nl-processing 1.0.2__tar.gz → 1.0.4__tar.gz

{nl_processing-1.0.2 → nl_processing-1.0.4}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: nl_processing
-Version: 1.0.2
+Version: 1.0.4
 Summary: Aggregate build for the nl_processing multi-package repository
 Requires-Python: >=3.12
 Description-Content-Type: text/markdown

{nl_processing-1.0.2 → nl_processing-1.0.4}/nl_processing.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: nl_processing
-Version: 1.0.2
+Version: 1.0.4
 Summary: Aggregate build for the nl_processing multi-package repository
 Requires-Python: >=3.12
 Description-Content-Type: text/markdown

{nl_processing-1.0.2 → nl_processing-1.0.4}/nl_processing.egg-info/SOURCES.txt

@@ -7,6 +7,7 @@ nl_processing.egg-info/requires.txt
 nl_processing.egg-info/top_level.txt
 packages/core/src/nl_processing/core/__init__.py
 packages/core/src/nl_processing/core/exceptions.py
+packages/core/src/nl_processing/core/image_encoding.py
 packages/core/src/nl_processing/core/models.py
 packages/core/src/nl_processing/core/ports.py
 packages/core/src/nl_processing/core/prompts.py
@@ -46,6 +47,14 @@ packages/translate_text/src/nl_processing/translate_text/__init__.py
 packages/translate_text/src/nl_processing/translate_text/service.py
 packages/translate_text/src/nl_processing/translate_text/prompts/generate_nl_ru_prompt.py
 packages/translate_text/src/nl_processing/translate_text/prompts/nl_ru.json
+packages/translate_text_from_image/src/nl_processing/translate_text_from_image/__init__.py
+packages/translate_text_from_image/src/nl_processing/translate_text_from_image/benchmark.py
+packages/translate_text_from_image/src/nl_processing/translate_text_from_image/service.py
+packages/translate_text_from_image/src/nl_processing/translate_text_from_image/prompts/__init__.py
+packages/translate_text_from_image/src/nl_processing/translate_text_from_image/prompts/generate_nl_ru_prompt.py
+packages/translate_text_from_image/src/nl_processing/translate_text_from_image/prompts/nl_ru.json
+packages/translate_text_from_image/src/nl_processing/translate_text_from_image/prompts/examples/dutch_handwritten_mixed.jpg
+packages/translate_text_from_image/src/nl_processing/translate_text_from_image/prompts/examples/dutch_vocabulary_wide.jpg
 packages/translate_word/src/nl_processing/translate_word/__init__.py
 packages/translate_word/src/nl_processing/translate_word/service.py
 packages/translate_word/src/nl_processing/translate_word/prompts/generate_nl_ru_prompt.py

{nl_processing-1.0.2/packages/extract_text_from_image/src/nl_processing/extract_text_from_image → nl_processing-1.0.4/packages/core/src/nl_processing/core}/image_encoding.py

@@ -2,9 +2,10 @@ import base64
 import pathlib
 
 import cv2
-from nl_processing.core.exceptions import UnsupportedImageFormatError
 import numpy
 
+from nl_processing.core.exceptions import UnsupportedImageFormatError
+
 SUPPORTED_EXTENSIONS = {".png", ".jpg", ".jpeg", ".gif", ".webp"}
 
 

{nl_processing-1.0.2 → nl_processing-1.0.4}/packages/core/src/nl_processing/core/models.py

@@ -28,6 +28,7 @@ class PartOfSpeech(Enum):
     INTERJECTION = "interjection"
     PROPER_NOUN_PERSON = "proper_noun_person"
     PROPER_NOUN_COUNTRY = "proper_noun_country"
+    PHRASE = "phrase"
 
 
 class ExtractedText(BaseModel):

{nl_processing-1.0.2 → nl_processing-1.0.4}/packages/core/src/nl_processing/core/prompts.py

@@ -18,6 +18,9 @@ def build_translation_chain(
     prompts_dir: pathlib.Path,
     tool_schema: type[BaseModel],
     model: str,
+    reasoning_effort: str | None = None,
+    service_tier: str | None = None,
+    temperature: float | None = 0,
 ) -> RunnableSerializable:  # type: ignore[type-arg]
     """Validate a language pair, load its prompt, and return a prompt|llm chain.
 
@@ -31,6 +34,9 @@ def build_translation_chain(
         prompts_dir: Directory containing ``<src>_<tgt>.json`` prompt files.
         tool_schema: Pydantic model class to bind as a tool.
         model: OpenAI model identifier string.
+        reasoning_effort: Optional reasoning effort level for the model.
+        service_tier: Optional service tier for the OpenAI API.
+        temperature: LLM temperature (default 0 for deterministic output).
 
     Returns:
         A ``prompt | llm`` RunnableSerializable ready for ``ainvoke()``.
@@ -50,7 +56,12 @@ def build_translation_chain(
     prompt_file = f"{source_language.value}_{target_language.value}.json"
     prompt = load_prompt(str(prompts_dir / prompt_file))
 
-    llm = ChatOpenAI(model=model, temperature=0).bind_tools(
+    llm = ChatOpenAI(
+        model=model,
+        temperature=temperature,
+        reasoning_effort=reasoning_effort,
+        service_tier=service_tier,
+    ).bind_tools(
         [tool_schema],
         tool_choice=tool_schema.__name__,
     )

nl_processing-1.0.4/packages/extract_text_from_image/src/nl_processing/extract_text_from_image/image_encoding.py

@@ -0,0 +1,15 @@
+from nl_processing.core.image_encoding import (
+    SUPPORTED_EXTENSIONS,
+    encode_cv2_to_base64,
+    encode_path_to_base64,
+    get_image_format,
+    validate_image_format,
+)
+
+__all__ = [
+    "SUPPORTED_EXTENSIONS",
+    "encode_cv2_to_base64",
+    "encode_path_to_base64",
+    "get_image_format",
+    "validate_image_format",
+]

{nl_processing-1.0.2 → nl_processing-1.0.4}/packages/extract_text_from_image/src/nl_processing/extract_text_from_image/prompts/generate_nl_prompt.py

@@ -18,9 +18,9 @@ import tempfile
 
 from langchain_core.messages import AIMessage, HumanMessage, SystemMessage, ToolMessage
 from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder
+from nl_processing.core.image_encoding import encode_path_to_base64
 
 from nl_processing.extract_text_from_image.benchmark import generate_test_image
-from nl_processing.extract_text_from_image.image_encoding import encode_path_to_base64
 
 SYSTEM_INSTRUCTION = (
     "Je bent een tekst-extractie assistent. "
@@ -28,6 +28,8 @@ SYSTEM_INSTRUCTION = (
     "Behoud de originele documentstructuur als markdown "
     "(koppen, nadruk, regelafbrekingen). "
     "Negeer tekst in andere talen. "
+    "Als er geen Nederlandse tekst zichtbaar is, retourneer dan een lege string. "
+    "Beschouw Engelse tekst nooit als Nederlands en kopieer geen niet-Nederlandse tekst. "
     "Retourneer alleen de geëxtraheerde tekst, zonder commentaar of uitleg."
 )
 
@@ -92,7 +94,7 @@ EXAMPLE_5_EXPECTED = ""
 EXAMPLE_6_TEXT = "Please take your shoes off before entering the house"
 EXAMPLE_6_EXPECTED = ""
 
-EXAMPLE_7_TEXT = "Remember to bring your umbrella tomorrow"
+EXAMPLE_7_TEXT = "Remember to charge your phone before leaving tomorrow"
 EXAMPLE_7_EXPECTED = ""
 
 OUTPUT_PATH = Path(__file__).parent / "nl.json"