sparrow-parse 1.0.1__py3-none-any.whl → 1.0.3__py3-none-any.whl

sparrow_parse/__init__.py

@@ -1 +1 @@
-__version__ = '1.0.1'
+__version__ = '1.0.3'

sparrow_parse/extractors/vllm_extractor.py

@@ -25,11 +25,21 @@ class VLLMExtractor(object):
         if debug:
             print("Input data:", input_data)
 
+        # Handle both missing file_path and file_path=None as text-only inference
+        is_text_only = "file_path" not in input_data[0] or input_data[0]["file_path"] is None
+
+        if is_text_only:
+            # Ensure file_path exists and is None for consistency
+            input_data[0]["file_path"] = None
+            results = model_inference_instance.inference(input_data)
+            return results, 0
+
+        # Document data extraction inference (file_path exists and is not None)
         file_path = input_data[0]["file_path"]
         if self.is_pdf(file_path):
             return self._process_pdf(model_inference_instance, input_data, tables_only, crop_size, debug, debug_dir, mode)
-
-        return self._process_non_pdf(model_inference_instance, input_data, tables_only, crop_size, debug, debug_dir)
+        else:
+            return self._process_non_pdf(model_inference_instance, input_data, tables_only, crop_size, debug, debug_dir)
 
 
     def _process_pdf(self, model_inference_instance, input_data, tables_only, crop_size, debug, debug_dir, mode):
@@ -147,6 +157,14 @@ class VLLMExtractor(object):
         table_detector = TableDetector()
         cropped_tables = table_detector.detect_tables(file_path, local=False, debug_dir=debug_dir, debug=debug)
         results_array = []
+
+        # Check if no tables were found
+        if cropped_tables is None:
+            if debug:
+                print(f"No tables detected in {file_path}")
+            # Return a structured no-tables-found response instead of failing
+            return [json.dumps({"message": "No tables detected in the document", "status": "empty"})]
+
         temp_dir = tempfile.mkdtemp()
 
         for i, table in enumerate(cropped_tables):
@@ -216,6 +234,13 @@ if __name__ == "__main__":
     #     }
     # ]
     #
+    # # input_data = [
+    # #     {
+    # #         "file_path": None,
+    # #         "text_input": "why earth is spinning around the sun?"
+    # #     }
+    # # ]
+    #
     # # Now you can run inference without knowing which implementation is used
     # results_array, num_pages = extractor.run_inference(model_inference_instance, input_data, tables_only=False,
     #                                                    generic_query=False,

sparrow_parse/vllm/mlx_inference.py

@@ -106,31 +106,70 @@ class MLXInference(ModelInference):
         :param mode: Optional mode for inference ("static" for simple JSON output).
         :return: List of processed model responses.
         """
+        # Handle static mode
         if mode == "static":
             return [self.get_simple_json()]
 
         # Load the model and processor
         model, processor = self._load_model_and_processor(self.model_name)
         config = model.config
+        
+        # Determine if we're doing text-only or image-based inference
+        is_text_only = input_data[0].get("file_path") is None
+        
+        if is_text_only:
+            # Text-only inference
+            messages = input_data[0]["text_input"]
+            response = self._generate_text_response(model, processor, config, messages)
+            results = [self.process_response(response)]
+            print("Agent inference completed successfully")
+        else:
+            # Image-based inference
+            file_paths = self._extract_file_paths(input_data)
+            results = self._process_images(model, processor, config, file_paths, input_data)
+        
+        return results
 
-        # Prepare absolute file paths
-        file_paths = self._extract_file_paths(input_data)
-
+    def _generate_text_response(self, model, processor, config, messages):
+        """
+        Generate a text response for text-only inputs.
+        
+        :param model: The loaded model
+        :param processor: The loaded processor
+        :param config: Model configuration
+        :param messages: Input messages
+        :return: Generated response
+        """
+        prompt = apply_chat_template(processor, config, messages)
+        return generate(
+            model,
+            processor,
+            prompt,
+            max_tokens=4000,
+            temperature=0.0,
+            verbose=False
+        )
+
+    def _process_images(self, model, processor, config, file_paths, input_data):
+        """
+        Process images and generate responses for each.
+        
+        :param model: The loaded model
+        :param processor: The loaded processor
+        :param config: Model configuration
+        :param file_paths: List of image file paths
+        :param input_data: Original input data
+        :return: List of processed responses
+        """
         results = []
         for file_path in file_paths:
             image, width, height = self.load_image_data(file_path)
-
-            # Prepare messages for the chat model
-            if "mistral" in self.model_name.lower():
-                messages = input_data[0]["text_input"]
-            else:
-                messages = [
-                    {"role": "system", "content": "You are an expert at extracting structured text from image documents."},
-                    {"role": "user", "content": input_data[0]["text_input"]},
-                ]
-
+            
+            # Prepare messages based on model type
+            messages = self._prepare_messages(input_data, file_path)
+            
             # Generate and process response
-            prompt = apply_chat_template(processor, config, messages)  # Assuming defined
+            prompt = apply_chat_template(processor, config, messages)
             response = generate(
                 model,
                 processor,
@@ -142,11 +181,26 @@ class MLXInference(ModelInference):
                 verbose=False
             )
             results.append(self.process_response(response))
-
-            print("Inference completed successfully for: ", file_path)
-
+            print(f"Inference completed successfully for: {file_path}")
+        
         return results
 
+    def _prepare_messages(self, input_data, file_path):
+        """
+        Prepare the appropriate messages based on the model type.
+        
+        :param input_data: Original input data
+        :param file_path: Current file path being processed
+        :return: Properly formatted messages
+        """
+        if "mistral" in self.model_name.lower():
+            return input_data[0]["text_input"]
+        else:
+            return [
+                {"role": "system", "content": "You are an expert at extracting structured text from image documents."},
+                {"role": "user", "content": input_data[0]["text_input"]},
+            ]
+
     @staticmethod
     def _extract_file_paths(input_data):
         """

{sparrow_parse-1.0.1.dist-info → sparrow_parse-1.0.3.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: sparrow-parse
-Version: 1.0.1
+Version: 1.0.3
 Summary: Sparrow Parse is a Python package (part of Sparrow) for parsing and extracting information from documents.
 Home-page: https://github.com/katanaml/sparrow/tree/main/sparrow-data/parse
 Author: Andrej Baranovskij
@@ -15,16 +15,16 @@ Classifier: Programming Language :: Python :: 3.10
 Requires-Python: >=3.10
 Description-Content-Type: text/markdown
 Requires-Dist: rich
-Requires-Dist: transformers >=4.50.1
+Requires-Dist: transformers >=4.51.3
 Requires-Dist: torchvision >=0.21.0
 Requires-Dist: torch >=2.6.0
-Requires-Dist: sentence-transformers >=4.0.0
+Requires-Dist: sentence-transformers >=4.1.0
 Requires-Dist: numpy >=2.2.4
 Requires-Dist: pypdf >=5.4.0
 Requires-Dist: gradio-client >=1.7.2
 Requires-Dist: pdf2image >=1.17.0
-Requires-Dist: mlx >=0.24.1 ; sys_platform == "darwin" and platform_machine == "arm64"
-Requires-Dist: mlx-vlm ==0.1.21 ; sys_platform == "darwin" and platform_machine == "arm64"
+Requires-Dist: mlx >=0.25.0 ; sys_platform == "darwin" and platform_machine == "arm64"
+Requires-Dist: mlx-vlm ==0.1.23 ; sys_platform == "darwin" and platform_machine == "arm64"
 
 # Sparrow Parse
 

{sparrow_parse-1.0.1.dist-info → sparrow_parse-1.0.3.dist-info}/RECORD

@@ -1,8 +1,8 @@
-sparrow_parse/__init__.py,sha256=bMIenWosteoeUs51RbaWVZetIuzRhWymuyy-n0rfK0I,21
+sparrow_parse/__init__.py,sha256=MpVHFFoITiYyPltTb_qFrdeX2entdTm4x0PczXi3txY,21
 sparrow_parse/__main__.py,sha256=Xs1bpJV0n08KWOoQE34FBYn6EBXZA9HIYJKrE4ZdG78,153
 sparrow_parse/text_extraction.py,sha256=lirPpvz8tnwCMGmoHPK94-vCviybuRyQM-mpvhtp3uY,1124
 sparrow_parse/extractors/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-sparrow_parse/extractors/vllm_extractor.py,sha256=MYaT8ITaxEDH6TpSwMZl6WbOiLYqyFN_VQsCmVe4YkI,9535
+sparrow_parse/extractors/vllm_extractor.py,sha256=ZxYiSrdKWLcBXn4LUuvEcDH0q_Ua8xTzqmEF15puP08,10557
 sparrow_parse/helpers/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 sparrow_parse/helpers/image_optimizer.py,sha256=gUAJuNzRAB5ipgfhxTNss4MHbCPPkV5y-BSyrEHcJ0Y,2164
 sparrow_parse/helpers/pdf_optimizer.py,sha256=A2BVkb2JMqTJUz6bdfVzMmFSYaWn1QMav7UadMi0XJg,3423
@@ -13,9 +13,9 @@ sparrow_parse/vllm/huggingface_inference.py,sha256=EJnG6PesGKMc_0qGPN8ufE6pSnhAg
 sparrow_parse/vllm/inference_base.py,sha256=4mwGoAY63MB4cHZpV0czTkJWEzimmiTzqqzKmLNzgjw,820
 sparrow_parse/vllm/inference_factory.py,sha256=FTM65O-dW2WZchHOrNN7_Q3-FlVoAc65iSptuuUuClM,1166
 sparrow_parse/vllm/local_gpu_inference.py,sha256=aHoJTejb5xrXjWDIGu5RBQWEyRCOBCB04sMvO2Wyvg8,628
-sparrow_parse/vllm/mlx_inference.py,sha256=KjAftUIAWxYfctE3n1BKXA8jETM4WT3ESyx97eMA_8U,5954
-sparrow_parse-1.0.1.dist-info/METADATA,sha256=IADhZaH5VgE530Czablge_kn3mEAtlq8Zq1TubTkMv8,7229
-sparrow_parse-1.0.1.dist-info/WHEEL,sha256=yQN5g4mg4AybRjkgi-9yy4iQEFibGQmlz78Pik5Or-A,92
-sparrow_parse-1.0.1.dist-info/entry_points.txt,sha256=HV5nnQVtr2m-kn6hzY_ynp0zugNCcGovbmnfmQgOyhw,53
-sparrow_parse-1.0.1.dist-info/top_level.txt,sha256=n6b-WtT91zKLyCPZTP7wvne8v_yvIahcsz-4sX8I0rY,14
-sparrow_parse-1.0.1.dist-info/RECORD,,
+sparrow_parse/vllm/mlx_inference.py,sha256=vqIkfTd5rP8bnZ8K_CGVEWe_G3E4i3rwN9MfLBDiE3c,8000
+sparrow_parse-1.0.3.dist-info/METADATA,sha256=dIGBhBhtR5rSKj4RbT1PhyrWxKUVUq5AxbJ33FsKNlE,7229
+sparrow_parse-1.0.3.dist-info/WHEEL,sha256=yQN5g4mg4AybRjkgi-9yy4iQEFibGQmlz78Pik5Or-A,92
+sparrow_parse-1.0.3.dist-info/entry_points.txt,sha256=HV5nnQVtr2m-kn6hzY_ynp0zugNCcGovbmnfmQgOyhw,53
+sparrow_parse-1.0.3.dist-info/top_level.txt,sha256=n6b-WtT91zKLyCPZTP7wvne8v_yvIahcsz-4sX8I0rY,14
+sparrow_parse-1.0.3.dist-info/RECORD,,