from flask import Flask, render_template, request, jsonify import json,base64,io,os,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 dotenv import load_dotenv import pytesseract # from transformers import AutoProcessor, AutoModelForImageTextToText, AutoModelForVision2Seq from langchain_community.document_loaders.image_captions import ImageCaptionLoader from werkzeug.utils import secure_filename from langchain_groq import ChatGroq from langgraph.prebuilt import create_react_agent from pdf2image import convert_from_bytes #convert_from_path, import asyncio 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 matplotlib.offsetbox import OffsetImage, AnnotationBbox from io import BytesIO # https://prthm11-Scratch_vlm_v1.hf.space/ # https://huggingface.co/spaces/prthm11/Scratch_vlm_v1 # https://prthm11-scratch-vlm-v1.hf.space/process_pdf # def log_execution_time(func): # @functools.wraps(func) # def wrapper(*args, **kwargs): # start_time = time.time() # result = func(*args, **kwargs) # end_time = time.time() # logger.info(f"⏱ {func.__name__} executed in {end_time - start_time:.2f} seconds") # return result # return wrapper # ============================== # # INITIALIZE CLIP EMBEDDER # # ============================== # clip_embd = OpenCLIPEmbeddings() # Configure logging logging.basicConfig( level=logging.DEBUG, # Use INFO or ERROR in production format="%(asctime)s [%(levelname)s] %(message)s", handlers=[ logging.FileHandler("app.log"), logging.StreamHandler() ] ) logger = logging.getLogger(__name__) load_dotenv() # os.environ["GROQ_API_KEY"] = os.getenv("GROQ_API_KEY") groq_api_key = os.getenv("GROQ_API_KEY") llm = ChatGroq( # model="meta-llama/llama-4-scout-17b-16e-instruct", model = "meta-llama/llama-4-maverick-17b-128e-instruct", temperature=0, max_tokens=None, ) app = Flask(__name__) pytesseract.pytesseract.tesseract_cmd = r"C:\Program Files\Tesseract-OCR\tesseract.exe" poppler_path = r"C:\poppler-23.11.0\Library\bin" count = 0 OUTPUT_FOLDER = "OUTPUTS" DETECTED_IMAGE_FOLDER_PATH = os.path.join(OUTPUT_FOLDER, "DETECTED_IMAGE") IMAGE_FOLDER_PATH = os.path.join(OUTPUT_FOLDER, "SCANNED_IMAGE") JSON_FOLDER_PATH = os.path.join(OUTPUT_FOLDER, "EXTRACTED_JSON") for path in [OUTPUT_FOLDER, IMAGE_FOLDER_PATH, DETECTED_IMAGE_FOLDER_PATH, JSON_FOLDER_PATH]: os.makedirs(path, exist_ok=True) # class GameState(TypedDict): # image: str # pseudo_node: Optional[Dict] # # Refined SYSTEM_PROMPT with more explicit Scratch JSON rules, especially for variables # SYSTEM_PROMPT = """ # You are an expert AI assistant named GameScratchAgent, specialized in generating and modifying Scratch-VM 3.x game project JSON. # Your core task is to process game descriptions and existing Scratch JSON structures, then produce or update JSON segments accurately. # You possess deep knowledge of Scratch 3.0 project schema, informed by comprehensive reference materials. When generating or modifying the `blocks` section, pay extremely close attention to the following: # **Scratch Project JSON Schema Rules:** # 1. **Target Structure (`project.json`'s `targets` array):** # * Each object in the `targets` array represents a Stage or a Sprite. # * `isStage`: A boolean indicating if the target is the Stage (`true`) or a Sprite (`false`). # * `name`: The name of the Stage (e.g., `"Stage"`) or the Sprite (e.g., `"Cat"`). This property replaces `objName` found in older Scratch versions. # * `variables` dictionary: This dictionary maps unique variable IDs to arrays `[variable_name, initial_value, isCloudVariable?]`. # * `variable_name`: The user-defined name of the variable. # * `initial_value`: The variable's initial value, which can be a number or a string. # * `isCloudVariable?`: (Optional) A boolean indicating if it's a cloud variable (`true`) or a local variable (`false` or absent for regular variables). # * Example: `"myVarId123": ["score", 0]`, `"cloudVarId456": ["☁ High Score", "54", true]` # * `lists` dictionary: This dictionary maps unique list IDs to arrays `[list_name, [item1, item2, ...]]`. # * Example: `"myListId789": ["my list", ["apple", "banana"]]` # * `broadcasts` dictionary: This dictionary maps unique broadcast IDs to their names. # * Example: `"myBroadcastId": "Game Over"` # * `blocks` dictionary: This dictionary contains all the blocks belonging to this target. Keys are block IDs, values are block objects. # 2. **Block Structure (within a `target`'s `blocks` dictionary):** # * Every block object must have the following core properties: # * [cite_start]`opcode`: A unique internal identifier for the block's specific functionality (e.g., `"motion_movesteps"`, `"event_whenflagclicked"`)[cite: 31, 18, 439, 452]. # * `parent`: The ID of the block directly above it in the script stack (or `null` for a top-level block). # * `next`: The ID of the block directly below it in the script stack (or `null` for the end of a stack). # * `inputs`: An object defining values or blocks plugged into the block's input slots. Values are **arrays**. # * `fields`: An object defining dropdown menu selections or direct internal values within the block. Values are **arrays**. # * `shadow`: `true` if it's a shadow block (e.g., a default number input that can be replaced by another block), `false` otherwise. # * `topLevel`: `true` if it's a hat block or a standalone block (not connected to a parent), `false` otherwise. # 3. **`inputs` Property Details (for blocks plugged into input slots):** # * **Direct Block Connection (Reporter/Boolean block plugged in):** # * Format: `"": [1, ""]` # * Example: `"CONDITION": [1, "someBooleanBlockId"]` (e.g., for an `if` block). # * **Literal Value Input (Shadow block with a literal):** # * Format: `"": [1, [, ""]]` # * `type_code`: A numeric code representing the data type. Common codes include: `4` for number, `7` for string/text, `10` for string/message. # * `value_string`: The literal value as a string. # * Examples: # * Number: `"STEPS": [1, [4, "10"]]` (for `move 10 steps` block). # * String/Text: `"MESSAGE": [1, [7, "Hello"]]` (for `say Hello` block). # * String/Message (common for text inputs): `"MESSAGE": [1, [10, "Hello!"]]` (for `say Hello! for 2 secs`). # * **C-Block Substack (blocks within a loop or conditional):** # * Format: `"": [2, ""]` # * Common `SUBSTACK_NAME` values are `SUBSTACK` (for `if`, `forever`, `repeat`) and `SUBSTACK2` (for `else` in `if else`). # * Example: `"SUBSTACK": [2, "firstBlockInLoopId"]` # 4. **`fields` Property Details (for dropdowns or direct internal values):** # * Used for dropdown menus, variable names, list names, or other static selections directly within the block. # * Format: `"": ["", null]` # * Examples: # * Dropdown: `"KEY_OPTION": ["space", null]` (for `when space key pressed`). # * Variable Name: `"VARIABLE": ["score", null]` (for `set score to 0`). # * Direction (specific motion block): `"FORWARD_BACKWARD": ["forward", null]` (for `go forward layers`). # 5. **Unique IDs:** # * All block IDs, variable IDs, and list IDs must be unique strings (e.g., "myBlock123", "myVarId456", "myListId789"). Do NOT use placeholder strings like "block_id_here". # 6. **No Nested `blocks` Dictionary:** # * The `blocks` dictionary should only appear once per `target` (sprite/stage). Do NOT nest a `blocks` dictionary inside an individual block definition. Blocks that are part of a substack are linked via the `SUBSTACK` input. # 7. **Asset Properties (for Costumes/Sounds):** # * `assetId`, `md5ext`, `bitmapResolution`, `rotationCenterX`/`rotationCenterY` should be correctly associated with costume and sound objects within the `costumes` and `sounds` arrays. # **General Principles and Important Considerations:** # * **Backward Compatibility:** Adhere strictly to existing Scratch 3.0 opcodes and schema to ensure backward compatibility with older projects. [cite_start]Opcodes must remain consistent to prevent previously saved projects from failing to load or behaving unexpectedly[cite: 18, 19, 25, 65]. # * **Forgiving Inputs:** Recognize that Scratch is designed to be "forgiving in its interpretation of inputs." [cite_start]The Scratch VM handles potentially "invalid" inputs gracefully (e.g., converting a number to a string if expected, returning default values like zero or empty strings, or performing no action) rather than crashing[cite: 20, 21, 22, 38, 39, 41]. This implies that precise type matching for inputs might be handled internally by Scratch, allowing for some flexibility in how values are provided, but the agent should aim for the most common and logical type. # """ # SYSTEM_PROMPT_JSON_CORRECTOR =""" # You are an assistant that outputs JSON responses strictly following the given schema. # If the JSON you produce has any formatting errors, missing required fields, or invalid structure, you must identify the problems and correct them. # Always return only valid JSON that fully conforms to the schema below, enclosed in triple backticks (```), without any extra text or explanation. # If you receive an invalid or incomplete JSON response, fix it by: # - Adding any missing required fields with appropriate values. # - Correcting syntax errors such as missing commas, brackets, or quotes. # - Ensuring the JSON structure matches the schema exactly. # Remember: Your output must be valid JSON only, ready to be parsed without errors. # """ # # debugger and resolver agent for Scratch 3.0 # 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 # ) # # Helper function to load the block catalog from a JSON file # def _load_block_catalog(file_path: str) -> Dict: # """Loads the Scratch block catalog from a specified JSON file.""" # try: # with open(file_path, 'r') as f: # catalog = json.load(f) # logger.info(f"Successfully loaded block catalog from {file_path}") # return catalog # except FileNotFoundError: # logger.error(f"Error: Block catalog file not found at {file_path}") # # Return an empty dict or raise an error, depending on desired behavior # return {} # except json.JSONDecodeError as e: # logger.error(f"Error decoding JSON from {file_path}: {e}") # return {} # except Exception as e: # logger.error(f"An unexpected error occurred while loading {file_path}: {e}") # return {} # # --- Global variable for the block catalog --- # ALL_SCRATCH_BLOCKS_CATALOG = {} # BLOCK_CATALOG_PATH = r"blocks\blocks.json" # Define the path to your JSON file # HAT_BLOCKS_PATH = r"blocks\hat_blocks.json" # Path to the hat blocks JSON file # STACK_BLOCKS_PATH = r"blocks\stack_blocks.json" # Path to the stack blocks JSON file # REPORTER_BLOCKS_PATH = r"blocks\reporter_blocks.json" # Path to the reporter blocks JSON file # BOOLEAN_BLOCKS_PATH = r"blocks\boolean_blocks.json" # Path to the boolean blocks JSON file # C_BLOCKS_PATH = r"blocks\c_blocks.json" # Path to the C blocks JSON file # CAP_BLOCKS_PATH = r"blocks\cap_blocks.json" # 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_opcodes_functionalities = os.path.join(HAT_BLOCKS_PATH, "hat_blocks.txt") # boolean_block_data = _load_block_catalog(BOOLEAN_BLOCKS_PATH) # boolean_description = boolean_block_data["description"] # boolean_opcodes_functionalities = os.path.join(BOOLEAN_BLOCKS_PATH, "boolean_blocks.txt") # c_block_data = _load_block_catalog(C_BLOCKS_PATH) # c_description = c_block_data["description"] # c_opcodes_functionalities = os.path.join(C_BLOCKS_PATH, "c_blocks.txt") # cap_block_data = _load_block_catalog(CAP_BLOCKS_PATH) # cap_description = cap_block_data["description"] # cap_opcodes_functionalities = os.path.join(CAP_BLOCKS_PATH, "cap_blocks.txt") # reporter_block_data = _load_block_catalog(REPORTER_BLOCKS_PATH) # reporter_description = reporter_block_data["description"] # reporter_opcodes_functionalities = os.path.join(REPORTER_BLOCKS_PATH, "reporter_blocks.txt") # stack_block_data = _load_block_catalog(STACK_BLOCKS_PATH) # stack_description = stack_block_data["description"] # stack_opcodes_functionalities = os.path.join(STACK_BLOCKS_PATH, "stack_blocks.txt") # # Helper function to extract JSON from LLM response # def extract_json_from_llm_response(raw_response: str) -> dict: # # --- 1) Pull out the JSON code‑block if present --- # md = re.search(r"```(?:json)?\s*([\s\S]*?)\s*```", raw_response) # json_string = md.group(1).strip() if md else raw_response # # --- 2) Trim to the outermost { … } so we drop any prefix/suffix junk --- # first, last = json_string.find('{'), json_string.rfind('}') # if 0 <= first < last: # json_string = json_string[first:last+1] # # --- 3) PRE‑CLEANUP: remove stray assistant{…}, rogue assistant keys, fix boolean quotes --- # json_string = re.sub(r'\b\w+\s*{', '{', json_string) # json_string = re.sub(r'"assistant"\s*:', '', json_string) # json_string = re.sub(r'\b(false|true)"', r'\1', json_string) # logger.debug("Ran pre‑cleanup for stray tokens and boolean quotes.") # # --- 3.1) Fix stray inner quotes at start of name/list values --- # # e.g., { "name": " \"recent_scoress\"", ... } → "recent_scoress" # json_string = re.sub( # r'("name"\s*:\s*")\s*"', # r'\1', # json_string # ) # # --- 4) Escape all embedded quotes in any `logic` value up to the next key --- # def _esc(m): # prefix, body = m.group(1), m.group(2) # return prefix + body.replace('"', r'\"') # json_string = re.sub( # r'("logic"\s*:\s*")([\s\S]+?)(?=",\s*"[A-Za-z_]\w*"\s*:\s*)', # _esc, # json_string # ) # logger.debug("Escaped embedded quotes in logic fields.") # logger.debug("Quoted unquoted keys.") # # --- 6) Remove trailing commas before } or ] --- # json_string = re.sub(r',\s*(?=[}\],])', '', json_string) # json_string = re.sub(r',\s*,', ',', json_string) # logger.debug("Removed trailing commas.") # # --- 7) Balance braces: drop extra } at end if needed --- # ob, cb = json_string.count('{'), json_string.count('}') # if cb > ob: # excess = cb - ob # json_string = json_string.rstrip()[:-excess] # logger.debug(f"Stripped {excess} extra closing brace(s).") # # --- 8) Escape literal newlines in *all* string values --- # json_string = re.sub( # r'"((?:[^"\\]|\\.)*?)"', # lambda m: '"' + m.group(1).replace('\n', '\\n').replace('\r', '\\r') + '"', # json_string, # flags=re.DOTALL # ) # logger.debug("Escaped newlines in strings.") # # --- 9) Final parse attempt --- # try: # return json.loads(json_string) # except json.JSONDecodeError: # logger.error("Sanitized JSON still invalid:\n%s", json_string) # raise # # 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 # ) # # Node 6: Logic updating if any issue here # def plan_logic_aligner_node(state: GameState): # logger.info("--- Running plan_logic_aligner_node ---") # image = state.get("image", "") # refinement_prompt = f""" # You are an expert in Scratch 3.0 game development, specializing in understanding block relationships (stacked, nested). # "Analyze the Scratch code-block image and generate Pseudo-Code for what this logic appears to be doing." # From Image, you also have to detect a value of Key given in Text form "Script for: ". Below is the example # Example: "Script for: Bear", "Script for:" is a key and "Bear" is value. # --- Scratch 3.0 Block Reference --- # ### Hat Blocks # Description: {hat_description} # Blocks: # {hat_opcodes_functionalities} # ### Boolean Blocks # Description: {boolean_description} # Blocks: # {boolean_opcodes_functionalities} # ### C Blocks # Description: {c_description} # Blocks: # {c_opcodes_functionalities} # ### Cap Blocks # Description: {cap_description} # Blocks: # {cap_opcodes_functionalities} # ### Reporter Blocks # Description: {reporter_description} # Blocks: # {reporter_opcodes_functionalities} # ### Stack Blocks # Description: {stack_description} # Blocks: # {stack_opcodes_functionalities} # ----------------------------------- # Your task is to: # If you don't find any "Code-Blocks" then, # **Don't generate Pseudo Code, and pass the message "No Code-blocks" find... # If you find any "Code-Blocks" then, # 1. **Refine the 'logic'**: Make it precise, accurate, and fully aligned with the Game Description. Use Scratch‑consistent verbs and phrasing. **Do NOT** use raw double‑quotes inside the logic string. # 2. **Structural requirements**: # - **Numeric values** `(e.g., 0, 5, 0.2, -130)` **must** be in parentheses: `(0)`, `(5)`, `(0.2)`, `(-130)`. # - **AlphaNumeric values** `(e.g., hello, say 5, 4, hi!)` **must** be in parentheses: `(hello)`, `(say 5)`, `(4)`, `(hi!)`. # - **Variables** must be in the form `[variable v]` (e.g., `[score v]`), even when used inside expressions two example use `set [score v] to (1)` or `show variable ([speed v])`. # - **Dropdown options** must be in the form `[option v]` (e.g., `[Game Start v]`, `[blue sky v]`). example use `when [space v] key pressed`. # - **Reporter blocks** used as inputs must be double‑wrapped: `((x position))`, `((y position))`. example use `if <((y position)) = (-130)> then` or `(((x position)) * (1))`. # - **Boolean blocks** in conditions must be inside `< >`, including nested ones: `>`, `< and >`,`< or >`. # - **Other Boolean blocks** in conditions must be inside `< >`, including nested ones or values or variables: `<(block/value/variable) * (block/value/variable)>`,`<(block/value/variable) < (block/value/variable)>`, and example of another variable`<[apple v] contains [a v]?>`. # - **Operator expressions** must use explicit Scratch operator blocks, e.g.: # ``` # (([ballSpeed v]) * (1.1)) # ``` # - **Every hat block script must end** with a final `end` on its own line. # 3. **Pseudo‑code formatting**: # - Represent each block or nested block on its own line. # - Indent nested blocks by 4 spaces under their parent (`forever`, `if`, etc.). # - No comments or explanatory text—just the block sequence. # - a natural language breakdown of each step taken after the event, formatted as a multi-line string representing pseudo-code. Ensure clarity and granularity—each described action should map closely to a Scratch block or tight sequence. # 4. **Logic content**: # - Build clear flow for mechanics (movement, jumping, flying, scoring, collisions). # - Match each action closely to a Scratch block or tight sequence. # - Do **NOT** include any justification or comments—only the raw logic. # 5. **Examples for reference**: # **Correct** pattern for a simple start script: # ``` # when green flag clicked # switch backdrop to [blue sky v] # set [score v] to (0) # show variable [score v] # broadcast [Game Start v] # end # ``` # **Correct** pattern for updating the high score variable handling: # ``` # when I receive [Game Over v] # if <((score)) > (([High Score v]))> then # set [High Score v] to ([score v]) # end # switch backdrop to [Game Over v] # end # ``` # **Correct** pattern for level up and increase difficulty use: # ``` # when I receive [Level Up v] # change [level v] by (1) # set [ballSpeed v] to ((([ballSpeed v]) * (1.1))) # end # ``` # **Correct** pattern for jumping mechanics use: # ``` # when [space v] key pressed # if <((y position)) = (-100)> then # repeat (5) # change y by (100) # wait (0.1) seconds # change y by (-100) # wait (0.1) seconds # end # end # end # ``` # **Correct** pattern for continuos moving objects use: # ``` # when green flag clicked # go to x: (240) y: (-100) # set [speed v] to (-5) # show variable [speed v] # forever # change x by ([speed v]) # if <((x position)) < (-240)> then # go to x: (240) y: (-100) # end # end # end # ``` # **Correct** pattern for continuos moving objects use: # ``` # when green flag clicked # go to x: (240) y: (-100) # set [speed v] to (-5) # show variable [speed v] # forever # change x by ([speed v]) # if <((x position)) < (-240)> then # go to x: (240) y: (-100) # end # end # end # ``` # 6. **Donot** add any explaination of logic or comments to justify or explain just put the logic content in the json. # 7. **Output**: # Return **only** a JSON object, using double quotes everywhere: # ```json # {{ # "refined_logic":{{ # "name_variable": 'Value of "Sript for: "', # "pseudocode":"…your fully‑formatted pseudo‑code here…", # }} # }} # ``` # """ # image_input = { # "type": "image_url", # "image_url": { # "url": f"data:image/png;base64,{image}" # } # } # content = [ # {"type": "text", "text": refinement_prompt}, # image_input # ] # try: # # Invoke the main agent for logic refinement and relationship identification # response = agent.invoke({"messages": [{"role": "user", "content": content}]}) # llm_output_raw = response["messages"][-1].content.strip() # parsed_llm_output = extract_json_from_llm_response(llm_output_raw) # # result = parsed_llm_output # # Extract needed values directly # logic_data = parsed_llm_output.get("refined_logic", {}) # name_variable = logic_data.get("name_variable", "Unknown") # pseudocode = logic_data.get("pseudocode", "No logic extracted") # result = {"pseudo_node": { # "name_variable": name_variable, # "pseudocode": pseudocode # }} # print(f"result:\n\n {result}") # return result # except Exception as e: # logger.error(f"❌ plan_logic_aligner_node failed: {str(e)}") # return {"error": str(e)} # 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" # ) # 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 = { # #"image_path": image_path, # "pseudo_code": corrected_output # } # return result # except Exception as e_corr: # logger.error(f"Failed to correct JSON output for even after retry: {e_corr}") # scratch_keywords = [ # "move", "turn", "wait", "repeat", "if", "else", "broadcast", # "glide", "change", "forever", "when", "switch", # "next costume", "set", "show", "hide", "play sound", # "go to", "x position", "y position", "think", "say", # "variable", "stop", "clone", # "touching", "sensing", "pen", "clear","Scratch","Code","scratch blocks" # ] # filtered_sprites = {} # Prepare manipulated sprite JSON structure manipulated_json = {} img_elements = [] # @log_execution_time # --- FUNCTION: Extract images from saved PDF --- def extract_images_from_pdf(pdf_stream: io.BytesIO): ''' Extract images from PDF and generate structured sprite JSON ''' try: # pdf_filename = os.path.splitext(os.path.basename(pdf_stream))[0] # e.g., "scratch_crab" if isinstance(pdf_stream, io.BytesIO): # use a random ID since there's no filename pdf_id = uuid.uuid4().hex else: pdf_id = os.path.splitext(os.path.basename(pdf_stream))[0] # pdf_dir_path = os.path.dirname(pdf_stream).replace("/", "\\") # Create subfolders # extracted_image_subdir = os.path.join(DETECTED_IMAGE_FOLDER_PATH, pdf_filename) # json_subdir = os.path.join(JSON_FOLDER_PATH, pdf_filename) # os.makedirs(extracted_image_subdir, exist_ok=True) # os.makedirs(json_subdir, exist_ok=True) # pdf_bytes = pdf_stream.getvalue() try: elements = partition_pdf( file=pdf_stream, strategy="hi_res", extract_image_block_types=["Image"], hi_res_model_name="yolox", extract_image_block_to_payload=True, # Set to True to get base64 in output ) except Exception as e: raise RuntimeError( f"❌ Failed to extract images from PDF: {str(e)}") file_elements = [element.to_dict() for element in elements] sprite_count = 1 for el in file_elements: img_b64 = el["metadata"].get("image_base64") # with open(os.path.join(DETECTED_IMAGE_FOLDER_PATH, f"img_{sprite_count}.png")) as d_img: # d_img = img_b64 if not img_b64: continue # raw = base64.b64decode(img_b64) # im = Image.open(io.BytesIO(raw)).convert("RGB") # up = upscale_image(im, scale=2) # buf = io.BytesIO() # up.save(buf, format="PNG") # buf.seek(0) # img_elements.append(base64.b64encode(buf.getvalue()).decode()) # print(f"------------------------IMAGE ELEMENTS: \n{img_elements}") # auto_id = f"sprite_{sprite_count}" 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 # print(f"************MANIPULATED JSON: {manipulated_json}") # manipulated_json_path = os.path.join(JSON_FOLDER_PATH, "manipulated.json") # with open(manipulated_json_path, 'w') as f: # json.dump(manipulated_json, f, indent=2) # def is_code_block(name: str) -> bool: # for kw in scratch_keywords: # if kw.lower() in name.lower(): # return True # return False # # Filter out code block images # for key, value in manipulated_json.items(): # sprite_name = value.get("name", "") # if not is_code_block(sprite_name): # filtered_sprites[key] = value # else: # logger.info(f"🛑 Excluded code block-like image: {key}") return manipulated_json except Exception as e: raise RuntimeError(f"❌ Error in extract_images_from_pdf: {str(e)}") # @log_execution_time def similarity_matching(sprites_data: str, project_folder:str) -> str: logger.info("🔍 Running similarity matching...") # ============================== # # DEFINE PATHS # # ============================== # # backdrop_images_path = r"E:\Pratham\2025\Harsh Sir\Scratch Vision\images\Backdrops" # sprite_images_path = r"E:\Pratham\2025\Harsh Sir\Scratch Vision\images\sprites" # image_dirs = [backdrop_images_path, sprite_images_path] 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, texts, 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"]) # print(f"\nSPRITE_BASE64: \n{sprite_base64}\n\n") 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) # ========================================= # # Walk folders to collect all image paths # # ========================================= # folder_image_paths = ['E:\\Pratham\\2025\\Harsh Sir\\Scratch Vision\\images\\Backdrops\\bedroom\\8cc0b88d53345b3e337e8f028a32a4e7.png', 'E:\\Pratham\\2025\\Harsh Sir\\Scratch Vision\\images\\Backdrops\\baseball\\7be1f5b3e682813dac1f297e52ff7dca.png', 'E:\\Pratham\\2025\\Harsh Sir\\Scratch Vision\\images\\Backdrops\\beach_malibu\\050615fe992a00d6af0e664e497ebf53.png', 'E:\\Pratham\\2025\\Harsh Sir\\Scratch Vision\\images\\Backdrops\\castle\\951765ee7f7370f120c9df20b577c22f.png', 'E:\\Pratham\\2025\\Harsh Sir\\Scratch Vision\\images\\Backdrops\\hall\\ea86ca30b346f27ca5faf1254f6a31e3.png', 'E:\\Pratham\\2025\\Harsh Sir\\Scratch Vision\\images\\Backdrops\\jungle\\f4f908da19e2753f3ed679d7b37650ca.png', 'E:\\Pratham\\2025\\Harsh Sir\\Scratch Vision\\images\\sprites\\Batter\\baseball_sprite_motion_1.png', 'E:\\Pratham\\2025\\Harsh Sir\\Scratch Vision\\images\\sprites\\Bear\\bear_motion_2.png', 'E:\\Pratham\\2025\\Harsh Sir\\Scratch Vision\\images\\sprites\\Beetle\\46d0dfd4ae7e9bfe3a6a2e35a4905eae.png', 'E:\\Pratham\\2025\\Harsh Sir\\Scratch Vision\\images\\sprites\\cat\\cat_motion_1.png', 'E:\\Pratham\\2025\\Harsh Sir\\Scratch Vision\\images\\sprites\\Centaur\\2373556e776cad3ba4d6ee04fc34550b.png', 'E:\\Pratham\\2025\\Harsh Sir\\Scratch Vision\\images\\sprites\\Crab\\bear_element.png', 'E:\\Pratham\\2025\\Harsh Sir\\Scratch Vision\\images\\sprites\\Soccer Ball\\cat_football.png', 'E:\\Pratham\\2025\\Harsh Sir\\Scratch Vision\\images\\code_blocks\\script1.jpg', 'E:\\Pratham\\2025\\Harsh Sir\\Scratch Vision\\images\\code_blocks\\script2.jpg', 'E:\\Pratham\\2025\\Harsh Sir\\Scratch Vision\\images\\code_blocks\\script3.jpg', 'E:\\Pratham\\2025\\Harsh Sir\\Scratch Vision\\images\\code_blocks\\script4.jpg', 'E:\\Pratham\\2025\\Harsh Sir\\Scratch Vision\\images\\code_blocks\\script5.jpg', 'E:\\Pratham\\2025\\Harsh Sir\\Scratch Vision\\images\\code_blocks\\script6.jpg' ] # ============================== # # EMBED SPRITE IMAGES # # ============================== # sprite_features = clip_embd.embed_image(sprite_images_bytes) # # ============================== # # # EMBED FOLDER IMAGES (REF) # # # ============================== # # img_features = clip_embd.embed_image(folder_image_paths) # # ============================== # # # Store image embeddings # # # ============================== # # embedding_json = [] # for i, path in enumerate(folder_image_paths): # embedding_json.append({ # "name":os.path.basename(path), # "file-path": path, # "embeddings": list(img_features[i]) # }) # # # Save to embeddings.json # with open(f"{OUTPUT_FOLDER}/embeddings.json", "w") as f: # json.dump(embedding_json, f, indent=2) # ============================== # # COMPUTE SIMILARITIES # # ============================== # ref_embedding_path = f"{OUTPUT_FOLDER}/embeddings.json" with open(ref_embedding_path, "r") as f: embedding_json = json.load(f) # print(f"\n\n EMBEDDING JSON: {embedding_json}") img_matrix = np.array([img["embeddings"] for img in embedding_json]) sprite_matrix = np.array(sprite_features) # if sprite_matrix.size == 0 or img_matrix.size == 0: # raise RuntimeError("❌ No valid embeddings found for sprites or reference images.") similarity = np.matmul(sprite_matrix, img_matrix.T) # try: # similarity = np.matmul(sprite_matrix, img_matrix.T) # except ValueError as ve: # if "matmul" in str(ve) and "size" in str(ve): # logger.error("❌ Matrix multiplication failed due to shape mismatch. Likely due to empty or invalid embeddings.") # raise RuntimeError("Matrix shape mismatch: CLIP embedding input is invalid or empty.") # else: # raise most_similar_indices = np.argmax(similarity, axis=1) sprite_base_path = r'E:\Pratham\2025\Harsh Sir\Scratch Vision\images\sprites' # ============= Match and copy =============== project_data = [] #will hold loaded "sprite.json" contents for building the final "project.json" copied_folders = set() # prevents copying the same sprite folder more than once. # =============================================================== # # Loop through most similar images from Sprites folder # # → Copy sprite assets (excluding matched image + sprite.json) # # → Load sprite.json and append its data to project_data # # =============================================================== # #sprite_idx: index of sprite we are processing #match_idx: index of the reference image with the highest similarity for sprite_idx, matched_idx in enumerate(most_similar_indices): matched_image_path = folder_image_paths[matched_idx] print(f"------------ folder image paths: \n {folder_image_paths[matched_idx]}\n") matched_image_path = os.path.normpath(matched_image_path) matched_folder = os.path.dirname(matched_image_path) print(f"\nMATCHED_FOLDER: {matched_folder}\n") if not matched_folder.startswith(os.path.normpath(sprite_base_path)): continue folder_name = os.path.basename(matched_folder) print(f"FOLDER NAME: {folder_name}") print(f"================COPIED FOLDER: \n {copied_folders}\n") if matched_folder in copied_folders: continue copied_folders.add(matched_folder) logger.info(f"Matched sprites: {matched_image_path}") sprite_json_path = os.path.join(matched_folder, 'sprite.json') if not os.path.exists(sprite_json_path): logger.warning(f"sprite.json not found in: {matched_folder}") continue with open(sprite_json_path, 'r') as f: sprite_data = json.load(f) # print(f"SPRITE DATA: \n{sprite_data}") # Copy only non-matched files for fname in os.listdir(matched_folder): fpath = os.path.join(matched_folder, fname) if os.path.isfile(fpath) and fname not in {os.path.basename(matched_image_path), 'sprite.json'}: shutil.copy2(fpath, os.path.join(project_folder, fname)) # logger.info(f"Copied Sprite asset: {fname}") project_data.append(sprite_data) # ================================================================== # # Loop through most similar images from Backdrops folder # # → Copy matched image + other backdrop assets into project_folder # # → Load project.json and append its stage target to backdrop_data# # ================================================================== # backdrop_data = [] # for backdrop-related entries copied_backdrop_folders = set() # prevent duplicate backdrops # make sure backdrop_base_path is normalized backdrop_base_path = os.path.normpath(r'E:\Pratham\2025\Harsh Sir\Scratch Vision\images\Backdrops') for backdrop_idx, matched_idx in enumerate(most_similar_indices): matched_image_path = os.path.normpath(folder_image_paths[matched_idx]) # only handle backdrops if not matched_image_path.startswith(backdrop_base_path): continue matched_folder = os.path.dirname(matched_image_path) # skip if backdrop folder already processed if matched_folder in copied_backdrop_folders: continue copied_backdrop_folders.add(matched_folder) matched_filename = os.path.basename(matched_image_path) logger.info(f"Backdrop matched image: {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}") # 2) Copy other non‐matched files (e.g. extra costumes) into project_folder for fname in os.listdir(matched_folder): if fname in {matched_filename, 'project.json'}: continue 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}") # 3) Load and append the Stage target from this backdrop's project.json backdrop_json_path = os.path.join(matched_folder, 'project.json') if os.path.exists(backdrop_json_path): with open(backdrop_json_path, 'r') as f: backdrop_json_data = json.load(f) for target in backdrop_json_data.get("targets", []): if target.get("isStage"): backdrop_data.append(target) else: logger.warning(f"project.json not found in: {matched_folder}") # Merge JSON structure 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 Edg/138.0.0.0" } } for sprite in project_data: if not sprite.get("isStage", False): final_project["targets"].append(sprite) if backdrop_data: all_costumes, sounds = [], [] for idx, bd in enumerate(backdrop_data): all_costumes.extend(bd.get("costumes", [])) if idx == 0 and "sounds" in bd: sounds = bd["sounds"] final_project["targets"].append({ "isStage": True, "name": "Stage", "objName": "Stage", "variables": { "`jEk@4|i[#Fk?(8x)AV.-my variable": ["my variable", 0] }, "lists": {}, "broadcasts": {}, "blocks": {}, "comments": {}, "currentCostume": 1 if len(all_costumes) > 1 else 0, "costumes": all_costumes, "sounds": sounds, "volume": 100, "layerOrder": 0, "tempo": 60, "videoTransparency": 50, "videoState": "on", "textToSpeechLanguage": None }) else: logger.warning("⚠️ No backdrop matched. Using default static backdrop.") default_backdrop_path = os.path.normpath(r"E:\Pratham\2025\Harsh Sir\Scratch Vision\images\plain_white.svg") default_backdrop_name = os.path.basename(default_backdrop_path) default_backdrop_sound = os.path.normpath(r"E:\Pratham\2025\Harsh Sir\Scratch Vision\images\83a9787d4cb6f3b7632b4ddfebf74367.wav") default_backdrop_sound_name = os.path.basename(default_backdrop_sound) try: shutil.copy2(default_backdrop_path, os.path.join(project_folder, default_backdrop_name)) logger.info(f"✅ Default backdrop copied to project: {default_backdrop_name}") shutil.copy2(default_backdrop_sound, os.path.join(project_folder, default_backdrop_sound_name)) logger.info(f"✅ Default backdrop sound copied to project: {default_backdrop_sound_name}") except Exception as e: logger.error(f"❌ Failed to copy default backdrop assets: {e}") final_project["targets"].append({ "isStage": True, "name": "Stage", "objName": "Stage", "variables": {}, "lists": {}, "broadcasts": {}, "blocks": {}, "comments": {}, "currentCostume": 0, "costumes": [ { "assetId": default_backdrop_name.split(".")[0], "name": "defaultBackdrop", "md5ext": default_backdrop_name, "dataFormat": "png", "rotationCenterX": 240, "rotationCenterY": 180 } ], "sounds": [ { "name": "pop", "assetId": "83a9787d4cb6f3b7632b4ddfebf74367", "dataFormat": "wav", "format": "", "rate": 48000, "sampleCount": 1123, "md5ext": "83a9787d4cb6f3b7632b4ddfebf74367.wav" } ], "volume": 100, "layerOrder": 0, "tempo": 60, "videoTransparency": 50, "videoState": "on", "textToSpeechLanguage": None }) with open(project_json_path, 'w') as f: json.dump(final_project, f, indent=2) validate_and_copy_assets( action_planner_path=r"E:\Pratham\2025\Harsh Sir\Scratch Vision\action_plan.json", project_json_path=project_json_path, project_folder=project_folder, sprite_base_path=sprite_base_path, backdrop_base_path=backdrop_base_path, final_project = final_project ) # project_data.extend(extra_sprites) # backdrop_data.extend(extra_backdrops) # logger.info(f"🎉 Final project saved: {project_json_path}") return project_json_path def validate_and_copy_assets(action_planner_path, project_json_path, project_folder,sprite_base_path, backdrop_base_path, final_project): # step 1: Load action_plan.json and get the first sprite name after "Stage":{...} with open(action_planner_path, 'r', encoding='utf-8') as f: planner_data = json.load(f) script_for = None keys = list(planner_data.keys()) if len(keys) > 1: script_for = keys[1] else: print("No sprite name found after 'Stage'") return print(f"\n\n=================script_for = '{script_for}'\n\n") # step 2: check in project.json if sprite exists sprite_found = False with open(project_json_path, 'r', encoding='utf-8') as f: project_data = json.load(f) for target in project_data.get("targets", []): if target.get("name", "").lower() == script_for.lower(): sprite_found = True break if sprite_found: print(f"'{script_for}' found in project.json – no asset copy needed.") return # Step 3: Search in backdrops and sprites folders base_dirs = [sprite_base_path, backdrop_base_path] # project_data = [] found_sprites = [] # backdrop_data = [] found_backdrops = [] # search in sprites if os.path.exists(sprite_base_path): for sub_dir in os.listdir(sprite_base_path): folder_path = os.path.join(sprite_base_path, sub_dir) sprite_json_path = os.path.join(folder_path, "sprite.json") if os.path.isdir(folder_path) and os.path.exists(sprite_json_path): with open(sprite_json_path, 'r', encoding='utf-8') as f: sprite_info = json.load(f) if sprite_info.get("name", "").lower() == script_for.lower(): print(f"Found matching sprite in {sprite_json_path}") # Copy all assets except json for fname in os.listdir(folder_path): if fname=='sprite.json': continue shutil.copy2(os.path.join(folder_path, fname), os.path.join(project_folder, fname)) found_sprites.append(sprite_info) break # search in backdrops if os.path.exists(backdrop_base_path): for sub_dir in os.listdir(backdrop_base_path): folder_path = os.path.join(backdrop_base_path, sub_dir) proj_json_path = os.path.join(folder_path, "project.json") if os.path.isdir(folder_path) and os.path.exists(proj_json_path): with open(proj_json_path, 'r', encoding='utf-8') as f: bd_json = json.load(f) for tgt in bd_json.get("targets", []): if tgt.get("name", "").lower() == script_for.lower(): print(f"Found matching backdrop in {proj_json_path}") # Copy all assets except JSON for fname in os.listdir(folder_path): if fname == "project.json": continue shutil.copy2(os.path.join(folder_path, fname), os.path.join(project_folder, fname)) found_backdrops.append(tgt) break # return found_sprites, found_backdrops # Merge into final project.json for spr in found_sprites: if not spr.get("isStage", False): final_project["targets"].append(spr) for bd in found_backdrops: if bd.get("isStage", False): final_project["targets"].insert(0, bd) with open(project_json_path, 'w', encoding='utf-8') as f: json.dump(final_project, f, indent=2) if found_sprites or found_backdrops: print(f"✅ Updated {project_json_path} with missing '{script_for}' assets.") else: print(f"⚠️ No matching '{script_for}' assets found in sprites/backdrops.") # --- ASYNC PDF to Image Conversion --- async def convert_pdf_to_images_async(pdf_stream: io.BytesIO, dpi=150): loop = asyncio.get_event_loop() return await loop.run_in_executor(None, lambda: convert_bytes_to_image(pdf_stream.getvalue(), dpi)) # pdf_name = os.path.splitext(os.path.basename(pdf_path))[0] # output_image_folder = os.path.join(IMAGE_FOLDER_PATH, pdf_name) # loop = asyncio.get_event_loop() # with ThreadPoolExecutor() as pool: # # Pass poppler_path explicitly # result = await loop.run_in_executor( # pool, convert_pdf_to_images_sync, pdf_path, output_image_folder, dpi, poppler_path # ) # return result def convert_bytes_to_image(pdf_bytes: bytes, dpi: int): images = convert_from_bytes(pdf_bytes, dpi=dpi, poppler_path=poppler_path) # Save each page to an in-memory BytesIO and return a list of BytesIOs buffers = [] for img in images: buf = BytesIO() img.save(buf, format="PNG") buf.seek(0) buffers.append(buf) return buffers # # Blocking version used internally # def convert_pdf_to_images_sync(pdf_path, output_image_folder, dpi, poppler_path): # pdf_name = os.path.splitext(os.path.basename(pdf_path))[0] # output_image_folder = os.path.join("outputs", "SCANNED_IMAGE", pdf_name) # os.makedirs(output_image_folder, exist_ok=True) # print(f"[INFO] Converting PDF: {pdf_path}") # print(f"[INFO] Output folder: {output_image_folder}") # # print(f"[INFO] Using Poppler path: {poppler_path}") # try: # images = convert_from_path(pdf_path, dpi=dpi, poppler_path=poppler_path) # image_paths = [] # for i, img in enumerate(images): # output_path = os.path.join(output_image_folder, f"page_{i+1}.png") # img.save(output_path, "PNG") # print(f"[DEBUG] Saved: {output_path}") # image_paths.append(output_path) # return image_paths # except PDFInfoNotInstalledError as e: # raise RuntimeError(f"Poppler not installed or path incorrect: {str(e)}") # except Exception as e: # print(f"[ERROR] Failed to convert PDF: {e}") # raise # def delay_for_tpm_node(state: GameState): # logger.info("--- Running DelayForTPMNode ---") # time.sleep(10) # Adjust the delay as needed # logger.info("Delay completed.") # return state # # Build the LangGraph workflow # workflow = StateGraph(GameState) # # Add all nodes to the workflow # workflow.add_node("timer_delay",delay_for_tpm_node) # workflow.add_node("opcode_counter", plan_logic_aligner_node) # workflow.set_entry_point("timer_delay") # workflow.add_edge("timer_delay","opcode_counter") # workflow.add_edge("opcode_counter", END) # app_graph = workflow.compile() # def get_desc_pseudo(image_buf: BytesIO, pseudo_store: dict, project_folder: str): # """ # Takes a path to a code-block image and returns a dict with: # - 'pseudo_code': pseudo-code representing logic in Scratch block format # Stores the output into outputs/pseudo_output.json # """ # try: # # Load image and encode to base64 # # with open(image_buf, "rb") as image_file: # # image_bytes = image_file.read() # img_base64 = base64.b64encode(image_buf.getvalue()).decode() # # pseudo_store[image_key] = pseudo # # === Call LangGraph workflow (auto triggers plan_logic_aligner_node) === # state = app_graph.invoke({"image": img_base64}) # # { comment: to solve pseudo_output issue # # logic_refined = state.get("pseudo_node", {}).get("refined_logic", {}) # pseudo_node = state.get("pseudo_node", {}) # name_variable = pseudo_node.get("name_variable", "Unknown") # pseudocode = pseudo_node.get("pseudocode", "No logic extracted") # pseudo_store.setdefault(name_variable, []).append({"pseudo_code": pseudocode}) # # --- Extract fields --- # # refined = logic_refined.get("refined_logic", {}) # # {comment: to solve pseudo_output.json issue} # # name_variable = logic_refined.get("name_variable", "Unknown") # # pseudo_code_raw = logic_refined.get("pseudocode", "No logic extracted") # # if pseudo_store is not None: # # pseudo_store.setdefault(name_variable, []).append({"pseudo_code": pseudo_code_raw}) # #} # with open(os.path.join(project_folder, "pseudo_output.json"), "w") as f: # json.dump(pseudo_store, f, indent=2) # result = { # "name_variable": name_variable, # "pseudo_code": pseudocode # } # return result # except Exception as e: # logger.error(f"❌ get_desc_pseudo failed for {image_buf}: {e}") # return { # "error": str(e) # } # ============== Helper function to Upscale an Image ============== # def upscale_image(image: Image.Image, scale: int = 2) -> Image.Image: """ Upscales a PIL image by a given scale factor. """ try: width, height = image.size new_size = (width * scale, height * scale) upscaled_image = image.resize(new_size, Image.LANCZOS) logger.info(f"✅ Upscaled image to {new_size}") return upscaled_image except Exception as e: logger.error(f"❌ Error during image upscaling: {str(e)}") return image @app.route('/') def index(): return render_template('app_index.html') # API endpoint @app.route('/process_pdf', methods=['POST']) async def process_pdf(): start_time = time.time() try: logger.info("Received request to process PDF.") if 'pdf_file' not in request.files: logger.warning("No PDF file found in request.") return jsonify({"error": "Missing PDF file in form-data with key 'pdf_file'"}), 400 pdf_file = request.files['pdf_file'] if pdf_file.filename == '': return jsonify({"error": "Empty filename"}), 400 # ================================================= # # Generate Random UUID for project folder name # # ================================================= # random_id = str(uuid.uuid4()).replace('-', '') project_folder = os.path.join("outputs", f"project_{random_id}") # =========================================================================== # # Create empty json in project_{random_id} folder # # =========================================================================== # os.makedirs(project_folder, exist_ok=True) # Save the uploaded PDF temporarily # filename = secure_filename(pdf_file.filename) # # temp_dir = tempfile.mkdtemp() # # saved_pdf_path = os.path.join(temp_dir, filename) # saved_pdf_path = os.path.join(DETECTED_IMAGE_FOLDER_PATH, filename) # pdf_file.save(saved_pdf_path) pdf_bytes = pdf_file.read() pdf_stream = io.BytesIO(pdf_bytes) # logger.info(f"Created project folder: {project_folder}") logger.info(f"Saved uploaded PDF to: {pdf_stream}") # Extract & process json_path = None # output_path, result = extract_images_from_pdf(saved_pdf_path, json_path) # print(f"\n\n OUTPUT_PATH: \n{output_path}\n") manipulated_sprites = extract_images_from_pdf(pdf_stream) # print(f"\nRESULT: {result}\n") # project_output = similarity_matching(output_path, project_folder) project_output = similarity_matching(manipulated_sprites, project_folder) # Call the async function from sync code try: image_paths = await convert_pdf_to_images_async(pdf_stream) print("PDF converted to images:", image_paths) # # Create an in-memory store for pseudo-codes # pseudo_store = {} # [get_desc_pseudo(img_buf, pseudo_store, project_folder) for img_buf in image_paths] except Exception as e: print(f"Error processing PDF: {e}") # Convert in-memory images to base64 strings scanned_images_b64 = [ base64.b64encode(buf.getvalue()).decode("utf-8") for buf in image_paths ] total_time = time.time() - start_time # ⏳ End timer logger.info(f"⏱ Total processing time for PDF: {total_time:.2f} seconds") return jsonify({ "message": "✅ PDF processed successfully", # "output_json": output_path, # "sprites": filtered_sprites, "project_output_json": project_output, "scanned_images": scanned_images_b64, # "scanned_image_pseudo": pseudo_results }) except Exception as e: logger.exception("❌ Failed to process PDF") return jsonify({"error": f"❌ Failed to process PDF: {str(e)}"}), 500 if __name__ == '__main__': app.run(host='0.0.0.0', port=7860, debug=True)