docling-core 2.31.0__tar.gz → 2.31.2__tar.gz

{docling_core-2.31.0 → docling_core-2.31.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: docling-core
-Version: 2.31.0
+Version: 2.31.2
 Summary: A python library to define and validate data types in Docling.
 Home-page: https://github.com/docling-project
 License: MIT

{docling_core-2.31.0 → docling_core-2.31.2}/docling_core/transforms/chunker/hybrid_chunker.py

@@ -9,6 +9,7 @@ from functools import cached_property
 from typing import Any, Iterable, Iterator, Optional, Union
 
 from pydantic import BaseModel, ConfigDict, Field, computed_field, model_validator
+from transformers import PreTrainedTokenizerBase
 
 from docling_core.transforms.chunker.hierarchical_chunker import (
     ChunkingSerializerProvider,
@@ -70,23 +71,37 @@ class HybridChunker(BaseChunker):
     @model_validator(mode="before")
     @classmethod
     def _patch(cls, data: Any) -> Any:
-        if isinstance(data, dict) and (tokenizer := data.get("tokenizer")):
+        if isinstance(data, dict):
+            tokenizer = data.get("tokenizer")
             max_tokens = data.get("max_tokens")
-            if isinstance(tokenizer, BaseTokenizer):
-                pass
-            else:
+            if not isinstance(tokenizer, BaseTokenizer) and (
+                # some legacy param passed:
+                tokenizer is not None
+                or max_tokens is not None
+            ):
                 from docling_core.transforms.chunker.tokenizer.huggingface import (
                     HuggingFaceTokenizer,
                 )
 
+                warnings.warn(
+                    "Deprecated initialization parameter types for HybridChunker. "
+                    "For updated usage check out "
+                    "https://docling-project.github.io/docling/examples/hybrid_chunking/",
+                    DeprecationWarning,
+                    stacklevel=3,
+                )
+
                 if isinstance(tokenizer, str):
                     data["tokenizer"] = HuggingFaceTokenizer.from_pretrained(
                         model_name=tokenizer,
                         max_tokens=max_tokens,
                     )
-                else:
-                    # migrate previous HF-based tokenizers
-                    kwargs = {"tokenizer": tokenizer}
+                elif tokenizer is None or isinstance(
+                    tokenizer, PreTrainedTokenizerBase
+                ):
+                    kwargs = {
+                        "tokenizer": tokenizer or _get_default_tokenizer().tokenizer
+                    }
                     if max_tokens is not None:
                         kwargs["max_tokens"] = max_tokens
                     data["tokenizer"] = HuggingFaceTokenizer(**kwargs)

{docling_core-2.31.0 → docling_core-2.31.2}/docling_core/transforms/serializer/markdown.py

@@ -535,7 +535,7 @@ class MarkdownDocSerializer(DocSerializer):
     ) -> SerializationResult:
         """Serialize a document out of its parts."""
         text_res = "\n\n".join([p.text for p in parts if p.text])
-        if self.params.page_break_placeholder:
+        if self.requires_page_break():
             page_sep = self.params.page_break_placeholder or ""
             for full_match, _, _ in self._get_page_breaks(text=text_res):
                 text_res = text_res.replace(full_match, page_sep)
@@ -543,6 +543,6 @@ class MarkdownDocSerializer(DocSerializer):
         return create_ser_result(text=text_res, span_source=parts)
 
     @override
-    def requires_page_break(self):
+    def requires_page_break(self) -> bool:
         """Whether to add page breaks."""
         return self.params.page_break_placeholder is not None

{docling_core-2.31.0 → docling_core-2.31.2}/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "docling-core"
-version = "2.31.0"
+version = "2.31.2"
 description = "A python library to define and validate data types in Docling."
 license = "MIT"
 authors = [