vlmparse 0.1.4__tar.gz → 0.1.5__tar.gz

{vlmparse-0.1.4/vlmparse.egg-info → vlmparse-0.1.5}/PKG-INFO

@@ -1,7 +1,7 @@
 Metadata-Version: 2.4
 Name: vlmparse
-Version: 0.1.4
-Requires-Python: >=3.12.0
+Version: 0.1.5
+Requires-Python: >=3.11.0
 Description-Content-Type: text/markdown
 License-File: LICENSE
 Requires-Dist: devtools>=0.12.2
@@ -72,6 +72,19 @@ Supported Converters:
 
 ## Installation
 
+Simplest solution with only the cli: 
+
+```bash
+uv tool install vlmparse
+```
+
+If you want to run the granite-docling model or use the streamlit viewing app:
+
+```bash
+uv tool install vlmparse[docling_core,st_app]
+```
+
+If you prefer cloning the repository and using the local version:
 ```bash
 uv sync
 ```
@@ -86,10 +99,11 @@ Activate the virtual environment:
 ```bash
 source .venv/bin/activate
 ```
-Other solution: append uv run to all the commands below.
 
 ## CLI Usage
 
+Note that you can bypass the previous installation step and just add uvx before each of the commands below.
+
 ### Convert PDFs
 
 With a general VLM (requires setting your api key as an environment variable):

{vlmparse-0.1.4 → vlmparse-0.1.5}/README.md

@@ -18,6 +18,19 @@ Supported Converters:
 
 ## Installation
 
+Simplest solution with only the cli: 
+
+```bash
+uv tool install vlmparse
+```
+
+If you want to run the granite-docling model or use the streamlit viewing app:
+
+```bash
+uv tool install vlmparse[docling_core,st_app]
+```
+
+If you prefer cloning the repository and using the local version:
 ```bash
 uv sync
 ```
@@ -32,10 +45,11 @@ Activate the virtual environment:
 ```bash
 source .venv/bin/activate
 ```
-Other solution: append uv run to all the commands below.
 
 ## CLI Usage
 
+Note that you can bypass the previous installation step and just add uvx before each of the commands below.
+
 ### Convert PDFs
 
 With a general VLM (requires setting your api key as an environment variable):

{vlmparse-0.1.4 → vlmparse-0.1.5}/pyproject.toml

@@ -2,17 +2,13 @@
 requires = ["setuptools", "wheel"]
 build-backend = "setuptools.build_meta"
 
-[metadata]
-name = "vlmparse"
-version = "0.1.0"
-
 [project]
 name = "vlmparse"
-version = "0.1.4"
+version = "0.1.5"
 authors = []
 description = ""
 readme = "README.md"
-requires-python = ">=3.12.0"
+requires-python = ">=3.11.0"
 dependencies = [
     "devtools>=0.12.2",
     "docker>=7.1.0",

{vlmparse-0.1.4 → vlmparse-0.1.5}/tests/test_end2end.py

@@ -64,3 +64,48 @@ def test_convert_with_docker(file_path, model):
     assert doc.pages[1].text is not None
 
     # Server will be automatically stopped due to auto_stop=True
+    server.stop()
+
+
+# @pytest.mark.skipif(
+#     "RUN_DEPLOYMENT_VLLM" not in os.environ
+#     or os.environ["RUN_DEPLOYMENT_VLLM"] == "false"
+#     or "GPU_TEST_VLMPARSE" not in os.environ,
+#     reason="Skipping because RUN_DEPLOYMENT_VLLM is not set or is false or GPU_TEST is not set",
+# )
+# @pytest.mark.parametrize(
+#     "model",
+#     [
+#         "docling",
+#         "lightonocr",
+#         "dotsocr",
+#         "nanonets/Nanonets-OCR2-3B",
+#         "hunyuanocr",
+#         "olmocr-2-fp8",
+#         "paddleocrvl",
+#         "mineru25",
+#         "chandra",
+#         "deepseekocr",
+#         "granite-docling",
+#     ],
+# )
+# def test_converter_with_server_with_docker(file_path, model):
+#     """Test conversion with automatic Docker deployment (requires GPU due to vllm limitations)."""
+
+#     from vlmparse.converter_with_server import ConverterWithServer
+#     converter_with_server = ConverterWithServer(
+#         model=model,
+#         uri=None,
+#         gpus=os.environ["GPU_TEST_VLMPARSE"],
+#         with_vllm_server=True,
+#         concurrency=10,
+#     )
+
+#     docs = converter_with_server.parse([file_path])
+
+#     # Assertions
+#     assert len(docs) == 1
+#     doc = docs[0]
+#     assert len(doc.pages) == 2
+#     assert doc.pages[0].text is not None
+#     assert doc.pages[1].text is not None

{vlmparse-0.1.4 → vlmparse-0.1.5}/vlmparse/clients/docling.py

@@ -34,7 +34,7 @@ class DoclingDockerServerConfig(DockerServerConfig):
             "LOG_LEVEL": "DEBUG",  # Enable verbose logging
             # Performance Tuning
             # "UVICORN_WORKERS": "4",  # Increase web server workers (Default: 1)
-            # "DOCLING_SERVE_ENG_LOC_NUM_WORKERS": "4",  # Increase processing workers (Default: 2)
+            "DOCLING_SERVE_ENG_LOC_NUM_WORKERS": "16",  # Increase processing workers (Default: 2)
             "DOCLING_NUM_THREADS": "32",  # Increase torch threads (Default: 4)
         }
     )
@@ -62,8 +62,8 @@ class DoclingDockerServerConfig(DockerServerConfig):
 class DoclingConverterConfig(ConverterConfig):
     """Configuration for Docling converter client."""
 
+    base_url: str
     model_name: str = "docling"
-    base_url: str = "http://localhost:5001"
     timeout: int = 300
     api_kwargs: dict = {"output_format": "markdown", "image_export_mode": "referenced"}
 

{vlmparse-0.1.4 → vlmparse-0.1.5}/vlmparse/clients/dotsocr.py

@@ -8,6 +8,7 @@ from PIL import Image
 from pydantic import Field
 
 from vlmparse.clients.openai_converter import (
+    LLMParams,
     OpenAIConverterClient,
     OpenAIConverterConfig,
 )
@@ -28,6 +29,7 @@ class DotsOCRDockerServerConfig(DockerServerConfig):
     dockerfile_dir: str = str(DOCKERFILE_DIR / "dotsocr")
     command_args: list[str] = Field(
         default_factory=lambda: [
+            "/workspace/weights/DotsOCR",
             "--tensor-parallel-size",
             "1",
             "--gpu-memory-utilization",
@@ -44,12 +46,19 @@ class DotsOCRDockerServerConfig(DockerServerConfig):
             # "16384",
         ]
     )
-    add_model_key_to_server: bool = False
+    add_model_key_to_server: bool = True
     aliases: list[str] = Field(default_factory=lambda: ["dotsocr"])
 
     @property
     def client_config(self):
-        return DotsOCRConverterConfig(llm_params=self.llm_params)
+        return DotsOCRConverterConfig(
+            llm_params=LLMParams(
+                base_url=f"http://localhost:{self.docker_port}{self.get_base_url_suffix()}",
+            )
+        )
+
+    def get_base_url_suffix(self) -> str:
+        return "/v1"
 
 
 class DotsOCRConverterConfig(OpenAIConverterConfig):

{vlmparse-0.1.4 → vlmparse-0.1.5}/vlmparse/clients/mineru.py

@@ -1,6 +1,5 @@
 import asyncio
 import io
-import os
 
 import orjson
 from loguru import logger
@@ -20,18 +19,21 @@ class MinerUDockerServerConfig(DockerServerConfig):
     docker_image: str = "pulsia/mineru25apipulsia:latest"
     docker_port: int = 4299
     container_port: int = 8000
+    server_ready_indicators: list[str] = Field(
+        default_factory=lambda: ["Uvicorn running"]
+    )
 
     @property
     def client_config(self):
-        return MinerUConverterConfig(api_url=f"http://localhost:{self.docker_port}")
+        return MinerUConverterConfig(base_url=f"http://localhost:{self.docker_port}")
 
 
 class MinerUConverterConfig(ConverterConfig):
     """Configuration for MinerU API converter."""
 
-    base_url: str = Field(
-        default_factory=lambda: os.getenv("MINERU_API_URL", "http://localhost:4299")
-    )
+    base_url: str
+    model_name: str = "opendatalab/MinerU2.5-2509-1.2B"
+    aliases: list[str] = Field(default_factory=lambda: ["mineru25"])
     timeout: int = 600
 
     def get_client(self, **kwargs) -> "MinerUConverter":
@@ -54,13 +56,12 @@ class MinerUConverter(BaseConverter):
         super().__init__(config=config, **kwargs)
         from httpx import AsyncClient
 
-        self.client = AsyncClient(base_url=config.api_url, timeout=config.timeout)
+        self.client = AsyncClient(base_url=config.base_url, timeout=config.timeout)
 
     async def _async_inference_with_api(self, image) -> list:
         """Run async inference with MinerU API."""
 
         img_byte_arr = await asyncio.to_thread(to_bytes_io, image)
-
         response = await self.client.post(
             "process-image",
             files={"image": ("image.png", img_byte_arr, "image/png")},

{vlmparse-0.1.4 → vlmparse-0.1.5}/vlmparse/clients/openai_converter.py

@@ -92,6 +92,7 @@ class OpenAIConverterClient(BaseConverter):
             base_url=self.config.llm_params.base_url,
             api_key=self.config.llm_params.api_key,
             timeout=self.config.llm_params.timeout,
+            max_retries=self.config.llm_params.max_retries,
         )
 
     async def _get_chat_completion(

{vlmparse-0.1.4 → vlmparse-0.1.5}/vlmparse/converter_with_server.py

@@ -42,13 +42,13 @@ class ConverterWithServer:
             docker_config = docker_config_registry.get(
                 self.model, default=self.with_vllm_server
             )
-            if self.port is not None:
-                docker_config.docker_port = self.port
 
             if docker_config is not None:
+                if self.port is not None:
+                    docker_config.docker_port = self.port
                 docker_config.gpu_device_ids = gpu_device_ids
-                server = docker_config.get_server(auto_stop=True)
-                server.start()
+                self.server = docker_config.get_server(auto_stop=True)
+                self.server.start()
 
                 self.client = docker_config.get_client()
             else:
@@ -56,6 +56,7 @@ class ConverterWithServer:
 
         else:
             client_config = converter_config_registry.get(self.model, uri=self.uri)
+
             self.client = client_config.get_client()
 
     def parse(

{vlmparse-0.1.4 → vlmparse-0.1.5}/vlmparse/registries.py

@@ -108,6 +108,7 @@ for gemini_model in [
     "gemini-2.5-flash",
     "gemini-2.5-flash-lite",
     "gemini-3-pro-preview",
+    "gemini-3-flash-preview",
 ]:
     converter_config_registry.register(
         gemini_model,
@@ -120,12 +121,9 @@ for gemini_model in [
         ),
     )
 for openai_model in [
-    "gpt-5.1",
-    "gpt-5.1-mini",
-    "gpt-5.1-nano",
+    "gpt-5.2",
     "gpt-5",
     "gpt-5-mini",
-    "gpt-5-nano",
 ]:
     converter_config_registry.register(
         openai_model,

{vlmparse-0.1.4 → vlmparse-0.1.5}/vlmparse/servers/docker_server.py

@@ -78,7 +78,7 @@ class VLLMDockerServerConfig(DockerServerConfig):
         from vlmparse.clients.openai_converter import LLMParams
 
         return LLMParams(
-            base_url=f"http://localhost:{self.docker_port}/v1",
+            base_url=f"http://localhost:{self.docker_port}{self.get_base_url_suffix()}",
             model_name=self.default_model_name,
         )
 

{vlmparse-0.1.4 → vlmparse-0.1.5}/vlmparse/servers/utils.py

@@ -3,9 +3,8 @@ import time
 from contextlib import contextmanager
 from pathlib import Path
 
-from loguru import logger
-
 import docker
+from loguru import logger
 
 
 def _ensure_image_exists(
@@ -230,6 +229,8 @@ def get_model_from_uri(uri: str) -> str:
     for container in containers:
         c_uri = container.labels.get("vlmparse_uri")
         c_model = container.labels.get("vlmparse_model_name")
+        if c_uri is not None:
+            c_uri = c_uri.replace("localhost", "0.0.0.0")
 
         # Check if user URI matches container URI (ignoring /v1 suffix if missing)
         if c_uri and (

{vlmparse-0.1.4 → vlmparse-0.1.5/vlmparse.egg-info}/PKG-INFO

@@ -1,7 +1,7 @@
 Metadata-Version: 2.4
 Name: vlmparse
-Version: 0.1.4
-Requires-Python: >=3.12.0
+Version: 0.1.5
+Requires-Python: >=3.11.0
 Description-Content-Type: text/markdown
 License-File: LICENSE
 Requires-Dist: devtools>=0.12.2
@@ -72,6 +72,19 @@ Supported Converters:
 
 ## Installation
 
+Simplest solution with only the cli: 
+
+```bash
+uv tool install vlmparse
+```
+
+If you want to run the granite-docling model or use the streamlit viewing app:
+
+```bash
+uv tool install vlmparse[docling_core,st_app]
+```
+
+If you prefer cloning the repository and using the local version:
 ```bash
 uv sync
 ```
@@ -86,10 +99,11 @@ Activate the virtual environment:
 ```bash
 source .venv/bin/activate
 ```
-Other solution: append uv run to all the commands below.
 
 ## CLI Usage
 
+Note that you can bypass the previous installation step and just add uvx before each of the commands below.
+
 ### Convert PDFs
 
 With a general VLM (requires setting your api key as an environment variable):

{vlmparse-0.1.4 → vlmparse-0.1.5}/vlmparse.egg-info/SOURCES.txt

@@ -3,11 +3,8 @@ README.md
 pyproject.toml
 tests/test_all_converters_mocked.py
 tests/test_batch_parser.py
-tests/test_benchmark_tests.py
 tests/test_cli.py
 tests/test_end2end.py
-tests/test_process_and_run_benchmark.py
-tests/test_table_tests.py
 vlmparse/base_model.py
 vlmparse/build_doc.py
 vlmparse/cli.py
@@ -22,21 +19,6 @@ vlmparse.egg-info/dependency_links.txt
 vlmparse.egg-info/entry_points.txt
 vlmparse.egg-info/requires.txt
 vlmparse.egg-info/top_level.txt
-vlmparse/benchpdf2md/create_dataset.py
-vlmparse/benchpdf2md/run_benchmark.py
-vlmparse/benchpdf2md/utils.py
-vlmparse/benchpdf2md/bench_tests/benchmark_tsts.py
-vlmparse/benchpdf2md/bench_tests/utils.py
-vlmparse/benchpdf2md/olmocrbench/repeatdetect.py
-vlmparse/benchpdf2md/olmocrbench/run_olmocr_bench.py
-vlmparse/benchpdf2md/olmocrbench/tests.py
-vlmparse/benchpdf2md/olmocrbench/katex/__init__.py
-vlmparse/benchpdf2md/olmocrbench/katex/render.py
-vlmparse/benchpdf2md/st_visu_benchmark/app.py
-vlmparse/benchpdf2md/st_visu_benchmark/highligh_text.py
-vlmparse/benchpdf2md/st_visu_benchmark/test_form.py
-vlmparse/benchpdf2md/st_visu_benchmark/ui_elements.py
-vlmparse/benchpdf2md/st_visu_benchmark/utils.py
 vlmparse/clients/chandra.py
 vlmparse/clients/deepseekocr.py
 vlmparse/clients/docling.py