diff --git a/Makefile b/Makefile
index 2ec1704..834b5d9 100644
--- a/Makefile
+++ b/Makefile
@@ -4,7 +4,7 @@ lint:
ruff check . --fix
format-nb:
- black --ipynb examples/*.ipynb
+ black --ipynb docs/examples/*.ipynb
format: format-nb
black .
diff --git a/docs/examples/deepseek_demo.ipynb b/docs/examples/deepseek_demo.ipynb
new file mode 100644
index 0000000..db09bb5
--- /dev/null
+++ b/docs/examples/deepseek_demo.ipynb
@@ -0,0 +1,68 @@
+{
+ "cells": [
+ {
+ "cell_type": "markdown",
+ "metadata": {},
+ "source": [
+ "# Using Vision Parse with DeepSeek"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {},
+ "source": [
+ "This notebook demonstrates how to use Vision Parse with DeepSeek."
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "!pip install 'vision-parse[openai]' -Uqq # install the vision-parse package with openai"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "from vision_parse import VisionParser\n",
+ "import os\n",
+ "\n",
+ "# Initialize parser\n",
+ "parser = VisionParser(\n",
+ " model_name=\"deepseek-chat\",\n",
+ " api_key=os.getenv(\"DEEPSEEK_API_KEY\"),\n",
+ " temperature=0.9,\n",
+ " top_p=0.4,\n",
+ " image_mode=None,\n",
+ " detailed_extraction=True,\n",
+ " enable_concurrency=True,\n",
+ ")\n",
+ "\n",
+ "pdf_path = \"../tests/Texas-Holdem-Rules.pdf\"\n",
+ "markdown_pages = parser.convert_pdf(pdf_path)\n",
+ "\n",
+ "# Print the markdown pages\n",
+ "for i, page_content in enumerate(markdown_pages):\n",
+ " print(f\"\\n--- Page {i+1} ---\\n{page_content}\")"
+ ]
+ }
+ ],
+ "metadata": {
+ "kernelspec": {
+ "display_name": "personal",
+ "language": "python",
+ "name": "python3"
+ },
+ "language_info": {
+ "name": "python",
+ "version": "3.10.9"
+ }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2
+}
diff --git a/pyproject.toml b/pyproject.toml
index e4e0a7a..d8f4fe0 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -61,11 +61,11 @@ gemini = [
"google-generativeai==0.8.3",
]
openai = [
- "openai>=1.59.8",
+ "openai==1.59.8",
]
all = [
"google-generativeai==0.8.3",
- "openai>=1.59.8",
+ "openai==1.59.8",
]
[tool.hatch.build.targets.wheel]
diff --git a/src/vision_parse/llm.py b/src/vision_parse/llm.py
index 3a1dd6b..7690f49 100644
--- a/src/vision_parse/llm.py
+++ b/src/vision_parse/llm.py
@@ -306,7 +306,7 @@ async def _get_response(
):
if self.provider == "ollama":
return await self._ollama(base64_encoded, prompt, structured)
- elif self.provider == "openai":
+ elif self.provider == "openai" or self.provider == "deepseek":
return await self._openai(base64_encoded, prompt, structured)
elif self.provider == "gemini":
return await self._gemini(base64_encoded, prompt, structured)
diff --git a/tests/test_llm.py b/tests/test_llm.py
index 8d0aaf4..f411c37 100644
--- a/tests/test_llm.py
+++ b/tests/test_llm.py
@@ -293,6 +293,66 @@ async def test_gemini_generate_markdown(
assert mock_client.generate_content_async.call_count == 2
+@pytest.mark.asyncio
+@patch("openai.AsyncOpenAI")
+async def test_deepseek_generate_markdown(
+ MockAsyncOpenAI, sample_base64_image, mock_pixmap
+):
+ """Test markdown generation using Deepseek."""
+ mock_client = AsyncMock()
+ MockAsyncOpenAI.return_value = mock_client
+
+ # Mock structured analysis response
+ mock_parse = AsyncMock()
+ mock_parse.choices = [
+ AsyncMock(
+ message=AsyncMock(
+ content=json.dumps(
+ {
+ "text_detected": "Yes",
+ "tables_detected": "No",
+ "images_detected": "No",
+ "latex_equations_detected": "No",
+ "extracted_text": "Test content",
+ "confidence_score_text": 0.9,
+ }
+ )
+ )
+ )
+ ]
+
+ # Mock markdown conversion response
+ mock_create = AsyncMock()
+ mock_create.choices = [
+ AsyncMock(message=AsyncMock(content="# Test Header\n\nTest content"))
+ ]
+ # Set up side effects to return mock_parse first, then mock_create
+ mock_client.chat.completions.create = AsyncMock(
+ side_effect=[mock_parse, mock_create]
+ )
+
+ llm = LLM(
+ model_name="deepseek-chat",
+ api_key="test-key",
+ temperature=0.7,
+ top_p=0.7,
+ ollama_config=None,
+ openai_config=None,
+ gemini_config=None,
+ image_mode=None,
+ custom_prompt=None,
+ detailed_extraction=True,
+ enable_concurrency=True,
+ device=None,
+ num_workers=1,
+ )
+ result = await llm.generate_markdown(sample_base64_image, mock_pixmap, 0)
+
+ assert isinstance(result, str)
+ assert "Test content" in result
+ assert mock_client.chat.completions.create.call_count == 2
+
+
@pytest.mark.asyncio
@patch("ollama.AsyncClient")
async def test_ollama_base64_image_mode(
@@ -389,65 +449,3 @@ async def test_ollama_llm_error(mock_async_client, sample_base64_image, mock_pix
await llm.generate_markdown(sample_base64_image, mock_pixmap, 0)
assert "Ollama Model processing failed" in str(exc_info.value)
assert mock_client.chat.call_count == 1
-
-
-@pytest.mark.asyncio
-@patch("openai.AsyncOpenAI")
-async def test_openai_llm_error(MockAsyncOpenAI, sample_base64_image, mock_pixmap):
- """Test LLMError handling for OpenAI."""
- mock_client = AsyncMock()
- MockAsyncOpenAI.return_value = mock_client
-
- # Mock API error for markdown generation
- mock_client.chat.completions.create.side_effect = Exception("OpenAI API error")
-
- llm = LLM(
- model_name="gpt-4o",
- api_key="test-key",
- temperature=0.7,
- top_p=0.7,
- ollama_config=None,
- openai_config=None,
- gemini_config=None,
- image_mode=None,
- custom_prompt=None,
- detailed_extraction=True,
- enable_concurrency=True,
- device=None,
- num_workers=1,
- )
-
- with pytest.raises(LLMError) as exc_info:
- await llm.generate_markdown(sample_base64_image, mock_pixmap, 0)
- assert "OpenAI Model processing failed" in str(exc_info.value)
-
-
-@pytest.mark.asyncio
-@patch("google.generativeai.GenerativeModel")
-async def test_gemini_llm_error(MockGenerativeModel, sample_base64_image, mock_pixmap):
- """Test LLMError handling for Gemini."""
- mock_client = AsyncMock()
- MockGenerativeModel.return_value = mock_client
-
- # Mock API error
- mock_client.generate_content.side_effect = Exception("Gemini API error")
-
- llm = LLM(
- model_name="gemini-1.5-pro",
- api_key="test-key",
- temperature=0.7,
- top_p=0.7,
- ollama_config=None,
- openai_config=None,
- gemini_config=None,
- image_mode=None,
- custom_prompt=None,
- detailed_extraction=True,
- enable_concurrency=True,
- device=None,
- num_workers=1,
- )
-
- with pytest.raises(LLMError) as exc_info:
- await llm.generate_markdown(sample_base64_image, mock_pixmap, 0)
- assert "Gemini Model processing failed" in str(exc_info.value)
diff --git a/uv.lock b/uv.lock
index c26234a..e441ab0 100644
--- a/uv.lock
+++ b/uv.lock
@@ -1162,7 +1162,7 @@ wheels = [
[[package]]
name = "openai"
-version = "1.59.9"
+version = "1.59.8"
source = { registry = "https://pypi.org/simple" }
dependencies = [
{ name = "anyio", version = "4.5.2", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version < '3.11'" },
@@ -1175,9 +1175,9 @@ dependencies = [
{ name = "tqdm" },
{ name = "typing-extensions" },
]
-sdist = { url = "https://files.pythonhosted.org/packages/ec/2d/04faa92bac0341649223398503db4415d2f658a757d9d32bb68f3378ddd0/openai-1.59.9.tar.gz", hash = "sha256:ec1a20b0351b4c3e65c6292db71d8233515437c6065efd4fd50edeb55df5f5d2", size = 347134 }
+sdist = { url = "https://files.pythonhosted.org/packages/e0/c4/b4482784de63c7158f6c0afcb07fd66450ea6c912d6bddf9d7599f2eda25/openai-1.59.8.tar.gz", hash = "sha256:ac4bda5fa9819fdc6127e8ea8a63501f425c587244bc653c7c11a8ad84f953e1", size = 346775 }
wheels = [
- { url = "https://files.pythonhosted.org/packages/07/b4/57f1954a4560092ad8c45f07ad183eab9c8e093e0a1db829f9b506b2d5d1/openai-1.59.9-py3-none-any.whl", hash = "sha256:61a0608a1313c08ddf92fe793b6dbd1630675a1fe3866b2f96447ce30050c448", size = 455527 },
+ { url = "https://files.pythonhosted.org/packages/8c/cf/5b235e12ead3cd2098f9792776c966994c1bc558cba5799e12f3045227df/openai-1.59.8-py3-none-any.whl", hash = "sha256:a8b8ee35c4083b88e6da45406d883cf6bd91a98ab7dd79178b8bc24c8bfb09d9", size = 455567 },
]
[[package]]
@@ -1808,7 +1808,7 @@ wheels = [
[[package]]
name = "vision-parse"
-version = "0.1.12.dev2+g6d52db1.d20250122"
+version = "0.1.12.dev15+g595bb42"
source = { editable = "." }
dependencies = [
{ name = "jinja2" },
@@ -1854,8 +1854,8 @@ requires-dist = [
{ name = "nest-asyncio", specifier = ">=1.6.0" },
{ name = "numpy", specifier = ">=2.0.0" },
{ name = "ollama", specifier = ">=0.4.4" },
- { name = "openai", marker = "extra == 'all'", specifier = ">=1.59.8" },
- { name = "openai", marker = "extra == 'openai'", specifier = ">=1.59.8" },
+ { name = "openai", marker = "extra == 'all'", specifier = "==1.59.8" },
+ { name = "openai", marker = "extra == 'openai'", specifier = "==1.59.8" },
{ name = "opencv-python", specifier = ">=4.10.0.84" },
{ name = "pydantic", specifier = ">=2.0.0" },
{ name = "pymupdf", specifier = ">=1.22.0" },