean_ai_text(parts[0].strip()), clean_ai_text(parts[1].strip())
-            except Exception as e:
-                mystical, academic = f"Error: {str(e)}", f"Error: {str(e)}"
-
-        analytics_plot = generate_analytics_plots(detections, img_w, img_h)
+        # 3. Translate
+        mystical, academic = core_translate(mapped_words)
+        
+        # 4. Analytics
+        analytics_plot = core_analytics(detections, img_w, img_h)
+        
+        # 5. Reports
         text_report = f"Total Symbols: {len(detections)}\nUnique Codes: {', '.join(unique_codes)}"
-        formatted_mystical = f"""<div class="mystical-container"><h3>✨ THE ANCIENT WHISPER</h3><p>{mystical}</p></div>"""
         json_output = {"count": len(detections), "detections": detections}
+        formatted_mystical = f"""<div class="mystical-container"><h3>✨ THE ANCIENT WHISPER</h3><p>{mystical}</p></div>"""
 
-        return annotated_image, formatted_mystical, academic, analytics_plot, text_report, json_output, crops
+        return annotated_img, formatted_mystical, academic, analytics_plot, text_report, json_output, crops
 
     except Exception as e:
         return None, f"System Failure: {str(e)}", "", None, str(e), None, []
 
+# --- 6. MCP API FUNCTIONS (Granular) ---
+# These wrappers are exposed as API endpoints for Claude/MCP to use independently.
+
+def api_detect_only(image, conf):
+    """API: Returns JSON detection data and annotated image."""
+    ann_img, dets, _ = core_detect(image, conf)
+    return ann_img, {"count": len(dets), "detections": dets}
+
+def api_translate_only(keywords_text):
+    """API: Translates a comma-separated string of keywords."""
+    # Convert string input "sun, life" to list ["sun", "life"]
+    if isinstance(keywords_text, str):
+        keywords = [k.strip() for k in keywords_text.split(',')]
+    else:
+        keywords = keywords_text
+    mystical, academic = core_translate(keywords)
+    return mystical, academic
+
+def api_get_supported_codes():
+    """API: Returns the full Gardiner list."""
+    return gardiner_data
+
+def api_get_analytics(json_data):
+    """API: Generates plots from JSON detection data."""
+    # Mock image size if not provided, mostly for heatmap relative positions
+    dets = json_data.get("detections", [])
+    return core_analytics(dets, 1000, 1000)
+
+# --- 7. HTML GENERATORS ---
+
+CATEGORIES = {
+    'A': "Men & Monarchs", 'B': "Women & Human Activities", 'C': "Deities",
+    'D': "Parts of Human Body", 'E': "Mammals", 'F': "Parts of Mammals",
+    'G': "Birds", 'H': "Parts of Birds", 'I': "Reptiles & Amphibians",
+    'K': "Fishes", 'L': "Invertebrates", 'M': "Trees & Plants",
+    'N': "Sky, Earth, Water", 'O': "Buildings", 'P': "Ships",
+    'Q': "Furniture", 'R': "Temple Furniture", 'S': "Crowns & Dress",
+    'T': "Warfare & Hunting", 'U': "Agriculture & Crafts", 'V': "Rope & Baskets",
+    'W': "Vessels", 'X': "Loaves & Cakes", 'Y': "Writings & Games",
+    'Z': "Strokes & Figures", 'Aa': "Unclassified"
+}
+
+def generate_gardiner_html():
+    if not gardiner_data:
+        return "<tr><td colspan='4'>No data loaded. Please upload gardiner_codes.json.</td></tr>"
+
+    html_rows = ""
+    grouped = {}
+    for key, data in gardiner_data.items():
+        match = re.match(r"([A-Za-z]+)", data.get("Code", key))
+        prefix = match.group(1) if match else "Unk"
+        if prefix not in grouped: grouped[prefix] = []
+        grouped[prefix].append(data)
+    
+    sorted_prefixes = sorted(grouped.keys(), key=lambda x: (len(x), x))
+    
+    for prefix in sorted_prefixes:
+        cat_name = CATEGORIES.get(prefix, f"Category {prefix}")
+        html_rows += f"<tr><td colspan='4' class='category-header'>{cat_name}</td></tr>"
+        items = sorted(grouped[prefix], key=lambda x: int(re.search(r'\d+', x.get("Code", "0")).group()) if re.search(r'\d+', x.get("Code", "0")) else 0)
+        
+        for item in items:
+            html_rows += f"""
+            <tr>
+                <td style="font-weight:bold; color: #fff;">{item.get("Code", "?")}</td>
+                <td>{item.get("Description", "-")}</td>
+                <td style="font-family:serif; font-size:1.1em;">{item.get("Transliteration", "-")}</td>
+                <td><span class="type-badge">{item.get("Type", "-")}</span></td>
+            </tr>
+            """
+    return html_rows
+
+GARDINER_TABLE_CONTENT = generate_gardiner_html()
+
 # --- 8. UI STYLING & ASSETS ---
 
 cursor_url = "url('data:image/svg+xml;base64,PHN2ZyB4bWxucz0iaHR0cDovL3d3dy53My5vcmcvMjAwMC9zdmciIHdpZHRoPSIzMiIgaGVpZ2h0PSIzMiIgdmlld0JveD0iMCAwIDMyIDMyIj4KICA8ZyBmaWxsPSJub25lIiBzdHJva2U9IiNkNGFmMzciIHN0cm9rZS13aWR0aD0iMS41Ij4KICAgIDxwYXRoIGQ9Ik0xNiw4IEM2LDIwIDI2LDIwIDE2LDggWiIgZmlsbD0icmdiYSgyMTIsIDE3NSwgNTUsIDAuMSkiLz4KICAgIDxjaXJjbGUgY3g9IjE2IiBjeT0iMTUiIHI9IjMiIGZpbGw9IiNkNGFmMzciLz4KICAgIDxwYXRoIGQ9Ik0xNiwyMiBMMTYsMjggTDEwLDI4Ii8+CiAgPC9nPgo8L3N2Zz4=')"
@@ -383,12 +393,38 @@ claude_json_content = """{ "mcpServers": { "gradio": { "command": "npx", "args":
 
 trail_script = """<script>document.addEventListener('DOMContentLoaded', () => { document.addEventListener('mousemove', (e) => { if (Math.random() > 0.7) return; const dust = document.createElement('div'); dust.classList.add('gold-dust'); dust.style.left = e.clientX + 'px'; dust.style.top = e.clientY + 'px'; document.body.appendChild(dust); setTimeout(() => dust.remove(), 600); }); });</script>"""
 
-# --- 8. MAIN APP ASSEMBLY ---
+# --- 9. MAIN APP ASSEMBLY ---
 
 with gr.Blocks(title="Rosetta Decoder Ultimate") as demo:
     gr.HTML(f"<style>{custom_css}</style>")
     gr.HTML(trail_script)
     
+    # --- Hidden API Buttons for MCP (Granular Access) ---
+    # These buttons are not visible in the UI but expose the API endpoints for Claude
+    
+    # 1. Detect Only
+    btn_mcp_detect = gr.Button(visible=False)
+    api_detect_out_img = gr.Image(visible=False)
+    api_detect_out_json = gr.JSON(visible=False)
+    btn_mcp_detect.click(fn=api_detect_only, inputs=[gr.Image(visible=False), gr.Number(visible=False)], outputs=[api_detect_out_img, api_detect_out_json], api_name="detect_hieroglyphs")
+
+    # 2. Translate Only
+    btn_mcp_trans = gr.Button(visible=False)
+    api_trans_out_mystical = gr.Textbox(visible=False)
+    api_trans_out_academic = gr.Textbox(visible=False)
+    btn_mcp_trans.click(fn=api_translate_only, inputs=[gr.Textbox(visible=False)], outputs=[api_trans_out_mystical, api_trans_out_academic], api_name="translate_hieroglyphs")
+
+    # 3. Analytics Only
+    btn_mcp_analytics = gr.Button(visible=False)
+    api_analytics_out = gr.Plot(visible=False)
+    btn_mcp_analytics.click(fn=api_get_analytics, inputs=[gr.JSON(visible=False)], outputs=[api_analytics_out], api_name="get_analytics")
+
+    # 4. Get List
+    btn_mcp_list = gr.Button(visible=False)
+    api_list_out = gr.JSON(visible=False)
+    btn_mcp_list.click(fn=api_get_supported_codes, inputs=[], outputs=[api_list_out], api_name="get_supported_codes")
+
+    # --- Visible UI ---
     with gr.Row(elem_classes="header-row"):
         with gr.Column(scale=4): gr.HTML(header_html)
         with gr.Column(scale=1): btn_toggle = gr.Button("🌗 Day / Night", elem_classes="toggle-btn")