"""Thread-safe, production-ready PDF pipeline ================================================ A self-contained, thread-safe PDF conversion pipeline exploiting parallelism between pipeline stages and models. * **Per-run isolation** - every :py:meth:`execute` call uses its own bounded queues and worker threads so that concurrent invocations never share mutable state. * **Deterministic run identifiers** - pages are tracked with an internal *run-id* instead of relying on :pyfunc:`id`, which may clash after garbage collection. * **Explicit back-pressure & shutdown** - producers block on full queues; queue *close()* propagates downstream so stages terminate deterministically without sentinels. * **Minimal shared state** - heavyweight models are initialised once per pipeline instance and only read by worker threads; no runtime mutability is exposed. * **Strict typing & clean API usage** - code is fully annotated and respects *coding_rules.md*. """ from __future__ import annotations import itertools import logging import threading import time import warnings from collections import defaultdict, deque from dataclasses import dataclass, field from pathlib import Path from typing import Any, Callable, Iterable, Sequence, cast import numpy as np from docling_core.types.doc import ( DocItem, ImageRef, PageItem, PictureItem, Size, TableItem, ) from docling.backend.abstract_backend import AbstractDocumentBackend from docling.backend.pdf_backend import PdfDocumentBackend, PdfPageBackend from docling.datamodel.base_models import ( AssembledUnit, ConversionStatus, DoclingComponentType, ErrorItem, FailureCategory, Page, ) from docling.datamodel.document import ConversionResult from docling.datamodel.pipeline_options import ThreadedPdfPipelineOptions from docling.datamodel.settings import settings from docling.models.factories import ( get_layout_factory, get_ocr_factory, get_table_structure_factory, ) from docling.models.stages.code_formula.code_formula_vlm_model import ( CodeFormulaVlmModel, ) from docling.models.stages.heading_hierarchy.heading_hierarchy_model import ( HeadingHierarchyModel, ) from docling.models.stages.page_assemble.page_assemble_model import ( PageAssembleModel, PageAssembleOptions, ) from docling.models.stages.page_preprocessing.page_preprocessing_model import ( PagePreprocessingModel, PagePreprocessingOptions, ) from docling.models.stages.reading_order.readingorder_model import ( ReadingOrderModel, ReadingOrderOptions, ) from docling.pipeline.base_pipeline import ConvertPipeline from docling.utils.profiling import ProfilingScope, TimeRecorder from docling.utils.utils import chunkify _log = logging.getLogger(__name__) STAGE_FAILURE_CATEGORY = { "ocr": FailureCategory.INFERENCE_FAILURE, "layout": FailureCategory.INFERENCE_FAILURE, "table": FailureCategory.INFERENCE_FAILURE, "assemble": FailureCategory.INFERENCE_FAILURE, } # ────────────────────────────────────────────────────────────────────────────── # Helper data structures # ────────────────────────────────────────────────────────────────────────────── def _make_error_item( *, component_type: DoclingComponentType, module_name: str, error: Exception, category: FailureCategory, page_no: int | None = None, ) -> ErrorItem: return ErrorItem( component_type=component_type, module_name=module_name, error_message=str(error) or error.__class__.__name__, category=category, page_no=page_no, ) @dataclass class ThreadedItem: """Envelope that travels between pipeline stages.""" payload: Page | None run_id: int # Unique per *execute* call, monotonic across pipeline instance page_no: int conv_res: ConversionResult error: Exception | None = None failure: ErrorItem | None = None is_failed: bool = False @dataclass class ProcessingResult: """Aggregated outcome of a pipeline run.""" pages: list[Page] = field(default_factory=list) failed_pages: list[tuple[int, Exception, ErrorItem | None]] = field( default_factory=list ) total_expected: int = 0 @property def success_count(self) -> int: return len(self.pages) @property def failure_count(self) -> int: return len(self.failed_pages) @property def is_partial_success(self) -> bool: return 0 < self.success_count < self.total_expected @property def is_complete_failure(self) -> bool: return self.success_count == 0 and self.failure_count > 0 class ThreadedQueue: """Bounded queue with blocking put/ get_batch and explicit *close()* semantics.""" __slots__ = ("_closed", "_items", "_lock", "_max", "_not_empty", "_not_full") def __init__(self, max_size: int) -> None: self._max: int = max_size self._items: deque[ThreadedItem] = deque() self._lock = threading.Lock() self._not_full = threading.Condition(self._lock) self._not_empty = threading.Condition(self._lock) self._closed = False # ---------------------------------------------------------------- put() def put(self, item: ThreadedItem, timeout: float | None = None) -> bool: """Block until queue accepts *item* or is closed. Returns *False* if closed.""" with self._not_full: if self._closed: return False start = time.monotonic() while len(self._items) >= self._max and not self._closed: if timeout is not None: remaining = timeout - (time.monotonic() - start) if remaining <= 0: return False self._not_full.wait(remaining) else: self._not_full.wait() if self._closed: return False self._items.append(item) self._not_empty.notify() return True # ------------------------------------------------------------ get_batch() def get_batch(self, size: int, timeout: float | None = None) -> list[ThreadedItem]: """Return up to *size* items. Blocks until ≥1 item present or queue closed/timeout.""" with self._not_empty: start = time.monotonic() while not self._items and not self._closed: if timeout is not None: remaining = timeout - (time.monotonic() - start) if remaining <= 0: return [] self._not_empty.wait(remaining) else: self._not_empty.wait() batch: list[ThreadedItem] = [] while self._items and len(batch) < size: batch.append(self._items.popleft()) if batch: self._not_full.notify_all() return batch # ---------------------------------------------------------------- close() def close(self) -> None: with self._lock: self._closed = True self._not_empty.notify_all() self._not_full.notify_all() # -------------------------------------------------------------- property @property def closed(self) -> bool: return self._closed class ThreadedPipelineStage: """A single pipeline stage backed by one worker thread.""" def __init__( self, *, name: str, model: Any, batch_size: int, batch_timeout: float, queue_max_size: int, postprocess: Callable[[ThreadedItem], None] | None = None, timed_out_run_ids: set[int] | None = None, ) -> None: self.name = name self.model = model self.batch_size = batch_size self.batch_timeout = batch_timeout self.input_queue = ThreadedQueue(queue_max_size) self._outputs: list[ThreadedQueue] = [] self._thread: threading.Thread | None = None self._running = False self._postprocess = postprocess self._timed_out_run_ids = ( timed_out_run_ids if timed_out_run_ids is not None else set() ) # ---------------------------------------------------------------- wiring def add_output_queue(self, q: ThreadedQueue) -> None: self._outputs.append(q) # -------------------------------------------------------------- lifecycle def start(self) -> None: if self._running: return self._running = True self._thread = threading.Thread( target=self._run, name=f"Stage-{self.name}", daemon=False ) self._thread.start() def stop(self) -> None: if not self._running: return self._running = False self.input_queue.close() if self._thread is not None: # Give thread 15s to finish naturally before abandoning self._thread.join(timeout=15.0) if self._thread.is_alive(): _log.warning( "Stage %s thread did not terminate within 15s. " "Thread is likely stuck in a blocking call and will be abandoned (resources may leak).", self.name, ) # ------------------------------------------------------------------ _run def _run(self) -> None: try: while self._running: batch = self.input_queue.get_batch(self.batch_size, self.batch_timeout) if not batch and self.input_queue.closed: break processed = self._process_batch(batch) self._emit(processed) except Exception: # pragma: no cover - top-level guard _log.exception("Fatal error in stage %s", self.name) finally: for q in self._outputs: q.close() # ----------------------------------------------------- _process_batch() def _process_batch(self, batch: Sequence[ThreadedItem]) -> list[ThreadedItem]: """Run *model* on *batch* grouped by run_id to maximise batching.""" groups: dict[int, list[ThreadedItem]] = defaultdict(list) for itm in batch: groups[itm.run_id].append(itm) result: list[ThreadedItem] = [] for rid, items in groups.items(): # If run_id is timed out, skip processing but pass through items as-is # This allows already-completed work to flow through while aborting new work if rid in self._timed_out_run_ids: for it in items: it.is_failed = True if it.error is None: it.error = RuntimeError("document timeout exceeded") if it.failure is None: error = it.error or RuntimeError("document timeout exceeded") it.failure = _make_error_item( component_type=DoclingComponentType.PIPELINE, module_name=self.name, error=error, category=FailureCategory.TIMEOUT, page_no=it.page_no, ) result.extend(items) continue good: list[ThreadedItem] = [i for i in items if not i.is_failed] if not good: result.extend(items) continue try: # Filter out None payloads and ensure type safety pages_with_payloads = [ (i, i.payload) for i in good if i.payload is not None ] if len(pages_with_payloads) != len(good): # Some items have None payloads, mark all as failed for it in good: it.is_failed = True error = RuntimeError("Page payload is None") it.error = error it.failure = _make_error_item( component_type=DoclingComponentType.PIPELINE, module_name=self.name, error=error, category=FailureCategory.UNKNOWN, page_no=it.page_no, ) result.extend(items) continue pages: list[Page] = [payload for _, payload in pages_with_payloads] if _log.isEnabledFor(logging.DEBUG): _t_start = time.time() _t_mono = time.monotonic() processed_pages = list(self.model(good[0].conv_res, pages)) # type: ignore[arg-type] if _log.isEnabledFor(logging.DEBUG): _log.debug( "PIPELINE_PROFILING Stage %s: run_id=%d pages=%s start=%.3f end=%.3f duration=%.3fs", self.name, rid, [it.page_no for it in good], _t_start, time.time(), time.monotonic() - _t_mono, ) if len(processed_pages) != len(pages): # strict mismatch guard raise RuntimeError( f"Model {self.name} returned wrong number of pages" ) for idx, page in enumerate(processed_pages): result.append( ThreadedItem( payload=page, run_id=rid, page_no=good[idx].page_no, conv_res=good[idx].conv_res, ) ) except Exception as exc: _log.error( "Stage %s failed for run %d: %s", self.name, rid, exc, exc_info=True ) for it in good: it.is_failed = True it.error = exc it.failure = _make_error_item( component_type=DoclingComponentType.MODEL, module_name=self.name, error=exc, category=STAGE_FAILURE_CATEGORY.get( self.name, FailureCategory.UNKNOWN ), page_no=it.page_no, ) result.extend(items) return result # -------------------------------------------------------------- _emit() def _emit(self, items: Iterable[ThreadedItem]) -> None: for item in items: if self._postprocess is not None: self._postprocess(item) for q in self._outputs: if not q.put(item): _log.error("Output queue closed while emitting from %s", self.name) class PreprocessThreadedStage(ThreadedPipelineStage): """Pipeline stage that validates pre-attached backends and runs preprocessing.""" def __init__( self, *, batch_timeout: float, queue_max_size: int, model: Any, timed_out_run_ids: set[int] | None = None, ) -> None: super().__init__( name="preprocess", model=model, batch_size=1, batch_timeout=batch_timeout, queue_max_size=queue_max_size, timed_out_run_ids=timed_out_run_ids, ) def _process_batch(self, batch: Sequence[ThreadedItem]) -> list[ThreadedItem]: groups: dict[int, list[ThreadedItem]] = defaultdict(list) for itm in batch: groups[itm.run_id].append(itm) result: list[ThreadedItem] = [] for rid, items in groups.items(): # If run_id is timed out, skip processing but pass through items as-is # This allows already-completed work to flow through while aborting new work if rid in self._timed_out_run_ids: for it in items: it.is_failed = True if it.error is None: it.error = RuntimeError("document timeout exceeded") if it.failure is None: error = it.error or RuntimeError("document timeout exceeded") it.failure = _make_error_item( component_type=DoclingComponentType.PIPELINE, module_name=self.name, error=error, category=FailureCategory.TIMEOUT, page_no=it.page_no, ) result.extend(items) continue good = [i for i in items if not i.is_failed] if not good: result.extend(items) continue # Validate backends before the model call so that invalid-page # items are emitted exactly once, even if the model later raises. invalid: list[ThreadedItem] = [] valid: list[tuple[ThreadedItem, Page]] = [] for it in good: page = it.payload if page is None: it.is_failed = True error = RuntimeError("Page payload is None") it.error = error it.failure = _make_error_item( component_type=DoclingComponentType.PIPELINE, module_name=self.name, error=error, category=FailureCategory.UNKNOWN, page_no=it.page_no, ) invalid.append(it) elif page._backend is None: it.is_failed = True error = RuntimeError( "Page backend must be attached before preprocess" ) it.error = error it.failure = _make_error_item( component_type=DoclingComponentType.PIPELINE, module_name=self.name, error=error, category=FailureCategory.UNKNOWN, page_no=it.page_no, ) invalid.append(it) elif not page._backend.is_valid(): it.is_failed = True error = RuntimeError(f"Page {page.page_no} failed to parse.") it.error = error it.failure = _make_error_item( component_type=DoclingComponentType.DOCUMENT_BACKEND, module_name=self.name, error=error, category=FailureCategory.BACKEND_FAILURE, page_no=it.page_no, ) invalid.append(it) else: valid.append((it, page)) result.extend(invalid) if not valid: continue try: if _log.isEnabledFor(logging.DEBUG): _t_start = time.time() _t_mono = time.monotonic() pages = [page for _, page in valid] processed_pages = list( self.model(valid[0][0].conv_res, pages) # type: ignore[arg-type] ) if _log.isEnabledFor(logging.DEBUG): _log.debug( "PIPELINE_PROFILING Stage preprocess: run_id=%d pages=%s start=%.3f end=%.3f duration=%.3fs", rid, [it.page_no for it, _ in valid], _t_start, time.time(), time.monotonic() - _t_mono, ) if len(processed_pages) != len(pages): raise RuntimeError( "PagePreprocessingModel returned unexpected number of pages" ) for idx, processed_page in enumerate(processed_pages): result.append( ThreadedItem( payload=processed_page, run_id=rid, page_no=valid[idx][0].page_no, conv_res=valid[idx][0].conv_res, ) ) except Exception as exc: _log.error( "Stage preprocess failed for run %d, pages %s: %s", rid, [it.page_no for it, _ in valid], exc, exc_info=False, ) for it, _ in valid: it.is_failed = True it.error = exc it.failure = _make_error_item( component_type=DoclingComponentType.MODEL, module_name=self.name, error=exc, category=FailureCategory.UNKNOWN, page_no=it.page_no, ) result.extend(it for it, _ in valid) return result @dataclass class RunContext: """Wiring for a single *execute* call.""" stages: list[ThreadedPipelineStage] first_stage: ThreadedPipelineStage output_queue: ThreadedQueue timed_out_run_ids: set[int] = field(default_factory=set) # ────────────────────────────────────────────────────────────────────────────── # Main pipeline # ────────────────────────────────────────────────────────────────────────────── class StandardPdfPipeline(ConvertPipeline): """High-performance PDF pipeline with multi-threaded stages.""" def __init__(self, pipeline_options: ThreadedPdfPipelineOptions) -> None: super().__init__(pipeline_options) self.pipeline_options: ThreadedPdfPipelineOptions = pipeline_options self._run_seq = itertools.count(1) # deterministic, monotonic run ids self._page_sizes_by_no: dict[int, Size] = {} # initialise heavy models once self._init_models() # ──────────────────────────────────────────────────────────────────────── # Heavy-model initialisation & helpers # ──────────────────────────────────────────────────────────────────────── def _init_models(self) -> None: art_path = self.artifacts_path self.keep_images = ( self.pipeline_options.generate_page_images or self.pipeline_options.generate_picture_images or self.pipeline_options.generate_table_images ) self.preprocessing_model = PagePreprocessingModel( options=PagePreprocessingOptions( images_scale=self.pipeline_options.images_scale ) ) self.ocr_model = self._make_ocr_model(art_path) layout_factory = get_layout_factory( allow_external_plugins=self.pipeline_options.allow_external_plugins ) self.layout_model = layout_factory.create_instance( options=self.pipeline_options.layout_options, artifacts_path=art_path, accelerator_options=self.pipeline_options.accelerator_options, enable_remote_services=self.pipeline_options.enable_remote_services, ) table_factory = get_table_structure_factory( allow_external_plugins=self.pipeline_options.allow_external_plugins ) self.table_model = table_factory.create_instance( options=self.pipeline_options.table_structure_options, enabled=self.pipeline_options.do_table_structure, artifacts_path=art_path, accelerator_options=self.pipeline_options.accelerator_options, enable_remote_services=self.pipeline_options.enable_remote_services, ) self.assemble_model = PageAssembleModel(options=PageAssembleOptions()) self.reading_order_model = ReadingOrderModel(options=ReadingOrderOptions()) self.heading_hierarchy_model = HeadingHierarchyModel( options=self.pipeline_options.heading_hierarchy_options ) # --- optional enrichment ------------------------------------------------ # Create a copy to avoid mutating pipeline_options in-place, # which would change its hash and break pipeline caching (#3109). code_formula_opts = self.pipeline_options.code_formula_options.model_copy( update={ "extract_code": self.pipeline_options.do_code_enrichment, "extract_formulas": self.pipeline_options.do_formula_enrichment, } ) self.enrichment_pipe = [ # Code Formula Enrichment Model (using new VLM runtime system) CodeFormulaVlmModel( enabled=self.pipeline_options.do_code_enrichment or self.pipeline_options.do_formula_enrichment, artifacts_path=self.artifacts_path, options=code_formula_opts, accelerator_options=self.pipeline_options.accelerator_options, enable_remote_services=self.pipeline_options.enable_remote_services, ), *self.enrichment_pipe, ] self.keep_backend = any( ( self.pipeline_options.do_formula_enrichment, self.pipeline_options.do_code_enrichment, self.pipeline_options.do_picture_classification, self.pipeline_options.do_picture_description, self.pipeline_options.do_chart_extraction, ) ) # ---------------------------------------------------------------- helpers def _make_ocr_model(self, art_path: Path | None) -> Any: factory = get_ocr_factory( allow_external_plugins=self.pipeline_options.allow_external_plugins ) return factory.create_instance( options=self.pipeline_options.ocr_options, enabled=self.pipeline_options.do_ocr, artifacts_path=art_path, accelerator_options=self.pipeline_options.accelerator_options, ) def _release_page_resources(self, item: ThreadedItem) -> None: page = item.payload if page is None: return if not self.keep_images: page._image_cache = {} if not self.keep_backend and page._backend is not None: page._backend.unload() page._backend = None if not self.pipeline_options.generate_parsed_pages: page.parsed_page = None # ──────────────────────────────────────────────────────────────────────── # Build - thread pipeline # ──────────────────────────────────────────────────────────────────────── def _create_run_ctx(self) -> RunContext: opts = self.pipeline_options timed_out_run_ids: set[int] = set() preprocess = PreprocessThreadedStage( batch_timeout=opts.batch_polling_interval_seconds, queue_max_size=opts.queue_max_size, model=self.preprocessing_model, timed_out_run_ids=timed_out_run_ids, ) ocr = ThreadedPipelineStage( name="ocr", model=self.ocr_model, batch_size=opts.ocr_batch_size, batch_timeout=opts.batch_polling_interval_seconds, queue_max_size=opts.queue_max_size, timed_out_run_ids=timed_out_run_ids, ) layout = ThreadedPipelineStage( name="layout", model=self.layout_model, batch_size=opts.layout_batch_size, batch_timeout=opts.batch_polling_interval_seconds, queue_max_size=opts.queue_max_size, timed_out_run_ids=timed_out_run_ids, ) table = ThreadedPipelineStage( name="table", model=self.table_model, batch_size=opts.table_batch_size, batch_timeout=opts.batch_polling_interval_seconds, queue_max_size=opts.queue_max_size, timed_out_run_ids=timed_out_run_ids, ) assemble = ThreadedPipelineStage( name="assemble", model=self.assemble_model, batch_size=1, batch_timeout=opts.batch_polling_interval_seconds, queue_max_size=opts.queue_max_size, postprocess=self._release_page_resources, timed_out_run_ids=timed_out_run_ids, ) # wire stages output_q = ThreadedQueue(opts.queue_max_size) preprocess.add_output_queue(ocr.input_queue) ocr.add_output_queue(layout.input_queue) layout.add_output_queue(table.input_queue) table.add_output_queue(assemble.input_queue) assemble.add_output_queue(output_q) stages = [preprocess, ocr, layout, table, assemble] return RunContext( stages=stages, first_stage=preprocess, output_queue=output_q, timed_out_run_ids=timed_out_run_ids, ) # --------------------------------------------------------------------- build def _iter_requested_page_backends( self, backend: PdfDocumentBackend, expected_page_nos: list[int] ) -> Iterable[PdfPageBackend]: if backend.supports_random_page_access: for page_no in expected_page_nos: yield backend.load_page(page_no - 1) return expected_page_no_set = set(expected_page_nos) for page_backend in backend.iter_pages(): if page_backend.page_no in expected_page_no_set: yield page_backend def _get_expected_page_nos(self, conv_res: ConversionResult) -> list[int]: start_page, end_page = conv_res.input.limits.page_range return list( range( max(1, start_page), min(conv_res.input.page_count, end_page) + 1, ) ) def _build_document(self, conv_res: ConversionResult) -> ConversionResult: """Stream-build the document with a dedicated producer thread. Note: If a worker thread gets stuck in a blocking call (model inference or PDF backend iter_pages/get_size), that thread will be abandoned after a brief wait (15s) during cleanup. The thread continues running until the blocking call completes, potentially holding resources (e.g., pypdfium2_lock). """ self._page_sizes_by_no = {} run_id = next(self._run_seq) assert isinstance(conv_res.input._backend, PdfDocumentBackend) backend = conv_res.input._backend expected_page_nos = self._get_expected_page_nos(conv_res) if not expected_page_nos: conv_res.status = ConversionStatus.FAILURE return conv_res page_by_no: dict[int, Page] = {} for page_no in expected_page_nos: page = Page(page_no=page_no) conv_res.pages.append(page) page_by_no[page_no] = page total_pages: int = len(expected_page_nos) ctx: RunContext = self._create_run_ctx() for st in ctx.stages: st.start() proc = ProcessingResult(total_expected=total_pages) batch_size: int = 32 # drain chunk start_time = time.monotonic() timeout_exceeded = False producer_error: list[Exception] = [] def _completed_page_nos() -> set[int]: failed_page_nos = { page_no for page_no, _, _ in proc.failed_pages if page_no > 0 } return {page.page_no for page in proc.pages} | failed_page_nos def _produce_pages() -> None: try: for page_backend in self._iter_requested_page_backends( backend, expected_page_nos ): page = page_by_no.get(page_backend.page_no) if page is None: continue page._backend = page_backend try: page.size = page_backend.get_size() self._page_sizes_by_no[page.page_no] = page.size except Exception: if page_backend.is_valid(): raise if not ctx.first_stage.input_queue.put( ThreadedItem( payload=page, run_id=run_id, page_no=page.page_no, conv_res=conv_res, ) ): break except Exception as exc: producer_error.append(exc) _log.error("Producer failed for run %d: %s", run_id, exc, exc_info=True) finally: ctx.first_stage.input_queue.close() producer_thread = threading.Thread( target=_produce_pages, name=f"PageProducer-{run_id}", daemon=False ) producer_thread.start() try: while proc.success_count + proc.failure_count < total_pages: # Check timeout if ( self.pipeline_options.document_timeout is not None and not timeout_exceeded ): elapsed_time = time.monotonic() - start_time if elapsed_time > self.pipeline_options.document_timeout: _log.warning( f"Document processing time ({elapsed_time:.3f}s) " f"exceeded timeout of {self.pipeline_options.document_timeout:.3f}s" ) timeout_exceeded = True ctx.timed_out_run_ids.add(run_id) ctx.first_stage.input_queue.close() # Break immediately - don't wait for in-flight work break # Drain - pull whatever is ready from the output side out_batch = ctx.output_queue.get_batch(batch_size, timeout=0.05) for itm in out_batch: if itm.run_id != run_id: continue if itm.is_failed or itm.error: error = itm.error or RuntimeError("unknown error") proc.failed_pages.append((itm.page_no, error, itm.failure)) else: assert itm.payload is not None proc.pages.append(itm.payload) # Failure safety - downstream closed early if not out_batch and ctx.output_queue.closed: missing_page_nos = sorted( set(expected_page_nos) - _completed_page_nos() ) if missing_page_nos: error = ( producer_error[0] if producer_error else RuntimeError("pipeline terminated early") ) proc.failed_pages.extend( [ ( page_no, error, _make_error_item( component_type=DoclingComponentType.PIPELINE, module_name=self.__class__.__name__, error=error, category=FailureCategory.UNKNOWN, page_no=page_no, ), ) for page_no in missing_page_nos ] ) break # Mark remaining pages as failed if timeout occurred if timeout_exceeded: missing_page_nos = sorted( set(expected_page_nos) - _completed_page_nos() ) for page_no in missing_page_nos: error = RuntimeError("document timeout exceeded") proc.failed_pages.append( ( page_no, error, _make_error_item( component_type=DoclingComponentType.PIPELINE, module_name=self.__class__.__name__, error=error, category=FailureCategory.TIMEOUT, page_no=page_no, ), ) ) finally: for st in ctx.stages: st.stop() ctx.output_queue.close() producer_thread.join(timeout=15.0) if producer_thread.is_alive(): _log.warning( "Producer thread for run %d did not terminate within 15s and will be abandoned.", run_id, ) self._integrate_results(conv_res, proc, timeout_exceeded=timeout_exceeded) return conv_res # ---------------------------------------------------- integrate_results() def _integrate_results( self, conv_res: ConversionResult, proc: ProcessingResult, timeout_exceeded: bool = False, ) -> None: page_map = {p.page_no: p for p in proc.pages} # Only keep pages that successfully completed processing conv_res.pages = [ page_map[p.page_no] for p in conv_res.pages if p.page_no in page_map ] # Add error details from failed pages for page_no, error, failure in proc.failed_pages: if failure is not None: conv_res.errors.append(failure) continue conv_res.errors.append( _make_error_item( component_type=DoclingComponentType.PIPELINE, module_name=self.__class__.__name__, error=error or RuntimeError("Page failed to process."), category=FailureCategory.UNKNOWN, page_no=page_no if page_no > 0 else None, ) ) if timeout_exceeded and proc.total_expected > 0: # Timeout exceeded: add structured error and set PARTIAL_SUCCESS timeout_msg = ( f"Pipeline stage timeout: processed {len(proc.pages)}/{proc.total_expected} pages successfully, " f"{len(proc.failed_pages)} pages failed or incomplete." ) timeout_error = ErrorItem( component_type=DoclingComponentType.PIPELINE, module_name=self.__class__.__name__, error_message=timeout_msg, category=FailureCategory.TIMEOUT, ) conv_res.errors.append(timeout_error) conv_res.status = ConversionStatus.PARTIAL_SUCCESS elif proc.is_complete_failure: conv_res.status = ConversionStatus.FAILURE elif proc.is_partial_success: conv_res.status = ConversionStatus.PARTIAL_SUCCESS else: conv_res.status = ConversionStatus.SUCCESS if not self.keep_images: for p in conv_res.pages: p._image_cache = {} for p in conv_res.pages: if not self.keep_backend and p._backend is not None: p._backend.unload() if not self.pipeline_options.generate_parsed_pages: del p.parsed_page p.parsed_page = None # ---------------------------------------------------------------- assemble def _assemble_document(self, conv_res: ConversionResult) -> ConversionResult: elements, headers, body = [], [], [] with TimeRecorder(conv_res, "doc_assemble", scope=ProfilingScope.DOCUMENT): for p in conv_res.pages: if p.assembled: elements.extend(p.assembled.elements) headers.extend(p.assembled.headers) body.extend(p.assembled.body) conv_res.assembled = AssembledUnit( elements=elements, headers=headers, body=body ) conv_res.document = self.reading_order_model(conv_res) conv_res.document = self.heading_hierarchy_model(conv_res) # Generate page images in the output if self.pipeline_options.generate_page_images: for page in conv_res.pages: assert page.image is not None page_no = page.page_no conv_res.document.pages[page_no].image = ImageRef.from_pil( page.image, dpi=int(72 * self.pipeline_options.images_scale) ) # Generate images of the requested element types with warnings.catch_warnings(): # deprecated generate_table_images warnings.filterwarnings("ignore", category=DeprecationWarning) if ( self.pipeline_options.generate_picture_images or self.pipeline_options.generate_table_images ): scale = self.pipeline_options.images_scale for element, _level in conv_res.document.iterate_items(): if not isinstance(element, DocItem) or len(element.prov) == 0: continue if ( isinstance(element, PictureItem) and self.pipeline_options.generate_picture_images ) or ( isinstance(element, TableItem) and self.pipeline_options.generate_table_images ): page_ix = element.prov[0].page_no page = next( (p for p in conv_res.pages if p.page_no == page_ix), cast("Page", None), ) assert page is not None assert page.size is not None assert page.image is not None crop_bbox = ( element.prov[0] .bbox.scaled(scale=scale) .to_top_left_origin( page_height=page.size.height * scale ) ) cropped_im = page.image.crop(crop_bbox.as_tuple()) element.image = ImageRef.from_pil( cropped_im, dpi=int(72 * scale) ) # Aggregate confidence values for document: if len(conv_res.pages) > 0: with warnings.catch_warnings(): warnings.filterwarnings( "ignore", category=RuntimeWarning, message="Mean of empty slice|All-NaN slice encountered", ) conv_res.confidence.layout_score = float( np.nanmean( [c.layout_score for c in conv_res.confidence.pages.values()] ) ) conv_res.confidence.parse_score = float( np.nanquantile( [c.parse_score for c in conv_res.confidence.pages.values()], q=0.1, # parse score should relate to worst 10% of pages. ) ) conv_res.confidence.table_score = float( np.nanmean( [c.table_score for c in conv_res.confidence.pages.values()] ) ) conv_res.confidence.ocr_score = float( np.nanmean( [c.ocr_score for c in conv_res.confidence.pages.values()] ) ) # Add failed pages to DoclingDocument.pages to preserve page numbering # This ensures page break markers are generated for skipped/failed pages self._add_failed_pages_to_document( conv_res, expected_page_nos=self._get_expected_page_nos(conv_res) ) return conv_res def _add_failed_pages_to_document( self, conv_res: ConversionResult, expected_page_nos: list[int] ) -> None: """Add failed/skipped pages to DoclingDocument.pages. This ensures that page break markers are correctly generated for documents where some pages failed to parse. Without this, export functions would not know about the missing pages and would generate incorrect page break counts. The failed pages are added with their size information (if available from the backend) but without any content. """ if conv_res.document is None: return # Find pages that are missing from the document existing_page_nos = set(conv_res.document.pages.keys()) missing_page_nos = set(expected_page_nos) - existing_page_nos if not missing_page_nos: return for page_no in sorted(missing_page_nos): size = self._page_sizes_by_no.get(page_no, Size(width=0.0, height=0.0)) # Add the failed page to the document's pages dict conv_res.document.pages[page_no] = PageItem( page_no=page_no, size=size, image=None, ) _log.debug( "Added %d failed/skipped pages to document: %s", len(missing_page_nos), sorted(missing_page_nos), ) # ---------------------------------------------------------------- misc @classmethod def get_default_options(cls) -> ThreadedPdfPipelineOptions: return ThreadedPdfPipelineOptions() @classmethod def is_backend_supported(cls, backend: AbstractDocumentBackend) -> bool: return isinstance(backend, PdfDocumentBackend) def _determine_status(self, conv_res: ConversionResult) -> ConversionStatus: return conv_res.status def _unload(self, conv_res: ConversionResult) -> None: self._page_sizes_by_no = {} for p in conv_res.pages: if p._backend is not None: p._backend.unload() if conv_res.input._backend: conv_res.input._backend.unload()