Update app.py

app.py

@@ -2,44 +2,32 @@ from flask import Flask, request, jsonify, render_template, send_from_directory,
 import cv2, json,base64,io,os,tempfile,logging, re
 import numpy as np
 from unstructured.partition.pdf import partition_pdf
-from PIL import Image
-# from imutils.perspective import four_point_transform
+from PIL import Image, ImageOps, ImageEnhance
 from dotenv import load_dotenv
-import pytesseract
+# import pytesseract
 from werkzeug.utils import secure_filename
 from langchain_groq import ChatGroq
 from langgraph.prebuilt import create_react_agent
 from pdf2image import convert_from_path, convert_from_bytes
-from concurrent.futures import ThreadPoolExecutor
-from pdf2image.exceptions import PDFInfoNotInstalledError
 from typing import Dict, TypedDict, Optional, Any
 from langgraph.graph import StateGraph, END
 import uuid
 import shutil, time, functools
-from langchain_experimental.open_clip.open_clip import OpenCLIPEmbeddings
-from langchain_core.utils.utils import secret_from_env
-# from matplotlib.offsetbox import OffsetImage, AnnotationBbox
 from io import BytesIO
 from pathlib import Path
 import os
 from utils.block_relation_builder import block_builder, separate_scripts, transform_logic_to_action_flow, analyze_opcode_counts
-from langchain.chat_models import ChatOpenAI
-from langchain_openai import ChatOpenAI
-from pydantic import Field, SecretStr
 from difflib import get_close_matches
 import torch
 from transformers import AutoImageProcessor, AutoModel
-import faiss
-from sentence_transformers import SentenceTransformer
-
-# image tools
-from PIL import Image, ImageOps, ImageEnhance
+from pathlib import Path
+from io import BytesIO
+import torch
+import json
 import cv2
-
 # hashing & image-match
 from imagededup.methods import PHash
 from image_match.goldberg import ImageSignature
-
 # --- Config (tune threads as needed) ---
 # DINOv2 model id
 DINOV2_MODEL = "facebook/dinov2-small"
@@ -63,7 +51,6 @@ dinov2_model.eval()
 phash = PHash()
 gis = ImageSignature()
 
-
 def log_execution_time(func):
     @functools.wraps(func)
     def wrapper(*args, **kwargs):
@@ -98,12 +85,6 @@ llm = ChatGroq(
 
 app = Flask(__name__)
 
-# ============================== #
-#     TESSERACT CONFIGURATION    #
-# ============================== #
-pytesseract.pytesseract.tesseract_cmd = (r'/usr/bin/tesseract')
-
-# poppler_path = r"C:\poppler\Library\bin"
 backdrop_images_path = r"app\blocks\Backdrops"
 sprite_images_path = r"app\blocks\sprites"
 code_blocks_image_path = r"app\blocks\code_blocks"
@@ -119,11 +100,6 @@ SPRITE_DIR       = BLOCKS_DIR / "sprites"
 CODE_BLOCKS_DIR  = BLOCKS_DIR / "code_blocks"
 # === new: outputs rooted under BASE_DIR ===
 OUTPUT_DIR       = BASE_DIR / "outputs"
-INDEX_PATH = os.path.join(BLOCKS_DIR, "faiss_index.bin")
-PATHS_JSON_PATH = os.path.join(BLOCKS_DIR, "image_paths.json")
-# DETECTED_IMAGE_DIR = OUTPUT_DIR / "DETECTED_IMAGE"
-# SCANNED_IMAGE_DIR  = OUTPUT_DIR / "SCANNED_IMAGE"
-# JSON_DIR           = OUTPUT_DIR / "EXTRACTED_JSON"
 
 # Global variables to hold the model and index, loaded only once.
 MODEL = None
@@ -139,9 +115,6 @@ for d in (
     SPRITE_DIR,
     CODE_BLOCKS_DIR,
     OUTPUT_DIR,
-    # DETECTED_IMAGE_DIR,
-    # SCANNED_IMAGE_DIR,
-    # JSON_DIR,
 ):
     d.mkdir(parents=True, exist_ok=True)
       
@@ -195,108 +168,19 @@ RULES:
 4. If you can't find the data, use "Unknown" for name_variable and "No pseudocode found" for pseudocode
 """
 
-# debugger and resolver agent for Scratch 3.0
 # Main agent of the system agent for Scratch 3.0
 agent = create_react_agent(
     model=llm,
     tools=[], # No specific tools are defined here, but could be added later
     prompt=SYSTEM_PROMPT
 )
-# agent_2 = create_react_agent(
-#     model=llm2,
-#     tools=[], # No specific tools are defined here, but could be added later
-#     prompt=SYSTEM_PROMPT
-# )
+
 agent_json_resolver = create_react_agent(
     model=llm,
     tools=[], # No specific tools are defined here, but could be added later
     prompt=SYSTEM_PROMPT_JSON_CORRECTOR 
 )
 
-def load_model_and_index():
-    """
-    Loads the SentenceTransformer model, FAISS index, and image paths into global variables.
-    This function is called once on the first run to avoid reloading heavy assets.
-    """
-    global MODEL, FAISS_INDEX, IMAGE_PATHS
-
-    # This check ensures we only load everything once
-    if MODEL is None:
-        logger.info("Loading CLIP model 'clip-ViT-L-14' for the first time...")
-        MODEL = SentenceTransformer('clip-ViT-L-14')
-        logger.info("✅ CLIP model loaded.")
-
-        logger.info(f"Loading FAISS index from: {INDEX_PATH}")
-        FAISS_INDEX = faiss.read_index(INDEX_PATH)
-        logger.info("✅ FAISS index loaded.")
-
-        logger.info(f"Loading image paths from: {PATHS_JSON_PATH}")
-        with open(PATHS_JSON_PATH, "r") as f:
-            IMAGE_PATHS = json.load(f)
-        logger.info("✅ Image paths loaded.")
-
-import torch
-from transformers import AutoImageProcessor, AutoModel
-import numpy as np
-from PIL import Image
-from pathlib import Path
-from io import BytesIO
-import json
-
-# def init_dinov2(model_name: str = DINOV2_MODEL, device: torch.device = DEVICE):
-#     """
-#     Lazy-initialize DINOv2 processor & model (call once before embedding).
-#     """
-#     global _dinov2_processor, _dinov2_model
-#     if _dinov2_processor is None or _dinov2_model is None:
-#         _dinov2_processor = AutoImageProcessor.from_pretrained(model_name)
-#         _dinov2_model = AutoModel.from_pretrained(model_name)
-#         _dinov2_model.eval().to(device)
- 
- 
-# def embed_bytesio_list(bytesio_list, batch_size: int = 8):
-#     """
-#     Accepts a list of BytesIO objects (each contains an image).
-#     Returns: np.ndarray shape (N, D) of L2-normalized embeddings (dtype float32).
-#     """
-#     if _dinov2_processor is None or _dinov2_model is None:
-#         init_dinov2()
- 
-#     imgs = []
-#     for b in bytesio_list:
-#         with Image.open(b) as original_img:
-#             # Create a new image with a white background in RGB mode
-#             final_img = Image.new("RGB", original_img.size, (255, 255, 255))
-#             # Paste the original image onto the white background, using the alpha channel as a mask if it exists
-#             if original_img.mode == 'RGBA':
-#                 final_img.paste(original_img, mask=original_img.split()[-1])
-#             else:
-#                 final_img.paste(original_img)
-#             imgs.append(final_img.copy())
- 
-#     embs = []
-#     for i in range(0, len(imgs), batch_size):
-#         batch = imgs[i: i + batch_size]
-#         inputs = _dinov2_processor(images=batch, return_tensors="pt")
-#         inputs = {k: v.to(DEVICE) for k, v in inputs.items()}
-#         with torch.no_grad():
-#             out = _dinov2_model(**inputs)
-#             cls = out.last_hidden_state[:, 0, :]  # (B, D)
-#             cls = torch.nn.functional.normalize(cls, p=2, dim=1)
-#             embs.append(cls.cpu().numpy())
- 
-#     if not embs:
-#         return np.zeros((0, _dinov2_model.config.hidden_size), dtype=np.float32)
- 
-#     return np.vstack(embs).astype(np.float32)
- 
-# def l2_normalize_rows(a: np.ndarray, eps: float = 1e-12) -> np.ndarray:
-#     """
-#     Row-wise L2 normalization for numpy arrays.
-#     """
-#     norm = np.linalg.norm(a, axis=1, keepdims=True)
-#     return a / (norm + eps)
-
 # -----------------------
 #  SERIALIZABLE HELPER
 # -----------------------
@@ -457,8 +341,7 @@ def cosine_similarity(a, b):
 from collections import defaultdict
 import math
 
-def choose_top_candidates(embedding_results, phash_results, imgmatch_results, top_k=10,
-                          method_weights=(0.5, 0.3, 0.2), verbose=True):
+def choose_top_candidates(embedding_results, phash_results, imgmatch_results, top_k=10, method_weights=(0.5, 0.3, 0.2), verbose=True):
     """
     embedding_results: list of (path, emb_sim) where emb_sim roughly in [-1,1] (we'll clamp to 0..1)
     phash_results: list of (path, hamming, ph_sim) where ph_sim in [0,1]
@@ -600,8 +483,7 @@ def is_subpath(path: str, base: str) -> bool:
     try:
         p = os.path.normpath(os.path.abspath(path))
         b = os.path.normpath(os.path.abspath(base))
-        if os.name == "nt":
-            p = p.lower(); b = b.lower()
+        if os.name == "nt": p = p.lower(); b = b.lower()
         return os.path.commonpath([p, b]) == b
     except Exception:
         return False
@@ -619,7 +501,6 @@ def _load_block_catalog(block_type: str) -> Dict:
         catalog = json.loads(text)       # will raise JSONDecodeError if malformed
         logger.info(f"Successfully loaded block catalog from {catalog_path}")
         return catalog
-
     except FileNotFoundError:
         logger.error(f"Error: Block catalog file not found at {catalog_path}")
     except json.JSONDecodeError as e:
@@ -634,8 +515,7 @@ def get_block_by_opcode(catalog_data: dict, opcode: str) -> dict | None:
     Returns the block dict or None if not found.
     """
     for block in catalog_data["blocks"]:
-        if block.get("op_code") == opcode:
-            return block
+        if block.get("op_code") == opcode: return block
     return None
 
 # Helper function to find a block in all catalogs by opcode
@@ -646,8 +526,7 @@ def find_block_in_all(opcode: str, all_catalogs: list[dict]) -> dict | None:
     """
     for catalog in all_catalogs:
         blk = get_block_by_opcode(catalog, opcode)
-        if blk is not None:
-            return blk
+        if blk is not None: return blk
     return None
 
 def variable_intialization(project_data):
@@ -655,24 +534,18 @@ def variable_intialization(project_data):
     Updates variable and broadcast definitions in a Scratch project JSON,
     populating the 'variables' and 'broadcasts' sections of the Stage target
     and extracting initial values for variables.
-
-    Args:
-        project_data (dict): The loaded JSON data of the Scratch project.
-
-    Returns:
-        dict: The updated project JSON data.
+    Args: project_data (dict): The loaded JSON data of the Scratch project.
+    Returns: dict: The updated project JSON data.
     """
 
     stage_target = None
     for target in project_data['targets']:
-        if target.get('isStage'):
+        if target.get('isStage'): 
             stage_target = target
             break
-
     if stage_target is None:
         print("Error: Stage target not found in the project data.")
         return project_data
-
     # Ensure 'variables' and 'broadcasts' exist in the Stage target
     if "variables" not in stage_target:
         stage_target["variables"] = {}
@@ -741,14 +614,9 @@ def deduplicate_variables(project_data):
     """
     Removes duplicate variable entries in the 'variables' dictionary of the Stage target,
     prioritizing entries with non-empty values.
-
-    Args:
-        project_data (dict): The loaded JSON data of the Scratch project.
-
-    Returns:
-        dict: The updated project JSON data with deduplicated variables.
+    Args: project_data (dict): The loaded JSON data of the Scratch project.
+    Returns: dict: The updated project JSON data with deduplicated variables.
     """
-
     stage_target = None
     for target in project_data['targets']:
         if target.get('isStage'):
@@ -762,36 +630,17 @@ def deduplicate_variables(project_data):
     if "variables" not in stage_target:
         return project_data # No variables to deduplicate
 
-    # Use a temporary dictionary to store the preferred variable entry by name
-    # Format: {variable_name: [variable_id, variable_name, variable_value]}
     resolved_variables = {}
 
     for var_id, var_info in stage_target["variables"].items():
         var_name = var_info[0]
         var_value = var_info[1]
 
-        if var_name not in resolved_variables:
-            # If the variable name is not yet seen, add it
-            resolved_variables[var_name] = [var_id, var_name, var_value]
-        else:
-            # If the variable name is already seen, decide which one to keep
-            existing_id, existing_name, existing_value = resolved_variables[var_name]
-
-            # Prioritize the entry with a non-empty value
-            if var_value != "" and existing_value == "":
-                resolved_variables[var_name] = [var_id, var_name, var_value]
-            # If both have non-empty values, or both are empty, keep the current one (arbitrary choice, but consistent)
-            # The current logic will effectively keep the last one encountered that has a value,
-            # or the very last one if all are empty.
-            elif var_value != "" and existing_value != "":
-                 # If there are multiple non-empty values for the same variable name
-                 # this keeps the one from the most recent iteration.
-                 # For the given example, this will correctly keep "5".
-                resolved_variables[var_name] = [var_id, var_name, var_value]
-            elif var_value == "" and existing_value == "":
-                # If both are empty, just keep the current one (arbitrary)
-                resolved_variables[var_name] = [var_id, var_name, var_value]
-
+        if var_name not in resolved_variables: resolved_variables[var_name] = [var_id, var_name, var_value]
+        else: existing_id, existing_name, existing_value = resolved_variables[var_name]
+            if var_value != "" and existing_value == "": resolved_variables[var_name] = [var_id, var_name, var_value]
+            elif var_value != "" and existing_value != "": resolved_variables[var_name] = [var_id, var_name, var_value]
+            elif var_value == "" and existing_value == "": resolved_variables[var_name] = [var_id, var_name, var_value]
 
     # Reconstruct the 'variables' dictionary using the resolved entries
     new_variables_dict = {}
@@ -800,9 +649,7 @@ def deduplicate_variables(project_data):
         var_name_to_keep = var_data[1]
         var_value_to_keep = var_data[2]
         new_variables_dict[var_id_to_keep] = [var_name_to_keep, var_value_to_keep]
-
     stage_target["variables"] = new_variables_dict
-
     return project_data
 
 def variable_adder_main(project_data):
@@ -819,81 +666,81 @@ def variable_adder_main(project_data):
         print(f"Error error in the variable initialization opcodes: {e}")
 
 
-# --- Global variable for the block catalog ---
-ALL_SCRATCH_BLOCKS_CATALOG = {}
-BLOCK_CATALOG_PATH = "blocks" # Define the path to your JSON file
-HAT_BLOCKS_PATH = "hat_blocks" # Path to the hat blocks JSON file
-STACK_BLOCKS_PATH = "stack_blocks" # Path to the stack blocks JSON file
-REPORTER_BLOCKS_PATH = "reporter_blocks" # Path to the reporter blocks JSON file
-BOOLEAN_BLOCKS_PATH = "boolean_blocks" # Path to the boolean blocks JSON file
-C_BLOCKS_PATH = "c_blocks" # Path to the C blocks JSON file
-CAP_BLOCKS_PATH = "cap_blocks" # Path to the cap blocks JSON file
-
-# Load the block catalogs from their respective JSON files
-hat_block_data = _load_block_catalog(HAT_BLOCKS_PATH)
-hat_description = hat_block_data["description"]
-#hat_description = hat_block_data.get("description", "No description available")
-# hat_opcodes_functionalities = "\n".join([f"    - Opcode: {block['op_code']}, functionality: {block['functionality']} example: standalone use: {block['example_standalone']}" for block in hat_block_data["blocks"]])
-hat_opcodes_functionalities = "\n".join([
-    # f"    - Opcode: {block.get('op_code', 'N/A')}, functionality: {block.get('functionality', 'N/A')}, example: standalone use {block.get('example_standalone', 'N/A')}"
-    f"    - Opcode: {block.get('op_code', 'N/A')}, example: standalone use {block.get('example_standalone', 'N/A')}"
-    for block in hat_block_data.get("blocks", [])
-]) if isinstance(hat_block_data.get("blocks"), list) else "    No blocks information available."
-#hat_opcodes_functionalities = os.path.join(BLOCKS_DIR, "hat_blocks.txt")
-print("Hat blocks loaded successfully.", hat_description)
-
-boolean_block_data = _load_block_catalog(BOOLEAN_BLOCKS_PATH)
-boolean_description = boolean_block_data["description"]
-# boolean_opcodes_functionalities = "\n".join([f"    - Opcode: {block['op_code']}, functionality: {block['functionality']} example: standalone use: {block['example_standalone']}" for block in boolean_block_data["blocks"]])
-boolean_opcodes_functionalities = "\n".join([
-    # f"    - Opcode: {block.get('op_code', 'N/A')}, functionality: {block.get('functionality', 'N/A')}, example: standalone use {block.get('example_standalone', 'N/A')}"
-    f"    - Opcode: {block.get('op_code', 'N/A')}, example: standalone use {block.get('example_standalone', 'N/A')}"
-    for block in boolean_block_data.get("blocks", [])
-]) if isinstance(boolean_block_data.get("blocks"), list) else "    No blocks information available."
-#boolean_opcodes_functionalities = os.path.join(BLOCKS_DIR, "boolean_blocks.txt")
-
-c_block_data = _load_block_catalog(C_BLOCKS_PATH)
-c_description = c_block_data["description"]
-# c_opcodes_functionalities = "\n".join([f"    - Opcode: {block['op_code']}, functionality: {block['functionality']} example: standalone use: {block['example_standalone']}" for block in c_block_data["blocks"]])
-c_opcodes_functionalities = "\n".join([
-    # f"    - Opcode: {block.get('op_code', 'N/A')}, functionality: {block.get('functionality', 'N/A')}, example: standalone use {block.get('example_standalone', 'N/A')}"
-    f"    - Opcode: {block.get('op_code', 'N/A')}, example: standalone use {block.get('example_standalone', 'N/A')}"
-    for block in c_block_data.get("blocks", [])
-]) if isinstance(c_block_data.get("blocks"), list) else "    No blocks information available."
-#c_opcodes_functionalities = os.path.join(BLOCKS_DIR, "c_blocks.txt")
-
-cap_block_data = _load_block_catalog(CAP_BLOCKS_PATH)
-cap_description = cap_block_data["description"]                                                                                                                                                                                                                                                                                                                                                                                                                                                      
-# cap_opcodes_functionalities = "\n".join([f"    - Opcode: {block['op_code']}, functionality: {block['functionality']} example: standalone use: {block['example_standalone']}" for block in cap_block_data["blocks"]])
-cap_opcodes_functionalities = "\n".join([
-    # f"    - Opcode: {block.get('op_code', 'N/A')}, functionality: {block.get('functionality', 'N/A')}, example: standalone use {block.get('example_standalone', 'N/A')}"
-    f"    - Opcode: {block.get('op_code', 'N/A')}, example: standalone use {block.get('example_standalone', 'N/A')}"
-    for block in cap_block_data.get("blocks", [])
-]) if isinstance(cap_block_data.get("blocks"), list) else "    No blocks information available."
-#cap_opcodes_functionalities = os.path.join(BLOCKS_DIR, "cap_blocks.txt")
-
-reporter_block_data = _load_block_catalog(REPORTER_BLOCKS_PATH)
-reporter_description = reporter_block_data["description"]
-# reporter_opcodes_functionalities = "\n".join([f"    - Opcode: {block['op_code']}, functionality: {block['functionality']} example: standalone use: {block['example_standalone']}" for block in reporter_block_data["blocks"]])
-reporter_opcodes_functionalities = "\n".join([
-    # f"    - Opcode: {block.get('op_code', 'N/A')}, functionality: {block.get('functionality', 'N/A')}, example: standalone use {block.get('example_standalone', 'N/A')}"
-    f"    - Opcode: {block.get('op_code', 'N/A')}, example: standalone use {block.get('example_standalone', 'N/A')}"
-    for block in reporter_block_data.get("blocks", [])
-]) if isinstance(reporter_block_data.get("blocks"), list) else "    No blocks information available."
-#reporter_opcodes_functionalities = os.path.join(BLOCKS_DIR, "reporter_blocks.txt")
-
-stack_block_data = _load_block_catalog(STACK_BLOCKS_PATH)
-stack_description = stack_block_data["description"]
-# stack_opcodes_functionalities = "\n".join([f"    - Opcode: {block['op_code']}, functionality: {block['functionality']} example: standalone use: {block['example_standalone']}" for block in stack_block_data["blocks"]])
-stack_opcodes_functionalities = "\n".join([
-    # f"    - Opcode: {block.get('op_code', 'N/A')}, functionality: {block.get('functionality', 'N/A')}, example: standalone use {block.get('example_standalone', 'N/A')}"
-    f"    - Opcode: {block.get('op_code', 'N/A')}, example: standalone use {block.get('example_standalone', 'N/A')}"
-    for block in stack_block_data.get("blocks", [])
-]) if isinstance(stack_block_data.get("blocks"), list) else "    No blocks information available."
-#stack_opcodes_functionalities = os.path.join(BLOCKS_DIR, "stack_blocks.txt")
-
-# This makes ALL_SCRATCH_BLOCKS_CATALOG available globally
-ALL_SCRATCH_BLOCKS_CATALOG = _load_block_catalog(BLOCK_CATALOG_PATH)
+# # --- Global variable for the block catalog ---
+# ALL_SCRATCH_BLOCKS_CATALOG = {}
+# BLOCK_CATALOG_PATH = "blocks" # Define the path to your JSON file
+# HAT_BLOCKS_PATH = "hat_blocks" # Path to the hat blocks JSON file
+# STACK_BLOCKS_PATH = "stack_blocks" # Path to the stack blocks JSON file
+# REPORTER_BLOCKS_PATH = "reporter_blocks" # Path to the reporter blocks JSON file
+# BOOLEAN_BLOCKS_PATH = "boolean_blocks" # Path to the boolean blocks JSON file
+# C_BLOCKS_PATH = "c_blocks" # Path to the C blocks JSON file
+# CAP_BLOCKS_PATH = "cap_blocks" # Path to the cap blocks JSON file
+
+# # Load the block catalogs from their respective JSON files
+# hat_block_data = _load_block_catalog(HAT_BLOCKS_PATH)
+# hat_description = hat_block_data["description"]
+# #hat_description = hat_block_data.get("description", "No description available")
+# # hat_opcodes_functionalities = "\n".join([f"    - Opcode: {block['op_code']}, functionality: {block['functionality']} example: standalone use: {block['example_standalone']}" for block in hat_block_data["blocks"]])
+# hat_opcodes_functionalities = "\n".join([
+#     # f"    - Opcode: {block.get('op_code', 'N/A')}, functionality: {block.get('functionality', 'N/A')}, example: standalone use {block.get('example_standalone', 'N/A')}"
+#     f"    - Opcode: {block.get('op_code', 'N/A')}, example: standalone use {block.get('example_standalone', 'N/A')}"
+#     for block in hat_block_data.get("blocks", [])
+# ]) if isinstance(hat_block_data.get("blocks"), list) else "    No blocks information available."
+# #hat_opcodes_functionalities = os.path.join(BLOCKS_DIR, "hat_blocks.txt")
+# print("Hat blocks loaded successfully.", hat_description)
+
+# boolean_block_data = _load_block_catalog(BOOLEAN_BLOCKS_PATH)
+# boolean_description = boolean_block_data["description"]
+# # boolean_opcodes_functionalities = "\n".join([f"    - Opcode: {block['op_code']}, functionality: {block['functionality']} example: standalone use: {block['example_standalone']}" for block in boolean_block_data["blocks"]])
+# boolean_opcodes_functionalities = "\n".join([
+#     # f"    - Opcode: {block.get('op_code', 'N/A')}, functionality: {block.get('functionality', 'N/A')}, example: standalone use {block.get('example_standalone', 'N/A')}"
+#     f"    - Opcode: {block.get('op_code', 'N/A')}, example: standalone use {block.get('example_standalone', 'N/A')}"
+#     for block in boolean_block_data.get("blocks", [])
+# ]) if isinstance(boolean_block_data.get("blocks"), list) else "    No blocks information available."
+# #boolean_opcodes_functionalities = os.path.join(BLOCKS_DIR, "boolean_blocks.txt")
+
+# c_block_data = _load_block_catalog(C_BLOCKS_PATH)
+# c_description = c_block_data["description"]
+# # c_opcodes_functionalities = "\n".join([f"    - Opcode: {block['op_code']}, functionality: {block['functionality']} example: standalone use: {block['example_standalone']}" for block in c_block_data["blocks"]])
+# c_opcodes_functionalities = "\n".join([
+#     # f"    - Opcode: {block.get('op_code', 'N/A')}, functionality: {block.get('functionality', 'N/A')}, example: standalone use {block.get('example_standalone', 'N/A')}"
+#     f"    - Opcode: {block.get('op_code', 'N/A')}, example: standalone use {block.get('example_standalone', 'N/A')}"
+#     for block in c_block_data.get("blocks", [])
+# ]) if isinstance(c_block_data.get("blocks"), list) else "    No blocks information available."
+# #c_opcodes_functionalities = os.path.join(BLOCKS_DIR, "c_blocks.txt")
+
+# cap_block_data = _load_block_catalog(CAP_BLOCKS_PATH)
+# cap_description = cap_block_data["description"]                                                                                                                                                                                                                                                                                                                                                                                                                                                      
+# # cap_opcodes_functionalities = "\n".join([f"    - Opcode: {block['op_code']}, functionality: {block['functionality']} example: standalone use: {block['example_standalone']}" for block in cap_block_data["blocks"]])
+# cap_opcodes_functionalities = "\n".join([
+#     # f"    - Opcode: {block.get('op_code', 'N/A')}, functionality: {block.get('functionality', 'N/A')}, example: standalone use {block.get('example_standalone', 'N/A')}"
+#     f"    - Opcode: {block.get('op_code', 'N/A')}, example: standalone use {block.get('example_standalone', 'N/A')}"
+#     for block in cap_block_data.get("blocks", [])
+# ]) if isinstance(cap_block_data.get("blocks"), list) else "    No blocks information available."
+# #cap_opcodes_functionalities = os.path.join(BLOCKS_DIR, "cap_blocks.txt")
+
+# reporter_block_data = _load_block_catalog(REPORTER_BLOCKS_PATH)
+# reporter_description = reporter_block_data["description"]
+# # reporter_opcodes_functionalities = "\n".join([f"    - Opcode: {block['op_code']}, functionality: {block['functionality']} example: standalone use: {block['example_standalone']}" for block in reporter_block_data["blocks"]])
+# reporter_opcodes_functionalities = "\n".join([
+#     # f"    - Opcode: {block.get('op_code', 'N/A')}, functionality: {block.get('functionality', 'N/A')}, example: standalone use {block.get('example_standalone', 'N/A')}"
+#     f"    - Opcode: {block.get('op_code', 'N/A')}, example: standalone use {block.get('example_standalone', 'N/A')}"
+#     for block in reporter_block_data.get("blocks", [])
+# ]) if isinstance(reporter_block_data.get("blocks"), list) else "    No blocks information available."
+# #reporter_opcodes_functionalities = os.path.join(BLOCKS_DIR, "reporter_blocks.txt")
+
+# stack_block_data = _load_block_catalog(STACK_BLOCKS_PATH)
+# stack_description = stack_block_data["description"]
+# # stack_opcodes_functionalities = "\n".join([f"    - Opcode: {block['op_code']}, functionality: {block['functionality']} example: standalone use: {block['example_standalone']}" for block in stack_block_data["blocks"]])
+# stack_opcodes_functionalities = "\n".join([
+#     # f"    - Opcode: {block.get('op_code', 'N/A')}, functionality: {block.get('functionality', 'N/A')}, example: standalone use {block.get('example_standalone', 'N/A')}"
+#     f"    - Opcode: {block.get('op_code', 'N/A')}, example: standalone use {block.get('example_standalone', 'N/A')}"
+#     for block in stack_block_data.get("blocks", [])
+# ]) if isinstance(stack_block_data.get("blocks"), list) else "    No blocks information available."
+# #stack_opcodes_functionalities = os.path.join(BLOCKS_DIR, "stack_blocks.txt")
+
+# # This makes ALL_SCRATCH_BLOCKS_CATALOG available globally
+# ALL_SCRATCH_BLOCKS_CATALOG = _load_block_catalog(BLOCK_CATALOG_PATH)
         
 def extract_json_from_llm_response(raw_response: str) -> dict:
     """
@@ -1288,19 +1135,6 @@ end
         print(f"result:\n\n {result}")
         
     except json.JSONDecodeError as error_json:
-        # If JSON parsing fails, use the json resolver agent
-        # correction_prompt = (
-        #     "Your task is to correct the provided JSON string to ensure it is **syntactically perfect and adheres strictly to JSON rules**.\n"
-        #     "It must be a JSON object with `refined_logic` (string) and `block_relationships` (array of objects).\n"
-        #     f"- **Error Details**: {error_json}\n\n"
-        #     "**Strict Instructions for your response:**\n"
-        #     "1. **ONLY** output the corrected JSON. Do not include any other text or explanations.\n"
-        #     "2. Ensure all keys and string values are enclosed in **double quotes**. Escape internal quotes (`\\`).\n"
-        #     "3. No trailing commas. Correct nesting.\n\n"
-        #     "Here is the problematic JSON string to correct:\n"
-        #     f"```json\n{llm_output_raw}\n```\n"
-        #     "Corrected JSON:\n"
-        # )
         correction_prompt = f"""
 Fix this malformed response and return only the corrected JSON:
 
@@ -1317,7 +1151,6 @@ Extract the sprite name and pseudocode, then return in this exact format:
         try:
             correction_response = agent_json_resolver.invoke({"messages": [{"role": "user", "content": correction_prompt}]})
             corrected_output = extract_json_from_llm_response(correction_response['messages'][-1].content)
-            #block_relationships = corrected_output.get("block_relationships", [])
             result = corrected_output 
             print(f"result:\n\n {result}")                   
         except Exception as e_corr:
@@ -1328,8 +1161,6 @@ Extract the sprite name and pseudocode, then return in this exact format:
     state["pseudo_code"] = result
     state["temp_pseudo_code"] += [result]
     Data = state["temp_pseudo_code"]
-    # with open("debug_state.json", "w", encoding="utf-8") as f:
-    #     json.dump(state, f, indent=2, ensure_ascii=False)
     print(f"[OVREALL REFINED PSEUDO CODE LOGIC]: {result}")
     print(f"[OVREALL LISTS OF LOGICS]: {Data}")
     logger.info("Plan refinement and block relation analysis completed for all plans.")
@@ -1464,14 +1295,7 @@ def overall_block_builder_node_2(state: GameState):
                     logger.info(f"Action blocks added for sprite '{sprite_name}' by OverallBlockBuilderNode.")
                 except Exception as e:
                     logger.error(f"Error generating blocks for sprite '{sprite_name}': {e}")
-                    # Consider adding more specific error handling here if a malformed output
-                    # from block_builder should cause a specific state change, but generally
-                    # avoid nulling the entire project_json.
-                    
-    state["project_json"] = project_json
-    # with open("debug_state.json", "w", encoding="utf-8") as f:
-    #     json.dump(state, f, indent=2, ensure_ascii=False)
-    
+    state["project_json"] = project_json   
     return state
 
 # Node 6: variable adder node
@@ -1553,15 +1377,11 @@ def extract_images_from_pdf(pdf_stream: io.BytesIO):
 
         try:
             elements = partition_pdf(
-                # filename=str(pdf_path), # partition_pdf might expect a string
-                file=pdf_stream, # 'file=', inplace of 'filename'
+                file=pdf_stream,
                 strategy="hi_res",
                 extract_image_block_types=["Image"],
                 hi_res_model_name="yolox",
                 extract_image_block_to_payload=True,
-                # ocr_languages=ocr_lang,
-                # extract_images_in_pdf=False,  
-                # extract_image_block_output_dir=r"E:\Pratham\2025\Harsh Sir\Scratch Vision\images\pdf_output"
             )
             print(f"ELEMENTS")
         except Exception as e:
@@ -1578,260 +1398,19 @@ def extract_images_from_pdf(pdf_stream: io.BytesIO):
                 continue
            
             manipulated_json[f"Sprite {sprite_count}"] = {
-                # "id":auto_id,
-                # "name": name,
                 "base64": el["metadata"]["image_base64"],
                 "file-path": pdf_id,
-                # "description": description
             }
             sprite_count += 1
         return manipulated_json
     except Exception as e:
         raise RuntimeError(f"❌ Error in extract_images_from_pdf: {str(e)}")
-
-# def similarity_matching(sprites_data: str, project_folder: str) -> str:
-#     logger.info("🔍 Running similarity matching…")
-#     os.makedirs(project_folder, exist_ok=True)
-
-#     # ----------------------------------------
-#     # CHANGED: define normalized base-paths so startswith() checks work
-#     backdrop_base_path = os.path.normpath(str(BACKDROP_DIR))
-#     sprite_base_path   = os.path.normpath(str(SPRITE_DIR))
-#     code_blocks_path = os.path.normpath(str(CODE_BLOCKS_DIR))
-#     # ----------------------------------------
-
-#     project_json_path = os.path.join(project_folder, "project.json")
-
-#     # ==============================
-#     #    READ SPRITE METADATA
-#     # ==============================
-#     # with open(input_json_path, 'r') as f:
-#     #     sprites_data = json.load(f)
-
-#     sprite_ids, sprite_base64 = [], []
-#     for sid, sprite in sprites_data.items():
-#         sprite_ids.append(sid)
-#         # texts.append("This is " + sprite.get("description", sprite.get("name", "")))
-#         sprite_base64.append(sprite["base64"])
-
-#     sprite_images_bytes = []
-#     for b64 in sprite_base64:
-#         img = Image.open(BytesIO(base64.b64decode(b64.split(",")[-1]))).convert("RGB")
-#         buffer = BytesIO()
-#         img.save(buffer, format="PNG")
-#         buffer.seek(0)
-#         sprite_images_bytes.append(buffer)
-    
-#     # =========================================
-#     #  Build the list of all candidate images
-#     # =========================================
-#     folder_image_paths = [
-# BACKDROP_DIR/"Baseball 2.sb3"/"7be1f5b3e682813dac1f297e52ff7dca.png",
-# BACKDROP_DIR/"Beach Malibu.sb3"/"050615fe992a00d6af0e664e497ebf53.png",
-# BACKDROP_DIR/"Bedroom 3.sb3"/"8cc0b88d53345b3e337e8f028a32a4e7.png",
-# BACKDROP_DIR/"Blue Sky.sb3"/"e7c147730f19d284bcd7b3f00af19bb6.png",
-# BACKDROP_DIR/"Castle 2.sb3"/"951765ee7f7370f120c9df20b577c22f.png",
-# BACKDROP_DIR/"Colorful City.sb3"/"04d18ddd1b85f0ea30beb14b8da49f60.png",
-# BACKDROP_DIR/"Hall.sb3"/"ea86ca30b346f27ca5faf1254f6a31e3.png",
-# BACKDROP_DIR/"Jungle.sb3"/"f4f908da19e2753f3ed679d7b37650ca.png",
-# BACKDROP_DIR/"Soccer.sb3"/"04a63154f04b09494354090f7cc2f1b9.png",
-# BACKDROP_DIR/"Theater.sb3"/"c2b097bc5cdb6a14ef5485202bc5ee76.png",
-
-# SPRITE_DIR/"Batter.sprite3"/"592ee9ab2aeefe65cb4fb95fcd046f33.png",
-# SPRITE_DIR/"Batter.sprite3"/"9d193bef6e3d6d8eba6d1470b8bf9351.png",
-# SPRITE_DIR/"Batter.sprite3"/"baseball_sprite_motion_1.png",
-# SPRITE_DIR/"Batter.sprite3"/"bd4fc003528acfa847e45ff82f346eee.png",
-# SPRITE_DIR/"Batter.sprite3"/"fdfde4bcbaca0f68e83fdf3f4ef0c660.png",
-# SPRITE_DIR/"Bear.sprite3"/"6f303e972f33fcb7ef36d0d8012d0975.png",
-# SPRITE_DIR/"Bear.sprite3"/"bear_motion_2.png",
-# SPRITE_DIR/"Bear.sprite3"/"deef1eaa96d550ae6fc11524a1935024.png",
-# SPRITE_DIR/"Beetle.sprite3"/"46d0dfd4ae7e9bfe3a6a2e35a4905eae.png",
-# SPRITE_DIR/"Butterfly 1.sprite3"/"34b76c1835c6a7fc2c47956e49bb0f52.png",
-# SPRITE_DIR/"Butterfly 1.sprite3"/"49c9f952007d870a046cff93b6e5e098.png",
-# SPRITE_DIR/"Butterfly 1.sprite3"/"fe98df7367e314d9640bfaa54fc239be.png",
-# SPRITE_DIR/"Cat.sprite3"/"0fb9be3e8397c983338cb71dc84d0b25.png",
-# SPRITE_DIR/"Cat.sprite3"/"bcf454acf82e4504149f7ffe07081dbc.png",
-# SPRITE_DIR/"Centaur.sprite3"/"2373556e776cad3ba4d6ee04fc34550b.png",
-# SPRITE_DIR/"Centaur.sprite3"/"c00ffa6c5dd0baf9f456b897ff974377.png",
-# SPRITE_DIR/"Centaur.sprite3"/"d722329bd9373ad80625e5be6d52f3ed.png",
-# SPRITE_DIR/"Centaur.sprite3"/"d7aa990538915b7ef1f496d7e8486ade.png",
-# SPRITE_DIR/"City Bus.sprite3"/"7d7e26014a346b894db8ab1819f2167f.png",
-# SPRITE_DIR/"City Bus.sprite3"/"e9694adbff9422363e2ea03166015393.png",
-# SPRITE_DIR/"Crab.sprite3"/"49839aa1b0feed02a3c759db5f8dee71.png",
-# SPRITE_DIR/"Crab.sprite3"/"bear_element.png",
-# SPRITE_DIR/"Crab.sprite3"/"f7cdd2acbc6d7559d33be8675059c79e.png",
-# SPRITE_DIR/"Glow-G.sprite3"/"56839bc48957869d980c6f9b6f5a2a91.png",
-# SPRITE_DIR/"Jordyn.sprite3"/"00c8c464c19460df693f8d5ae69afdab.png",
-# SPRITE_DIR/"Jordyn.sprite3"/"768c4601174f0dfcb96b3080ccc3a192.png",
-# SPRITE_DIR/"Jordyn.sprite3"/"a7cc1e5f02b58ecc8095cfc18eef0289.png",
-# SPRITE_DIR/"Jordyn.sprite3"/"db4d97cbf24e2b8af665bfbf06f67fa0.png",
-# SPRITE_DIR/"Soccer Ball.sprite3"/"5d973d7a3a8be3f3bd6e1cd0f73c32b5.png",
-# SPRITE_DIR/"Soccer Ball.sprite3"/"cat_football.png",
-# SPRITE_DIR/"Star.sprite3"/"551629f2a64c1f3703e57aaa133effa6.png",
-# SPRITE_DIR/"Wizard.sprite3"/"55ba51188af86ca16ef30267e874c1ed.png",
-# SPRITE_DIR/"Wizard.sprite3"/"91d495085eb4d02a375c42f6318071e7.png",
-# SPRITE_DIR/"Wizard.sprite3"/"df943c9894ee4b9df8c5893ce30c2a5f.png",
-
-# # CODE_BLOCKS_DIR/"client_code_block_1.jpg",
-# # CODE_BLOCKS_DIR/"client_code_block_2.jpg",
-# CODE_BLOCKS_DIR/"script1.JPG",
-# CODE_BLOCKS_DIR/"script2.JPG",
-# CODE_BLOCKS_DIR/"script3.JPG",
-# CODE_BLOCKS_DIR/"script4.JPG",
-# CODE_BLOCKS_DIR/"script5.JPG",
-# CODE_BLOCKS_DIR/"script6.JPG",
-# CODE_BLOCKS_DIR/"script7.JPG",
-# CODE_BLOCKS_DIR/"script8.JPG",
-# CODE_BLOCKS_DIR/"script9.JPG",
-# CODE_BLOCKS_DIR/"static_white.png"]
-#     folder_image_paths = [os.path.normpath(str(p)) for p in folder_image_paths]
-#     # =========================================
-
-#     # -----------------------------------------
-#     #  Load reference embeddings from JSON
-#     # -----------------------------------------
-#     with open(f"{BLOCKS_DIR}/dinov2_embeddings.json", "r") as f:
-#         embedding_json = json.load(f)
-
-#     # ============================== #
-#     #      EMBED SPRITE IMAGES       #
-#     # ============================== #
-#     # ensure model is initialized (fast no-op after first call)
-#     init_dinov2()
-
-#     # embed the incoming sprite BytesIO images (same data structure you already use)
-#     sprite_matrix = embed_bytesio_list(sprite_images_bytes, batch_size=8)  # shape (N, D)
-
-#     # load reference embeddings from JSON (they must be numeric lists)
-#     img_matrix = np.array([img["embeddings"] for img in embedding_json], dtype=np.float32)
-
-#     # normalize both sides (important — stored embeddings may not be normalized)
-#     sprite_matrix = l2_normalize_rows(sprite_matrix)
-#     img_matrix = l2_normalize_rows(img_matrix)
-    
-#     # =========================================
-#     #  Compute similarities & pick best match
-#     # =========================================
-#     similarity = np.matmul(sprite_matrix, img_matrix.T)
-#     most_similar_indices = np.argmax(similarity, axis=1)
-
-#     # =========================================
-#     #  Copy matched sprite assets + collect data
-#     # =========================================
-#     project_data   = []
-#     copied_folders = set()
-
-#     for sprite_idx, matched_idx in enumerate(most_similar_indices):
-#         matched_image_path = folder_image_paths[matched_idx]
-#         matched_folder     = os.path.dirname(matched_image_path)
-
-#         # CHANGED: use our new normalized sprite_base_path
-#         if not matched_folder.startswith(sprite_base_path):
-#             continue
-
-#         if matched_folder in copied_folders:
-#             continue
-#         copied_folders.add(matched_folder)
-#         logger.info(f"Matched sprite: {matched_image_path}")
-
-#         sprite_json_path = os.path.join(matched_folder, 'sprite.json')
-#         if not os.path.exists(sprite_json_path):
-#             logger.warning(f"No sprite.json in {matched_folder}")
-#             continue
-
-#         with open(sprite_json_path, 'r') as f:
-#             sprite_info = json.load(f)
-#         # copy all non‐matched files
-#         for fname in os.listdir(matched_folder):
-#             if fname in (os.path.basename(matched_image_path), 'sprite.json'):
-#                 continue
-#             shutil.copy2(os.path.join(matched_folder, fname),
-#                          os.path.join(project_folder, fname))
-#         project_data.append(sprite_info)
-
-#     # =========================================
-#     #  Copy matched backdrop assets + collect
-#     # =========================================
-#     backdrop_data = []
-#     copied_backdrop_folders = set()
-#     for backdrop_idx, matched_idx in enumerate(most_similar_indices):
-#         matched_image_path = folder_image_paths[matched_idx]
-#         matched_folder     = os.path.dirname(matched_image_path)
-#         matched_filename = os.path.basename(matched_image_path)
-        
-#         # CHANGED: use our new normalized backdrop_base_path
-#         if not matched_folder.startswith(backdrop_base_path):
-#             continue
-
-#         # skip if backdrop folder already processed
-#         if matched_folder in copied_backdrop_folders:
-#             continue
-#         copied_backdrop_folders.add(matched_folder)
-
-#         logger.info(f"Matched backdrop: {matched_image_path}")
-
-#         # 1) Copy the matched backdrop image itself
-#         try:
-#             shutil.copy2(
-#                 matched_image_path,
-#                 os.path.join(project_folder, matched_filename)
-#             )
-#             logger.info(f"✅ Copied matched backdrop image {matched_filename} to {project_folder}")
-#         except Exception as e:
-#             logger.error(f"❌ Failed to copy matched backdrop {matched_image_path}: {e}")
         
-#         # copy non‐matched files
-#         for fname in os.listdir(matched_folder):
-#             # if fname in (os.path.basename(matched_image_path), 'project.json'):
-#             if fname in {matched_filename, 'project.json'}:
-#                 continue
-#             # shutil.copy2(os.path.join(matched_folder, fname),
-#             #              os.path.join(project_folder, fname))
-#             src = os.path.join(matched_folder, fname)
-#             dst = os.path.join(project_folder, fname)
-#             if os.path.isfile(src):
-#                 try:
-#                     shutil.copy2(src, dst)
-#                     logger.info(f"Copied additional backdrop asset {fname} to project folder")
-#                 except Exception as e:
-#                     logger.error(f"Failed to copy {src}: {e}")
-
-#         # append the stage‐target from its project.json
-#         pj = os.path.join(matched_folder, 'project.json')
-#         if os.path.exists(pj):
-#             with open(pj, 'r') as f:
-#                 bd_json = json.load(f)
-#             for tgt in bd_json.get("targets", []):
-#                 if tgt.get("isStage"):
-#                     backdrop_data.append(tgt)
-#         else:
-#             logger.warning(f"No project.json in {matched_folder}")
-
-
-#     # =========================================
-#     #  Merge into final Scratch project.json
-#     # =========================================
-#     final_project = {
-#         "targets": [], "monitors": [], "extensions": [],
-#         "meta": {
-#             "semver": "3.0.0",
-#             "vm": "11.3.0",
-#             "agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/138.0.0.0 Safari/537.36"
-#         }
-#     }
-#     # sprites first
-#     for spr in project_data:
-#         if not spr.get("isStage", False):
-#             final_project["targets"].append(spr)
 ''' It appends all the list and paths from json files and pick the best match's path'''
-
 def similarity_matching(sprites_data: dict, project_folder: str, top_k: int = 1, min_similarity: float = None) -> str:
     print("🔍 Running similarity matching…")
     os.makedirs(project_folder, exist_ok=True)
 
-    # backdrop_base_path = r"D:\DEV PATEL\2025\scratch_VLM\scratch_agent\blocks\Backdrops"
-    # sprite_base_path   = r"D:\DEV PATEL\2025\scratch_VLM\scratch_agent\blocks\sprites"
-    # code_blocks_path = r"D:\DEV PATEL\2025\scratch_VLM\scratch_agent\blocks\code_blocks"
     backdrop_base_path = os.path.normpath(str(BACKDROP_DIR))
     sprite_base_path   = os.path.normpath(str(SPRITE_DIR))
     code_blocks_path = os.path.normpath(str(CODE_BLOCKS_DIR))