parserPDF / ui /gradio_ui.py
semmyk's picture
baseline08_beta0.2.2_30Sept25: fix oauth_token. - fixing timeout: progress tracker
42d6e84
raw
history blame
52.5 kB
# ui/gradio_ui.py
import gradio as gr
from concurrent.futures import ProcessPoolExecutor, as_completed
import asyncio
from pathlib import Path, WindowsPath
from typing import Optional, Union #, Dict, List, Any, Tuple
from huggingface_hub import get_token
from numpy import append, iterable
#import file_handler
import file_handler.file_utils
from utils.config import TITLE, DESCRIPTION, DESCRIPTION_PDF_HTML, DESCRIPTION_PDF, DESCRIPTION_HTML, DESCRIPTION_MD
from utils.utils import is_dict, is_list_of_dicts
from file_handler.file_utils import zip_processed_files, process_dicts_data, collect_pdf_paths, collect_html_paths, collect_markdown_paths, create_outputdir ## should move to handling file
from file_handler.file_utils import find_file
from utils.get_config import get_config_value
#from llm.hf_client import HFChatClient ## SMY: unused. See converters.extraction_converter
from llm.provider_validator import is_valid_provider, suggest_providers
from llm.llm_login import is_loggedin_huggingface, login_huggingface
from converters.extraction_converter import DocumentConverter as docconverter #DocumentExtractor #as docextractor
from converters.pdf_to_md import PdfToMarkdownConverter, init_worker
#from converters.md_to_pdf import MarkdownToPdfConverter
#from converters.html_to_md import HtmlToMarkdownConverter ##SMY: PENDING: implementation
import traceback ## Extract, format and print information about Python stack traces.
from utils.logger import get_logger
logger = get_logger(__name__) ##NB: setup_logging() ## set logging
# Instantiate converters class once – they are stateless
pdf2md_converter = PdfToMarkdownConverter()
#html2md_converter = HtmlToMarkdownConverter()
#md2pdf_converter = MarkdownToPdfConverter()
# User eXperience: Load Marker models ahead of time if not already loaded in reload mode
## SMY: 29Sept2025 - Came across https://github.com/xiaoyao9184/docker-marker/tree/master/gradio
from converters.extraction_converter import load_models
from globals import config_load_models
try:
if not config_load_models.model_dict:
config_load_models.model_dict = load_models()
'''if 'model_dict' not in globals():
global model_dict
model_dict = load_models()'''
except Exception as exc:
#tb = traceback.format_exc() #exc.__traceback__
logger.exception(f"βœ— Error loading models (reload): {exc}") #\n{tb}")
raise RuntimeError(f"βœ— Error loading models (reload): {exc}") #\n{tb}")
def get_login_token( api_token_arg, oauth_token: gr.OAuthToken | None=None,):
""" Use user's supplied token or Get token from logged-in users, else from token stored on the machine. Return token"""
#oauth_token = get_token() if oauth_token is not None else api_token_arg
if api_token_arg != '': # or not None: #| None:
oauth_token = api_token_arg
elif oauth_token:
oauth_token = oauth_token
else: get_token()
return oauth_token.token if oauth_token else '' ##token value or empty string
# pool executor to convert files called by Gradio
##SMY: TODO: future: refactor to gradio_process.py and
## pull options to cli-options{"output_format":, "output_dir_string":, "use_llm":, "page_range":, "force_ocr":, "debug":, "strip_existing_ocr":, "disable_ocr_math""}
def convert_batch(
pdf_files, #: list[str],
pdf_files_count: int,
provider: str,
model_id: str,
#base_url: str
hf_provider: str,
endpoint: str,
backend_choice: str,
system_message: str,
max_tokens: int,
temperature: float,
top_p: float,
stream: bool,
api_token_gr: str,
#max_workers: int,
#max_retries: int,
openai_base_url: str = "https://router.huggingface.co/v1",
openai_image_format: Optional[str] = "webp",
max_workers: Optional[int] = 4,
max_retries: Optional[int] = 2,
output_format: str = "markdown",
#output_dir: Optional[Union[str, Path]] = "output_dir",
output_dir_string: str = "output_dir_default",
use_llm: bool = False, #Optional[bool] = False, #True,
page_range: str = None, #Optional[str] = None,
tz_hours: str = None,
oauth_token: gr.OAuthToken | None=None,
progress: gr.Progress = gr.Progress(), #Progress tracker to keep tab on pool queue executor
): #-> str:
"""
Handles the conversion process using multiprocessing.
Spins up a pool and converts all uploaded files in parallel.
Aggregates per-file logs into one string.
Receives Gradio component values, starting with the list of uploaded file paths
"""
# login: Update the Gradio UI to improve user-friendly eXperience - commencing
#yield gr.update(interactive=False), f"Commencing Processing ... Getting login", {"process": "Commencing Processing"}, f"dummy_log.log"
#progress((0,16), f"Commencing Processing ...")
# get token from logged-in user:
api_token = get_login_token(api_token_arg=api_token_gr, oauth_token=oauth_token)
##SMY: Strictly debug. Must not be live
#logger.log(level=30, msg="Commencing: get_login_token", extra={"api_token]": api_token, "api_token_gr": api_token_gr})
try:
##SMY: might deprecate. To replace with oauth login from Gradio ui or integrate cleanly.
#login_huggingface(api_token) ## attempt login if not already logged in. NB: HF CLI login prompt would not display in Process Worker.
if is_loggedin_huggingface() and (api_token is None or api_token == ""):
api_token = get_token() ##SMY: might be redundant
elif is_loggedin_huggingface() is False and api_token:
login_huggingface(api_token)
# login: Update the Gradio UI to improve user-friendly eXperience
#yield gr.update(interactive=False), f"login to HF: Processing files...", {"process": "Processing files"}, f"dummy_log.log"
else:
pass
# login: Update the Gradio UI to improve user-friendly eXperience
#yield gr.update(interactive=False), f"Not logged in to HF: Processing files...", {"process": "Processing files"}, f"dummy_log.log"
except Exception as exc: # Catch all exceptions
tb = traceback.format_exc()
logger.exception(f"βœ— Error during login_huggingface β†’ {exc}\n{tb}", exc_info=True) # Log the full traceback
return [gr.update(interactive=True), f"βœ— An error occurred during login_huggingface β†’ {exc}\n{tb}", {"Error":f"Error: {exc}"}, f"dummy_log.log"] # return the exception message
#progress((1,16), desc=f"Log in: {is_loggedin_huggingface}")
## debug
#logger.log(level=30, msg="pdf_files_inputs", extra={"input_arg[0]:": pdf_files[0]})
#if not files:
if not pdf_files or pdf_files is None: ## Check if files is None. This handles the case where no files are uploaded.
logger.log(level=30, msg="Initialising ProcessPool: No files uploaded.", extra={"pdf_files": pdf_files, "files_len": pdf_files_count})
#outputs=[log_output, files_individual_JSON, files_individual_downloads],
return [gr.update(interactive=True), "Initialising ProcessPool: No files uploaded.", {"Upload":"No files uploaded"}, f"dummy_log.log"]
#progress((2,16), desc=f"Getting configuration values")
# Get config values if not provided
config_file = find_file("config.ini") ##from file_handler.file_utils
model_id = get_config_value(config_file, "MARKER_CAP", "MODEL_ID") if not model_id else model_id
openai_base_url = get_config_value(config_file, "MARKER_CAP", "OPENAI_BASE_URL") if not openai_base_url else openai_base_url
openai_image_format = get_config_value(config_file, "MARKER_CAP", "OPENAI_IMAGE_FORMAT") if not openai_image_format else openai_image_format
max_workers = get_config_value(config_file, "MARKER_CAP", "MAX_WORKERS") if not max_workers else max_workers
max_retries = get_config_value(config_file, "MARKER_CAP", "MAX_RETRIES") if not max_retries else max_retries
output_format = get_config_value(config_file, "MARKER_CAP", "OUTPUT_FORMAT") if not output_format else output_format
output_dir_string = str(get_config_value(config_file, "MARKER_CAP", "OUTPUT_DIR") if not output_dir_string else output_dir_string)
use_llm = get_config_value(config_file, "MARKER_CAP", "USE_LLM") if not use_llm else use_llm
page_range = get_config_value(config_file,"MARKER_CAP", "PAGE_RANGE") if not page_range else page_range
#progress((3,16), desc="Retrieved configuration values")
# Create the initargs tuple from the Gradio inputs: # 'files' is an iterable, and handled separately.
#progress((4,16), desc=f"Initialiasing init_args")
yield gr.update(interactive=False), f"Initialising init_args", {"process": "Processing files ..."}, f"dummy_log.log"
init_args = (
provider,
model_id,
#base_url,
hf_provider,
endpoint,
backend_choice,
system_message,
max_tokens,
temperature,
top_p,
stream,
api_token,
openai_base_url,
openai_image_format,
max_workers,
max_retries,
output_format,
output_dir_string,
use_llm,
page_range,
)
#global docextractor ##SMY: deprecated.
try:
results = [] ## initialised pool result holder
# Create a pool with init_worker initialiser
logger.log(level=30, msg="Initialising ProcessPoolExecutor: pool:", extra={"pdf_files": pdf_files, "files_len": len(pdf_files), "model_id": model_id, "output_dir": output_dir_string}) #pdf_files_count
#progress((5,16), desc=f"Initialising ProcessPoolExecutor: Processing Files ...")
yield gr.update(interactive=False), f"Initialising ProcessPoolExecutor: Processing Files ...", {"process": "Processing files ..."}, f"dummy_log.log"
with ProcessPoolExecutor(
max_workers=max_workers,
initializer=init_worker,
initargs=init_args
) as pool:
#logger.log(level=30, msg="Initialising ProcessPoolExecutor: pool:", extra={"pdf_files": pdf_files, "files_len": len(pdf_files), "model_id": model_id, "output_dir": output_dir_string}) #pdf_files_count
#progress((6,16), desc=f"Starting ProcessPool queue: Processing Files ...")
# Update the Gradio UI to improve user-friendly eXperience
#outputs=[process_button, log_output, files_individual_JSON, files_individual_downloads],
# Map the files (pdf_files) to the conversion function (pdf2md_converter.convert_file)
# The 'docconverter' argument is implicitly handled by the initialiser
#futures = [pool.map(pdf2md_converter.convert_files, f) for f in pdf_files]
#logs = [f.result() for f in as_completed(futures)]
#futures = [pool.submit(pdf2md_converter.convert_files, file) for file in pdf_files]
#logs = [f.result() for f in futures]
try:
#(7,16), desc=f"ProcessPoolExecutor: Creating output_dir")
yield gr.update(interactive=False), f"Creating output_dir ...", {"process": "Processing files ..."}, f"dummy_log.log"
pdf2md_converter.output_dir_string = output_dir_string ##SMY: attempt setting directly to resolve pool.map iterable
#progress((8,16), desc=f"ProcessPoolExecutor: Created output_dir.")
yield gr.update(interactive=False), f"Created output_dir ...", {"process": "Processing files ..."}, f"dummy_log.log"
except Exception as exc:
# Raise the exception to stop the Gradio app: exception to halt execution
logger.exception("Error during creating output_dir", exc_info=True) # Log the full traceback
traceback.print_exc() # Print the exception traceback
#return f"An error occurred during pool.map: {str(exc)}", f"Error: {exc}", f"Error: {exc}" ## return the exception message
# Update the Gradio UI to improve user-friendly eXperience
yield gr.update(interactive=True), f"An error occurred creating output_dir: {str(exc)}", {"Error":f"Error: {exc}"}, f"dummy_log.log" ## return the exception message
try:
#progress((9,16), desc=f"ProcessPoolExecutor: Pooling file conversion ...")
yield gr.update(interactive=True), f"ProcessPoolExecutor: Pooling file conversion ...", {"process": "Processing files ..."}, f"dummy_log.log"
# Use progress.tqdm to integrate with the executor map
#results = pool.map(pdf2md_converter.convert_files, pdf_files) ##SMY iterables #max_retries #output_dir_string)
for result_interim in progress.tqdm(
iterable=pool.map(pdf2md_converter.convert_files, pdf_files), total=len(pdf_files)
):
results.append(result_interim)
#progress((10,16), desc=f"ProcessPoolExecutor: Pooling file conversion result: [{str(result_interim)}[:20]]")
# Update the Gradio UI to improve user-friendly eXperience
yield gr.update(interactive=True), f"ProcessPoolExecutor: Pooling file conversion result: [{str(result_interim)}[:20]]", {"process": "Processing files ..."}, f"dummy_log.log"
#progress((11,16), desc=f"ProcessPoolExecutor: Got Results from files conversion")
yield gr.update(interactive=True), f"rocessPoolExecutor: Got Results from files conversion: [{str(result_interim)}[:20]]", {"process": "Processing files ..."}, f"dummy_log.log"
except Exception as exc:
# Raise the exception to stop the Gradio app: exception to halt execution
logger.exception("Error during pooling file conversion", exc_info=True) # Log the full traceback
traceback.print_exc() # Print the exception traceback
return [gr.update(interactive=True), f"An error occurred during pool.map: {str(exc)}", {"Error":f"Error: {exc}"}, f"dummy_log.log"] ## return the exception message
# Update the Gradio UI to improve user-friendly eXperience
#yield gr.update(interactive=True), f"An error occurred during pool.map: {str(exc)}", {"Error":f"Error: {exc}"}, f"dummy_log.log" ## return the exception message
#'''
try:
logger.log(level=20, msg="ProcessPoolExecutor pool result:", extra={"results": str(results)})
logs = []
logs_files_images = []
#logs.extend(results) ## performant pythonic
#logs = list[results] ##
logs = [result for result in results] ## pythonic list comprehension
## logs : [file , images , filepath, image_path]
#logs_files_images = logs_files.extend(logs_images) #zip(logs_files, logs_images) ##SMY: in progress
logs_count = 0
#for log in logs:
for i, log in enumerate(logs):
logs_files_images.append(log.get("filepath") if is_dict(log) or is_list_of_dicts(logs) else "Error or no file_path") # isinstance(log, (dict, str))
logs_files_images.extend(list(image for image in log.get("image_path", "Error or no image_path")))
i_image = log.get("images", 0)
# Update the Gradio UI to improve user-friendly eXperience
#yield gr.update(interactive=False), f"Processing files: {logs_files_images[logs_count]}", {"process": "Processing files"}, f"dummy_log.log"
logs_count = i+i_image
#progress((12,16), desc="Processing results from files conversion") ##rekickin
#logs_files_images.append(logs_filepath) ## to del
#logs_files_images.extend(logs_images) ## to del
except Exception as exc:
logger.exception("Error during processing results logs β†’ {exc}\n{tb}", exc_info=True) # Log the full traceback
traceback.print_exc() # Print the exception traceback
return [gr.update(interactive=True), f"An error occurred during processing results logs: {str(exc)}\n{tb}", {"Error":f"Error: {exc}"}, f"dummy_log.log"] ## return the exception message
#yield gr.update(interactive=True), f"An error occurred during processing results logs: {str(exc)}\n{tb}", {"Error":f"Error: {exc}"}, f"dummy_log.log" ## return the exception message
#'''
except Exception as exc:
tb = traceback.format_exc()
logger.exception(f"βœ— Error during ProcessPoolExecutor β†’ {exc}\n{tb}" , exc_info=True) # Log the full traceback
#traceback.print_exc() # Print the exception traceback
yield gr.update(interactive=True), f"βœ— An error occurred during ProcessPoolExecutorβ†’ {exc}\n{tb}", {"Error":f"Error: {exc}"}, f"dummy_log.log" # return the exception message
'''
logger.log(level=20, msg="ProcessPoolExecutor pool result:", extra={"results": str(results)})
logs = []
#logs.extend(results) ## performant pythonic
#logs = list[results] ##
logs = [result for result in results] ## pythonic list comprehension
'''
# Zip Processed md Files and images. Insert to first index
try: ##from file_handler.file_utils
#progress((13,16), desc="Zipping processed files and images")
zipped_processed_files = zip_processed_files(root_dir=f"data/{output_dir_string}", file_paths=logs_files_images, tz_hours=tz_hours, date_format='%d%b%Y_%H-%M-%S') #date_format='%d%b%Y'
logs_files_images.insert(0, zipped_processed_files)
#logs_files_images.insert(1, "====================")
#progress((14,16), desc="Zipped processed files and images")
#yield gr.update(interactive=False), f"Processing zip and files: {logs_files_images}", {"process": "Processing files"}, f"dummy_log.log"
except Exception as exc:
tb = traceback.format_exc()
logger.exception(f"βœ— Error during zipping processed files β†’ {exc}\n{tb}" , exc_info=True) # Log the full traceback
#traceback.print_exc() # Print the exception traceback
#return gr.update(interactive=True), f"βœ— An error occurred during zipping files β†’ {exc}\n{tb}", f"Error: {exc}", f"Error: {exc}" # return the exception message
yield gr.update(interactive=True), f"βœ— An error occurred during zipping files β†’ {exc}\n{tb}", {"Error":f"Error: {exc}"}, f"dummy_log.log" # return the exception message
# Return processed files log
try:
#progress((15,16), desc="Formatting processed log results")
## # Convert logs list of dicts to formatted json string
logs_return_formatted_json_string = file_handler.file_utils.process_dicts_data(logs) #"\n".join(log for log in logs) ##SMY outputs to gr.JSON component with no need for json.dumps(data, indent=)
#logs_files_images_return = "\n".join(path for path in logs_files_images) ##TypeError: sequence item 0: expected str instance, WindowsPath found
##convert the List of Path objects to List of string for gr.Files output
#logs_files_images_return = list(str(path) for path in logs_files_images)
## # Convert any Path objects to strings, but leave strings as-is
logs_files_images_return = list(str(path) if isinstance(path, Path) else path for path in logs_files_images)
logger.log(level=20, msg="File conversion complete. Sending outcome to Gradio:", extra={"logs_files_image_return": str(logs_files_images_return)}) ## debug: FileNotFoundError: [WinError 2] The system cannot find the file specified: 'Error or no image_path'
#progress((16,16), desc="Complete processing and formatting file processing results")
#outputs=[process_button, log_output, files_individual_JSON, files_individual_downloads],
#return "\n".join(logs), "\n".join(logs_files_images) #"\n".join(logs_files)
yield gr.update(interactive=True), gr.update(value=logs_return_formatted_json_string), gr.update(value=logs_return_formatted_json_string, visible=True), gr.update(value=logs_files_images_return, visible=True) ##SMY: redundant
return [gr.update(interactive=True), gr.update(value=logs_return_formatted_json_string), gr.update(value=logs_return_formatted_json_string, visible=True), gr.update(value=logs_files_images_return, visible=True)]
#yield gr.update(interactive=True), logs_return_formatted_json_string, logs_return_formatted_json_string, logs_files_images_return
#return [gr.update(interactive=True), logs_return_formatted_json_string, logs_return_formatted_json_string, logs_files_images_return]
except Exception as exc:
tb = traceback.format_exc()
logger.exception(f"βœ— Error during returning result logs β†’ {exc}\n{tb}" , exc_info=True) # Log the full traceback
#traceback.print_exc() # Print the exception traceback
#return [gr.update(interactive=True), f"βœ— An error occurred during returning result logsβ†’ {exc}\n{tb}", {"Error":f"Error: {exc}"}, f"dummy_log.log"] # return the exception message
yield [gr.update(interactive=True), f"βœ— An error occurred during returning result logsβ†’ {exc}\n{tb}", {"Error":f"Error: {exc}"}, f"dummy_log.log"] # return the exception message
#return "\n".join(log for log in logs), "\n".join(str(path) for path in logs_files_images)
#print(f'logs_files_images: {"\n".join(str(path) for path in logs_files_images)}')
# files wrapping into list ##SMY: Flagged for deprecation
def pdf_files_wrap(files: list[str]):
# explicitly wrap file object in a list
return [files] if not isinstance(files, list) else files
#return [files]
##====================
## SMY: moved to logic file: See pdf_to_md.py. Currently unused
def convert_pdfs_to_md(file: gr.File | None, folder: str | None) -> dict:
"""
Gradio callback for PDF β†’ Markdown.
Accepts either a single file or a folder path (recursively).
Leverages Marker, a pipeline of deep learning models, for conversion
Returns a dictionary of filename β†’ Markdown string.
"""
if not file and not folder:
return {"error": "Please provide a PDF file or a folder."}
pdf_paths = []
# Single file
if file:
pdf_path = Path(file.name)
pdf_paths.append(pdf_path)
# Folder (recursively)
if folder:
try:
pdf_paths.extend(collect_pdf_paths(folder))
except Exception as exc:
logger.exception("Folder traversal failed.")
return {"error": str(exc)}
if not pdf_paths:
return {"error": "No PDF files found."}
results = pdf2md_converter.batch_convert(pdf_paths)
# Gradio expects a dict of {filename: content}
return results
## SMY: to be implemented AND to refactor and moved to logic file
def convert_md_to_pdf(file: gr.File | None, folder: str | None) -> list[gr.File]:
"""
Gradio callback for Markdown β†’ PDF.
Returns a list of generated PDF files (as Gradio File objects).
"""
if not file and not folder:
return []
md_paths = []
# Single file
if file:
md_path = Path(file.name)
md_paths.append(md_path)
# Folder
if folder:
try:
md_paths.extend(collect_markdown_paths(folder))
except Exception as exc:
logger.exception("Folder traversal failed.")
return []
if not md_paths:
return []
output_dir = Path("./generated_pdfs")
output_dir.mkdir(exist_ok=True)
pdf_files = md2pdf_converter.batch_convert(md_paths, output_dir)
# Convert to Gradio File objects
gr_files = [gr.File(path=str(p)) for p in pdf_files]
return gr_files
## SMY: to refactor and moved to logic file. Currently unused
'''
def convert_htmls_to_md(file: gr.File | None, folder: str | None) -> dict:
"""
Gradio callback for HTML β†’ Markdown.
Accepts either a single file or a folder path (recursively).
Returns a dictionary of filename β†’ Markdown string.
"""
if not file and not folder:
return {"error": "Please provide a HTML file or a folder."}
html_paths = []
# Single file
if file:
html_path = Path(file.name)
html_paths.append(html_path)
# Folder (recursively)
if folder:
try:
html_paths.extend(collect_html_paths(folder))
except Exception as exc:
logger.exception("Folder traversal failed.")
return {"error": str(exc)}
if not html_paths:
return {"error": "No HTML files found."}
results = html2md_converter.batch_convert(html_paths)
# Gradio expects a dict of {filename: content}
return results
'''
##====================
def build_interface() -> gr.Blocks:
"""
Assemble the Gradio Blocks UI.
"""
# Use custom CSS to style the file component
custom_css = """
.file-or-directory-area {
border: 2px dashed #ccc;
padding: 20px;
text-align: center;
border-radius: 8px;
margin-bottom: 10px;
display: flex;
flex-direction: column;
align-items: center;
}
.file-or-directory-area:hover {
border-color: #007bff;
background-color: #f8f9fa;
}
.gradio-upload-btn {
margin-top: 10px;
}
"""
##SMY: flagged; to move to file_handler.file_utils
def is_file_with_extension(path_obj: Path) -> bool:
"""
Checks if a pathlib.Path object is a file and has a non-empty extension.
"""
path_obj = path_obj if isinstance(path_obj, Path) else Path(path_obj) if isinstance(path_obj, str) else None
return path_obj.is_file() and bool(path_obj.suffix)
##SMY: flagged; to move to file_handler.file_utils
def accumulate_files(uploaded_files, current_state):
"""
Accumulates newly uploaded files with the existing state.
"""
# Initialize state if it's the first run
if current_state is None:
current_state = []
# If no files were uploaded in this interaction, return the current state unchanged
if not uploaded_files:
return current_state, f"No new files uploaded. Still tracking {len(current_state)} file(s)."
# Get the temporary paths of the newly uploaded files
# call is_file_with_extension to check if pathlib.Path object is a file and has a non-empty extension
new_file_paths = [f.name for f in uploaded_files if is_file_with_extension(Path(f.name))] #Path(f.name) and Path(f.name).is_file() and bool(Path(f.name).suffix)] #Path(f.name).suffix.lower() !=""]
# Concatenate the new files with the existing ones in the state
updated_files = current_state + new_file_paths
updated_filenames = [Path(f).name for f in updated_files]
# Return the updated state and a message to the user
#file_info = "\n".join(updated_files)
filename_info = "\n".join(updated_filenames)
#message = f"Accumulated {len(updated_files)} file(s) total.\n\nAll file paths:\n{file_info}"
message = f"Accumulated {len(updated_files)} file(s) total: \n{filename_info}"
return updated_files, message
# with gr.Blocks(title=TITLE) as demo
with gr.Blocks(title=TITLE, css=custom_css) as demo:
gr.Markdown(f"## {DESCRIPTION}")
# Clean UI: Model parameters hidden in collapsible accordion
with gr.Accordion("βš™οΈ LLM Model Settings", open=False):
gr.Markdown(f"#### **Backend Configuration**")
system_message = gr.Textbox(
label="System Message",
lines=2,
)
with gr.Row():
provider_dd = gr.Dropdown(
choices=["huggingface", "openai"],
label="Provider",
value="huggingface",
#allow_custom_value=True,
)
backend_choice = gr.Dropdown(
choices=["model-id", "provider", "endpoint"],
label="HF Backend Choice",
) ## SMY: ensure HFClient maps correctly
model_tb = gr.Textbox(
label="Model ID",
value="meta-llama/Llama-4-Maverick-17B-128E-Instruct", #image-Text-to-Text #"openai/gpt-oss-120b", ##Text-to-Text
)
endpoint_tb = gr.Textbox(
label="Endpoint",
placeholder="Optional custom endpoint",
)
with gr.Row():
max_token_sl = gr.Slider(
label="Max Tokens",
minimum=1,
maximum=131172, #65536, #32768, #16384, #8192,
value=1024, #512,
step=1,
)
temperature_sl = gr.Slider(
label="Temperature",
minimum=0.0,
maximum=1.0,
value=0.0,
step=0.1, #0.01
)
top_p_sl = gr.Slider(
label="Top-p",
minimum=0.0,
maximum=1.0,
value=0.1,
step=0.1, #0.01
)
with gr.Column():
stream_cb = gr.Checkbox(
label="LLM Streaming",
value=False,
)
#tz_hours_tb = gr.Textbox(value=None, label="TZ Hours", placeholder="Timezone in numbers", max_lines=1,)
tz_hours_num = gr.Number(label="TZ Hours", placeholder="Timezone in numbers", min_width=5,)
with gr.Row():
api_token_tb = gr.Textbox(
label="API Token [OPTIONAL]",
type="password",
placeholder="hf_xxx or openai key"
)
hf_provider_dd = gr.Dropdown(
choices=["fireworks-ai", "together-ai", "openrouter-ai", "hf-inference"],
value="fireworks-ai",
label="Provider",
allow_custom_value=True, # let users type new providers as they appear
)
# Clean UI: Model parameters hidden in collapsible accordion
with gr.Accordion("βš™οΈ Marker Settings", open=False):
gr.Markdown(f"#### **Marker Configuration**")
with gr.Row():
openai_base_url_tb = gr.Textbox(
label="OpenAI Base URL: Default HuggingFace",
value="https://router.huggingface.co/v1",
lines=1,
max_lines=1,
)
openai_image_format_dd = gr.Dropdown(
choices=["webp", "png", "jpeg"],
label="OpenAI Image Format",
value="webp",
)
output_format_dd = gr.Dropdown(
choices=["markdown", "html"], #, "json", "chunks"], ##SMY: To be enabled later
#choices=["markdown", "html", "json", "chunks"],
label="Output Format",
value="markdown",
)
output_dir_tb = gr.Textbox(
label="Output Directory",
value="output_dir", #"output_md",
lines=1,
max_lines=1,
)
with gr.Row():
max_workers_sl = gr.Slider(
label="Max Worker",
minimum=1,
maximum=7,
value=4,
step=1
)
max_retries_sl = gr.Slider(
label="Max Retry",
minimum=1,
maximum=3,
value=2,
step=1 #0.01
)
use_llm_cb = gr.Checkbox(
label="Use LLM for Marker conversion",
value=False
)
page_range_tb = gr.Textbox(
label="Page Range (Optional)",
placeholder="Example: 0,1-5,8,12-15",
lines=1,
max_lines=1,
)
with gr.Accordion("πŸ€— HuggingFace Client Logout", open=True): #, open=False):
# Logout controls
with gr.Row():
#hf_login_logout_btn = gr.LoginButton(value="Sign in to HuggingFace πŸ€—", logout_value="Clear Session & Logout of HF: ({})", variant="huggingface")
hf_login_logout_btn = gr.LoginButton(value="Sign in to HuggingFace πŸ€—", logout_value="Logout of HF: ({}) πŸ€—", variant="huggingface")
#logout_btn = gr.Button("Logout from session & HF (inference) Client", variant="stop", )
logout_status_md = gr.Markdown(visible=True) #visible=False)
# The gr.State component to hold the accumulated list of files
uploaded_file_list = gr.State([]) ##NB: initial value of `gr.State` must be able to be deepcopied
# --- PDF & HTML β†’ Markdown tab ---
with gr.Tab(" πŸ“„ PDF & HTML ➜ Markdown"):
gr.Markdown(f"#### {DESCRIPTION_PDF_HTML}")
### flag4deprecation #earlier implementation
'''
pdf_files = gr.File(
label="Upload PDF, HTML or PDF and HTMLfiles",
file_count="directory", ## handle directory and files upload #"multiple",
type="filepath",
file_types=["pdf", ".pdf"],
#size="small",
)
pdf_files_count = gr.TextArea(label="Files Count", interactive=False, lines=1)
with gr.Row():
btn_pdf_count = gr.Button("Count Files")
#btn_pdf_upload = gr.UploadButton("Upload files")
btn_pdf_convert = gr.Button("Convert PDF(s)")
'''
with gr.Column(elem_classes=["file-or-directory-area"]):
with gr.Row():
file_btn = gr.UploadButton(
#file_btn = gr.File(
label="Upload Multiple Files",
file_count="multiple",
file_types=["file"],
#height=25, #"sm",
size="sm",
elem_classes=["gradio-upload-btn"]
)
dir_btn = gr.UploadButton(
#dir_btn = gr.File(
label="Upload a Directory",
file_count="directory",
#file_types=["file"], #Warning: The `file_types` parameter is ignored when `file_count` is 'directory'
#height=25, #"0.5",
size="sm",
elem_classes=["gradio-upload-btn"]
)
with gr.Accordion("Display uploaded", open=True):
# Displays the accumulated file paths
output_textbox = gr.Textbox(label="Accumulated Files", lines=3) #, max_lines=4) #10
with gr.Row():
process_button = gr.Button("Process All Uploaded Files", variant="primary")
clear_button = gr.Button("Clear All Uploads", variant="secondary")
# --- PDF β†’ Markdown tab ---
with gr.Tab(" πŸ“„ PDF ➜ Markdown (Flag for DEPRECATION)", interactive=False, visible=True): #False
gr.Markdown(f"#### {DESCRIPTION_PDF}")
files_upload_pdf = gr.File(
label="Upload PDF files",
file_count="directory", ## handle directory and files upload #"multiple",
type="filepath",
file_types=["pdf", ".pdf"],
#size="small",
)
files_count = gr.TextArea(label="Files Count", interactive=False, lines=1) #pdf_files_count
with gr.Row():
btn_pdf_count = gr.Button("Count Files")
#btn_pdf_upload = gr.UploadButton("Upload files")
btn_pdf_convert = gr.Button("Convert PDF(s)")
# --- πŸ“ƒ HTML β†’ Markdown tab ---
with gr.Tab("πŸ•ΈοΈ HTML ➜ Markdown: (Flag for DEPRECATION)", interactive=False, visible=False):
gr.Markdown(f"#### {DESCRIPTION_HTML}")
files_upload_html = gr.File(
label="Upload HTML files",
file_count="multiple",
type="filepath",
file_types=["html", ".html", "htm", ".htm"]
)
#btn_html_convert = gr.Button("Convert HTML(s)")
html_files_count = gr.TextArea(label="Files Count", interactive=False, lines=1)
with gr.Row():
btn_html_count = gr.Button("Count Files")
#btn_pdf_upload = gr.UploadButton("Upload files")
btn_html_convert = gr.Button("Convert PDF(s)")
# --- Markdown β†’ PDF tab ---
with gr.Tab("PENDING: Markdown ➜ PDF", interactive=False):
gr.Markdown(f"#### {DESCRIPTION_MD}")
md_files = gr.File(
label="Upload Markdown files",
file_count="multiple",
type="filepath",
file_types=["md", ".md"]
)
btn_md_convert = gr.Button("Convert Markdown to PDF)")
output_pdf = gr.Gallery(label="Generated PDFs", elem_id="pdf_gallery")
'''
md_input = gr.File(label="Upload a single Markdown file", file_count="single")
md_folder_input = gr.Textbox(
label="Or provide a folder path (recursively)",
placeholder="/path/to/folder",
)
convert_md_btn = gr.Button("Convert Markdown to PDF")
output_pdf = gr.Gallery(label="Generated PDFs", elem_id="pdf_gallery")
convert_md_btn.click(
fn=convert_md_to_pdf,
inputs=[md_input, md_folder_input],
outputs=output_pdf,
)
'''
# A Files component to display individual processed files as download links
with gr.Accordion("⏬ View and Download processed files", open=True): #, open=False
##SMY: future
zip_btn = gr.DownloadButton("Download Zip file of all processed files", visible=False) #.Button()
# Placeholder to download zip file of processed files
download_zip_file = gr.File(label="Download processed Files (ZIP)", interactive=False, visible=False) #, height="1"
with gr.Row():
files_individual_JSON = gr.JSON(label="Serialised JSON list", max_height=250, visible=False)
files_individual_downloads = gr.Files(label="Individual Processed Files", visible=False)
## Displays processed file paths
with gr.Accordion("View processing log", open=True): #open=False):
log_output = gr.Textbox(
label="Conversion Logs",
lines=5,
#max_lines=25,
#interactive=False
)
# Initialise gr.State
state_max_workers = gr.State(4) #max_workers_sl,
state_max_retries = gr.State(2) #max_retries_sl,
state_tz_hours = gr.State(value=None)
state_api_token = gr.State(None)
processed_file_state = gr.State([]) ##SMY: future: View and Download processed files
def update_state_stored_value(new_component_input):
""" Updates stored state: use for max_workers and max_retries """
return new_component_input
# Update gr.State values on slider components change. NB: initial value of `gr.State` must be able to be deepcopied
max_workers_sl.change(update_state_stored_value, inputs=max_workers_sl, outputs=state_max_workers)
max_retries_sl.change(update_state_stored_value, inputs=max_retries_sl, outputs=state_max_retries)
tz_hours_num.change(update_state_stored_value, inputs=tz_hours_num, outputs=state_tz_hours)
api_token_tb.change(update_state_stored_value, inputs=api_token_tb, outputs=state_api_token)
# LLM Setting: Validate provider on change; warn but allow continue
def on_provider_change(provider_value: str):
if not provider_value:
return
if not is_valid_provider(provider_value):
sug = suggest_providers(provider_value)
extra = f" Suggestions: {', '.join(sug)}." if sug else ""
gr.Warning(
f"Provider not on HF provider list. See https://huggingface.co/docs/inference-providers/index.{extra}"
)
hf_provider_dd.change(on_provider_change, inputs=hf_provider_dd, outputs=None)
# HuggingFace Client Logout
'''def get_login_token(state_api_token_arg, oauth_token: gr.OAuthToken | None=None):
#oauth_token = get_token() if oauth_token is not None else state_api_token
#oauth_token = oauth_token if oauth_token else state_api_token_arg
if oauth_token:
print(oauth_token)
return oauth_token
else:
oauth_token = get_token()
print(oauth_token)
return oauth_token'''
#'''
def do_logout(): ##SMY: use with clear_state() as needed
try:
#ok = docextractor.client.logout()
ok = docconverter.client.logout()
# Reset token textbox on successful logout
#msg = "βœ… Logged out of HuggingFace and cleared tokens. Remember to log out of HuggingFace completely." if ok else "⚠️ Logout failed."
msg = "βœ… Session Cleared. Remember to close browser." if ok else "⚠️ HF client closing failed."
return msg
#return gr.update(value=""), gr.update(visible=True, value=msg), gr.update(value="Sign in to HuggingFace πŸ€—"), gr.update(value="Clear session")
except AttributeError:
msg = "⚠️ HF client closing failed."
return msg
#return gr.update(value=""), gr.update(visible=True, value=msg), gr.update(value="Sign in to HuggingFace πŸ€—"), gr.update(value="Clear session", interactive=False)
#'''
def do_logout_hf():
try:
ok = docconverter.client.logout()
# Reset token textbox on successful logout
msg = "βœ… Session Cleared. Remember to close browser." if ok else "⚠️ Logout & Session Cleared"
#return gr.update(value=""), gr.update(visible=True, value=msg), gr.update(value="Sign in to HuggingFace πŸ€—"), gr.update(value="Clear session", interactive=False)
return msg
#yield msg ## generator for string
except AttributeError:
msg = "⚠️ Logout. No HF session"
return msg
#yield msg ## generator for string
#def custom_do_logout(hf_login_logout_btn_arg: gr.LoginButton, state_api_token_arg: gr.State):
def custom_do_logout():
#global state_api_token
''' ##SMY: TO DELETE
try:
state_api_token_get= get_token() if "Clear Session & Logout of HF" in hf_login_logout_btn_arg.value else state_api_token_arg.value
except AttributeError:
#state_api_token_get= get_token() if "Clear Session & Logout of HF" in hf_login_logout_btn_arg else state_api_token_arg
state_api_token_get = get_login_token(state_api_token_arg)
'''
#do_logout()
#return gr.update(value="Sign in to HuggingFace πŸ€—")
msg = do_logout_hf()
##debug
#msg = "βœ… Session Cleared. Remember to close browser." if "Clear Session & Logout of HF" in hf_login_logout_btn else "⚠️ Logout" # & Session Cleared"
return gr.update(value="Sign in to HuggingFace πŸ€—"), gr.update(value=""), gr.update(visible=True, value=msg) #, state_api_token_arg
#yield gr.update(value="Sign in to HuggingFace πŸ€—"), gr.update(value=""), gr.update(visible=True, value=msg)
# Files, status, session clearing
def clear_state():
"""
Clears the accumulated state of uploaded file list, output textbox, files and directory upload.
"""
#msg = f"Files list cleared: {do_logout()}" ## use as needed
msg = f"Files list cleared."
yield [], msg, '', ''
#return [], f"Files list cleared.", [], []
#hf_login_logout_btn.click(fn=custom_do_logout, inputs=None, outputs=hf_login_logout_btn)
##unused
###hf_login_logout_btn.click(fn=custom_do_logout, inputs=[hf_login_logout_btn, state_api_token], outputs=[hf_login_logout_btn, api_token_tb, logout_status_md, state_api_token])
###logout_btn.click(fn=do_logout, inputs=None, outputs=[api_token_tb, logout_status_md, hf_login_logout_btn, logout_btn])
#logout_btn.click(fn=clear_state, inputs=None, outputs=[uploaded_file_list, output_textbox, log_output, api_token_tb])
hf_login_logout_btn.click(fn=custom_do_logout, inputs=None, outputs=[hf_login_logout_btn, api_token_tb, logout_status_md]) #, state_api_token])
# --- PDF & HTML β†’ Markdown tab ---
# Event handler for the multiple file upload button
file_btn.upload(
fn=accumulate_files,
inputs=[file_btn, uploaded_file_list],
outputs=[uploaded_file_list, output_textbox]
)
# Event handler for the directory upload button
dir_btn.upload(
fn=accumulate_files,
inputs=[dir_btn, uploaded_file_list],
outputs=[uploaded_file_list, output_textbox]
)
# Event handler for the "Clear" button
clear_button.click(
fn=clear_state,
inputs=None,
outputs=[uploaded_file_list, output_textbox, file_btn, dir_btn],
)
# file inputs
## [wierd] NB: inputs_arg is a list of Gradio component objects, not the values of those components.
## inputs_arg variable captures the state of these components at the time the list is created.
## When btn_convert.click() is called later, it uses the list as it was initially defined
##
## SMY: Gradio component values are not directly mutable.
## Instead, you should pass the component values to a function,
## and then use the return value of the function to update the component.
## Discarding for now. #//TODO: investigate further.
## SMY: Solved: using gr.State
inputs_arg = [
#pdf_files,
##pdf_files_wrap(pdf_files), # wrap pdf_files in a list (if not already)
uploaded_file_list,
files_count, #pdf_files_count,
provider_dd,
model_tb,
hf_provider_dd,
endpoint_tb,
backend_choice,
system_message,
max_token_sl,
temperature_sl,
top_p_sl,
stream_cb,
api_token_tb, #state_api_token, #api_token_tb,
#gr.State(4), # max_workers
#gr.State(3), # max_retries
openai_base_url_tb,
openai_image_format_dd,
state_max_workers, #gr.State(4), #max_workers_sl,
state_max_retries, #gr.State(2), #max_retries_sl,
output_format_dd,
output_dir_tb,
use_llm_cb,
page_range_tb,
tz_hours_num, #state_tz_hours
]
## debug
#logger.log(level=30, msg="About to execute btn_pdf_convert.click", extra={"files_len": pdf_files_count, "pdf_files": pdf_files})
try:
#logger.log(level=30, msg="input_arg[0]: {input_arg[0]}")
process_button.click(
#pdf_files.upload(
fn=convert_batch,
inputs=inputs_arg,
outputs=[process_button, log_output, files_individual_JSON, files_individual_downloads],
)
except Exception as exc:
tb = traceback.format_exc()
logger.exception(f"βœ— Error during process_button.click β†’ {exc}\n{tb}", exc_info=True)
msg = "βœ— An error occurred during process_button.click" # β†’
#return f"βœ— An error occurred during process_button.click β†’ {exc}\n{tb}"
return gr.update(interactive=True), f"{msg} β†’ {exc}\n{tb}", f"{msg} β†’ {exc}", f"{msg} β†’ {exc}"
##gr.File .upload() event, fire only after a file has been uploaded
# Event handler for the pdf file upload button
files_upload_pdf.upload(
fn=accumulate_files,
inputs=[files_upload_pdf, uploaded_file_list],
outputs=[uploaded_file_list, log_output]
)
#inputs_arg[0] = files_upload
btn_pdf_convert.click(
#pdf_files.upload(
fn=convert_batch,
outputs=[btn_pdf_convert, log_output, files_individual_JSON, files_individual_downloads],
inputs=inputs_arg,
)
# )
# reuse the same business logic for HTML tab
# Event handler for the pdf file upload button
files_upload_html.upload(
fn=accumulate_files,
inputs=[files_upload_html, uploaded_file_list],
outputs=[uploaded_file_list, log_output]
)
#inputs_arg[0] = html_files
btn_html_convert.click(
fn=convert_batch,
inputs=inputs_arg,
outputs=[btn_html_convert,log_output, files_individual_JSON, files_individual_downloads]
)
def get_file_count(file_list):
"""
Counts the number of files in the list.
Args:
file_list (list): A list of temporary file objects.
Returns:
str: A message with the number of uploaded files.
"""
if file_list:
return f"{len(file_list)}", f"Upload: {len(file_list)} files: \n {file_list}" #{[pdf_files.value]}"
else:
return "No files uploaded.", "No files uploaded." # Count files button
btn_pdf_count.click(
fn=get_file_count,
inputs=[files_upload_pdf],
outputs=[files_count, log_output]
)
btn_html_count.click(
fn=get_file_count,
inputs=[files_upload_html],
outputs=[html_files_count, log_output]
)
# Validate files upload on change; warn but allow continue
def on_pdf_files_change(pdf_files_value: list[str]):
# explicitly wrap file object in a list
pdf_files_value = pdf_files_wrap(pdf_files_value)
#if not isinstance(pdf_files_value, list):
# pdf_files_value = [pdf_files_value]
pdf_files_path = [file.name for file in pdf_files_value]
pdf_files_len = len(pdf_files_value) #len(pdf_files_path)
if pdf_files_value:
#return
return pdf_files_path, pdf_files_len
#pdf_files.change(on_pdf_files_change, inputs=pdf_files, outputs=[log_output, pdf_files_count]) #, postprocess=False) ##debug
return demo