sparrow-parse 0.5.5__py3-none-any.whl → 1.0.2__py3-none-any.whl

sparrow_parse/__init__.py

@@ -1 +1 @@
-__version__ = '0.5.5'
+__version__ = '1.0.2'

sparrow_parse/extractors/vllm_extractor.py

@@ -147,6 +147,14 @@ 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):
@@ -198,7 +206,7 @@ if __name__ == "__main__":
     # # export HF_TOKEN="hf_"
     # config = {
     #     "method": "mlx",  # Could be 'huggingface', 'mlx' or 'local_gpu'
-    #     "model_name": "mlx-community/Qwen2.5-VL-7B-Instruct-8bit",
+    #     "model_name": "mlx-community/Mistral-Small-3.1-24B-Instruct-2503-8bit",
     #     # "hf_space": "katanaml/sparrow-qwen2-vl-7b",
     #     # "hf_token": os.getenv('HF_TOKEN'),
     #     # Additional fields for local GPU inference
@@ -211,8 +219,8 @@ if __name__ == "__main__":
     #
     # input_data = [
     #     {
-    #         "file_path": "/Users/andrejb/Work/katana-git/sparrow/sparrow-ml/llm/data/bonds_table.png",
-    #         "text_input": "retrieve document data. return response in JSON format"
+    #         "file_path": "sparrow_parse/images/bonds_table.png",
+    #         "text_input": "retrieve all data. return response in JSON format"
     #     }
     # ]
     #
@@ -226,4 +234,5 @@ if __name__ == "__main__":
     #
     # for i, result in enumerate(results_array):
     #     print(f"Result for page {i + 1}:", result)
-    # print(f"Number of pages: {num_pages}")
+    # print(f"Number of pages: {num_pages}")
+

sparrow_parse/text_extraction.py

@@ -1,30 +1,35 @@
 from mlx_vlm import load, apply_chat_template, generate
 from mlx_vlm.utils import load_image
 
-# For test purposes, we will use a sample image
 
 # Load model and processor
-qwen_vl_model, qwen_vl_processor = load("mlx-community/Qwen2.5-VL-7B-Instruct-8bit")
-qwen_vl_config = qwen_vl_model.config
+# vl_model, vl_processor = load("mlx-community/Mistral-Small-3.1-24B-Instruct-2503-8bit")
+vl_model, vl_processor = load("mlx-community/Qwen2.5-VL-7B-Instruct-8bit")
+vl_config = vl_model.config
 
-image = load_image("images/graph.png")
+image = load_image("images/bonds_table.png")
 
 messages = [
     {"role": "system", "content": "You are an expert at extracting text from images. Format your response in json."},
-    {"role": "user", "content": "Extract the names, labels and y coordinates from the image."}
+    {"role": "user", "content": "retrieve [{\"instrument_name\":\"str\", \"valuation\":\"int\"}]. return response in JSON format"}
 ]
 
+# message = "retrieve all data. return response in JSON format"
+# message = "retrieve [{\"instrument_name\":\"str\", \"valuation\":\"int\"}]. return response in JSON format"
+
 # Apply chat template
-prompt = apply_chat_template(qwen_vl_processor, qwen_vl_config, messages)
+prompt = apply_chat_template(vl_processor, vl_config, messages)
+# prompt = apply_chat_template(vl_processor, vl_config, message)
 
 # Generate text
-qwen_vl_output = generate(
-    qwen_vl_model,
-    qwen_vl_processor,
+vl_output = generate(
+    vl_model,
+    vl_processor,
     prompt,
     image,
     max_tokens=1000,
-    temperature=0.7,
+    temperature=0,
+    verbose=False
 )
 
-print(qwen_vl_output)
+print(vl_output)

sparrow_parse/vllm/mlx_inference.py

@@ -39,21 +39,40 @@ class MLXInference(ModelInference):
     def process_response(self, output_text):
         """
         Process and clean the model's raw output to format as JSON.
-
-        :param output_text: Raw output text from the model.
-        :return: A formatted JSON string or the original text in case of errors.
         """
         try:
-            cleaned_text = (
-                output_text.strip("[]'")
-                .replace("```json\n", "")
-                .replace("\n```", "")
-                .replace("'", "")
-            )
-            formatted_json = json.loads(cleaned_text)
+            # Check if we have markdown code block markers
+            if "```" in output_text:
+                # Handle markdown-formatted output
+                json_start = output_text.find("```json")
+                if json_start != -1:
+                    # Extract content between ```json and ```
+                    content = output_text[json_start + 7:]
+                    json_end = content.rfind("```")
+                    if json_end != -1:
+                        content = content[:json_end].strip()
+                        formatted_json = json.loads(content)
+                        return json.dumps(formatted_json, indent=2)
+
+            # Handle raw JSON (no markdown formatting)
+            # First try to find JSON array or object patterns
+            for pattern in [r'\[\s*\{.*\}\s*\]', r'\{.*\}']:
+                import re
+                matches = re.search(pattern, output_text, re.DOTALL)
+                if matches:
+                    potential_json = matches.group(0)
+                    try:
+                        formatted_json = json.loads(potential_json)
+                        return json.dumps(formatted_json, indent=2)
+                    except:
+                        pass
+
+            # Last resort: try to parse the whole text as JSON
+            formatted_json = json.loads(output_text.strip())
             return json.dumps(formatted_json, indent=2)
-        except json.JSONDecodeError as e:
-            print(f"Failed to parse JSON in MLX inference backend: {e}")
+
+        except Exception as e:
+            print(f"Failed to parse JSON: {e}")
             return output_text
 
 
@@ -102,10 +121,13 @@ class MLXInference(ModelInference):
             image, width, height = self.load_image_data(file_path)
 
             # Prepare messages for the chat model
-            messages = [
-                {"role": "system", "content": "You are an expert at extracting structured text from image documents."},
-                {"role": "user", "content": input_data[0]["text_input"]},
-            ]
+            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"]},
+                ]
 
             # Generate and process response
             prompt = apply_chat_template(processor, config, messages)  # Assuming defined

{sparrow_parse-0.5.5.dist-info → sparrow_parse-1.0.2.dist-info}/METADATA

@@ -1,15 +1,13 @@
 Metadata-Version: 2.1
 Name: sparrow-parse
-Version: 0.5.5
+Version: 1.0.2
 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
 Author-email: andrejus.baranovskis@gmail.com
-License: UNKNOWN
 Project-URL: Homepage, https://github.com/katanaml/sparrow/tree/main/sparrow-data/parse
 Project-URL: Repository, https://github.com/katanaml/sparrow
 Keywords: llm,vllm,ocr,vision
-Platform: UNKNOWN
 Classifier: Operating System :: OS Independent
 Classifier: License :: OSI Approved :: GNU General Public License v3 (GPLv3)
 Classifier: Topic :: Software Development
@@ -17,16 +15,16 @@ Classifier: Programming Language :: Python :: 3.10
 Requires-Python: >=3.10
 Description-Content-Type: text/markdown
 Requires-Dist: rich
-Requires-Dist: transformers>=4.49.0
-Requires-Dist: torchvision>=0.21.0
-Requires-Dist: torch>=2.6.0
-Requires-Dist: sentence-transformers>=3.3.1
-Requires-Dist: numpy>=2.1.3
-Requires-Dist: pypdf>=5.2.0
-Requires-Dist: gradio-client>=1.7.2
-Requires-Dist: pdf2image>=1.17.0
-Requires-Dist: mlx>=0.23.1; sys_platform == "darwin" and platform_machine == "arm64"
-Requires-Dist: mlx-vlm==0.1.14; sys_platform == "darwin" and platform_machine == "arm64"
+Requires-Dist: transformers >=4.50.1
+Requires-Dist: torchvision >=0.21.0
+Requires-Dist: torch >=2.6.0
+Requires-Dist: sentence-transformers >=4.0.0
+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"
 
 # Sparrow Parse
 
@@ -187,5 +185,3 @@ If your organization is seeking to utilize Sparrow under a proprietary license,
 ## License
 
 Licensed under the GPL 3.0. Copyright 2020-2025 Katana ML, Andrej Baranovskij. [Copy of the license](https://github.com/katanaml/sparrow/blob/main/LICENSE).
-
-

{sparrow_parse-0.5.5.dist-info → sparrow_parse-1.0.2.dist-info}/RECORD

@@ -1,8 +1,8 @@
-sparrow_parse/__init__.py,sha256=nrKeY2xA6SXRPdgHDxi2HLkFNpXRuW6MkqwC0reZpy8,21
+sparrow_parse/__init__.py,sha256=C8nyPP5-54GgYCcP38Lbel_pRimOW-Ra4bw6Vzp2lmE,21
 sparrow_parse/__main__.py,sha256=Xs1bpJV0n08KWOoQE34FBYn6EBXZA9HIYJKrE4ZdG78,153
-sparrow_parse/text_extraction.py,sha256=JtUU7swvV12xBai5S9ICxWWWrUlkpZTZqvUnbz1h5Mk,834
+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=Cf2sVgxDExj2ud4G6z9JnirVclTgPIEe9YSoCfTkW4k,9563
+sparrow_parse/extractors/vllm_extractor.py,sha256=uRSXzCQzjXujg1n1ozDitSPQoCfO435Nog7yO1IxWiU,9874
 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=MUuW56f-aKnVmeMAATxKLxsovEMmp1qlgtlmW8J2C7M,4899
-sparrow_parse-0.5.5.dist-info/METADATA,sha256=l9s3Vi-5KVtryw0a1Z7AXqQHIKuUMbPQu0XNeilRofU,7254
-sparrow_parse-0.5.5.dist-info/WHEEL,sha256=tZoeGjtWxWRfdplE7E3d45VPlLNQnvbKiYnx7gwAy8A,92
-sparrow_parse-0.5.5.dist-info/entry_points.txt,sha256=8CrvTVTTcz1YuZ8aRCYNOH15ZOAaYLlcbYX3t28HwJY,54
-sparrow_parse-0.5.5.dist-info/top_level.txt,sha256=n6b-WtT91zKLyCPZTP7wvne8v_yvIahcsz-4sX8I0rY,14
-sparrow_parse-0.5.5.dist-info/RECORD,,
+sparrow_parse/vllm/mlx_inference.py,sha256=KjAftUIAWxYfctE3n1BKXA8jETM4WT3ESyx97eMA_8U,5954
+sparrow_parse-1.0.2.dist-info/METADATA,sha256=K4XNgj-PpegO8aLAe32aOZ3D8kh6lnMX0po2wXTxn-w,7229
+sparrow_parse-1.0.2.dist-info/WHEEL,sha256=yQN5g4mg4AybRjkgi-9yy4iQEFibGQmlz78Pik5Or-A,92
+sparrow_parse-1.0.2.dist-info/entry_points.txt,sha256=HV5nnQVtr2m-kn6hzY_ynp0zugNCcGovbmnfmQgOyhw,53
+sparrow_parse-1.0.2.dist-info/top_level.txt,sha256=n6b-WtT91zKLyCPZTP7wvne8v_yvIahcsz-4sX8I0rY,14
+sparrow_parse-1.0.2.dist-info/RECORD,,

{sparrow_parse-0.5.5.dist-info → sparrow_parse-1.0.2.dist-info}/WHEEL

@@ -1,5 +1,5 @@
 Wheel-Version: 1.0
-Generator: bdist_wheel (0.45.1)
+Generator: bdist_wheel (0.41.2)
 Root-Is-Purelib: true
 Tag: py3-none-any
 

{sparrow_parse-0.5.5.dist-info → sparrow_parse-1.0.2.dist-info}/entry_points.txt

@@ -1,3 +1,2 @@
 [console_scripts]
 sparrow-parse = sparrow_parse:main
-