Omniparse v2 #112
Omniparse v2 #112
Conversation
Migrate PDF Processing Core from Marker to Docling
WalkthroughThis update transitions the codebase from using Marker and Surya OCR models to Docling models for document parsing, updates licensing and attribution to MIT and Docling, and migrates PDF, PPT, and DOC parsing logic to Docling-based workflows. The project configuration is modernized, dependencies updated, and documentation revised to reflect these changes. Changes
Sequence Diagram(s)sequenceDiagram
participant User
participant API (FastAPI)
participant DoclingConverter
participant Florence2Model
participant Utils
User->>API: Uploads PDF/PPT/DOC file
API->>DoclingConverter: convert(file bytes)
DoclingConverter-->>API: Parsed document object
API->>DoclingConverter: Export to markdown
DoclingConverter-->>API: Markdown text
API->>Utils: encode_images(document_name, document, responseDocument)
Utils-->>API: Images encoded in response
API-->>User: Returns responseDocument with text, metadata, images
Estimated code review effort🎯 4 (Complex) | ⏱️ ~35 minutes Poem
Note ⚡️ Unit Test Generation is now available in beta!Learn more here, or try it out under "Finishing Touches" below. ✨ Finishing Touches
🧪 Generate unit tests
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. 🪧 TipsChatThere are 3 ways to chat with CodeRabbit:
SupportNeed help? Create a ticket on our support page for assistance with any issues or questions. Note: Be mindful of the bot's finite context window. It's strongly recommended to break down tasks such as reading entire modules into smaller chunks. For a focused discussion, use review comments to chat about specific files and their changes, instead of using the PR comments. CodeRabbit Commands (Invoked using PR comments)
Other keywords and placeholders
CodeRabbit Configuration File (
|
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 5
🔭 Outside diff range comments (9)
omniparse/media/__init__.py (2)
52-55:temp_audio_pathmay be undefined →UnboundLocalErrorduring cleanup
If the function raisesValueError(invalid input), thefinallyblock still referencestemp_audio_path, which was never set.- finally: - # Clean up the temporary file - if os.path.exists(temp_audio_path): - os.remove(temp_audio_path) + finally: + # Clean up the temporary file if it was created + if 'temp_audio_path' in locals() and os.path.exists(temp_audio_path): + os.remove(temp_audio_path)
90-95: Same undefined-variable risk forvideo_path/audio_pathinparse_video
A validation error before these variables are assigned will crash the cleanup block.- finally: - # Clean up the temporary files - if os.path.exists(video_path): - os.remove(video_path) - if os.path.exists(audio_path): - os.remove(audio_path) + finally: + # Clean up the temporary files if they were created + if 'video_path' in locals() and os.path.exists(video_path): + os.remove(video_path) + if 'audio_path' in locals() and os.path.exists(audio_path): + os.remove(audio_path)pyproject.toml (1)
35-47:tool.uvindexes: platform marker syntax might be ignored
sys_platform == 'linux' or sys_platform == 'win32'is not valid PEP 508 syntax inside TOML arrays.
Use a separate table per marker or wrap with extras according to uv docs, e.g.torch = [ { index = "pytorch-cu124", markers = "sys_platform == 'linux' or sys_platform == 'win32'" } ]Validate with
uv pip install -r pyproject.tomlbefore merging.Dockerfile (1)
1-1: Fix CUDA version inconsistency.The CUDA_VERSION argument is set to "11.8.0" but the PyTorch installation uses the cu124 index, creating an inconsistency that could cause confusion or deployment issues.
Apply this diff to align the CUDA version:
-ARG CUDA_VERSION="11.8.0" +ARG CUDA_VERSION="12.4.0"Also applies to: 49-49
omniparse/image/__init__.py (1)
45-108: Consider adding validation for Docling converter availabilityThe function should verify that the Docling converter is initialized before use.
Add this validation at the beginning of the function:
def parse_image(image_name, input_data, model_state) -> dict: + if not model_state.docling_converter: + raise ValueError("Docling converter not initialized. Please load document models first.") + temp_files = []omniparse/documents/router.py (3)
112-112: Fix incorrect file suffix for Word documentsThe parse_doc_endpoint uses ".ppt" suffix for Word documents, which is incorrect.
- with tempfile.NamedTemporaryFile(delete=False, suffix=".ppt") as tmp_ppt: + with tempfile.NamedTemporaryFile(delete=False, suffix=".docx") as tmp_doc: + tmp_doc.write(await file.read()) + tmp_doc.flush() + input_path = tmp_doc.name - tmp_ppt.write(await file.read()) - tmp_ppt.flush() - input_path = tmp_ppt.name
62-152: Refactor duplicate code between PPT and DOC endpointsThe parse_ppt_endpoint and parse_doc_endpoint have nearly identical implementations. This violates the DRY principle.
Create a shared helper function:
async def _convert_office_to_pdf_and_parse(file: UploadFile, suffix: str): """Convert Office documents to PDF and parse with Docling.""" with tempfile.NamedTemporaryFile(delete=False, suffix=suffix) as tmp_file: tmp_file.write(await file.read()) tmp_file.flush() input_path = tmp_file.name output_dir = tempfile.mkdtemp() try: command = [ "libreoffice", "--headless", "--convert-to", "pdf", "--outdir", output_dir, input_path, ] result = subprocess.run(command, check=True, capture_output=True, text=True) output_pdf_path = os.path.join( output_dir, os.path.splitext(os.path.basename(input_path))[0] + ".pdf" ) with open(output_pdf_path, "rb") as pdf_file: pdf_bytes = pdf_file.read() source = DocumentStream(name=file.filename, stream=BytesIO(pdf_bytes)) filetype = os.path.splitext(file.filename)[1].lstrip('.').upper() out_meta = {"filename": file.filename, "filetype": filetype} docling_result = model_state.docling_converter.convert(source) full_text = docling_result.document.export_to_markdown() out_meta["block_stats"] = { "images": len(docling_result.document.pictures), "tables": len(docling_result.document.tables), } result = responseDocument(text=full_text, metadata=out_meta) encode_images(file.filename, docling_result.document, result) return result except subprocess.CalledProcessError as e: raise HTTPException( status_code=500, detail=f"LibreOffice conversion failed: {e.stderr}" ) finally: os.remove(input_path) if os.path.exists(output_pdf_path): os.remove(output_pdf_path) os.rmdir(output_dir)Then simplify the endpoints:
@document_router.post("/ppt") async def parse_ppt_endpoint(file: UploadFile = File(...)): try: result = await _convert_office_to_pdf_and_parse(file, ".pptx") return JSONResponse(content=result.model_dump()) except Exception as e: raise HTTPException(status_code=500, detail=str(e)) @document_router.post("/docs") async def parse_doc_endpoint(file: UploadFile = File(...)): try: result = await _convert_office_to_pdf_and_parse(file, ".docx") return JSONResponse(content=result.model_dump()) except Exception as e: raise HTTPException(status_code=500, detail=str(e))
167-208: Add proper cleanup with try/finally blockFile cleanup should be in a finally block to ensure it happens even if an exception occurs.
Wrap the logic in try/finally:
with tempfile.NamedTemporaryFile(delete=False) as tmp_file: tmp_file.write(await file.read()) tmp_file.flush() input_path = tmp_file.name +output_dir = None +try: if file_ext.lower() in {".ppt", ".pptx", ".doc", ".docx"}: output_dir = tempfile.mkdtemp() # ... conversion logic ... # ... parsing logic ... result = responseDocument(text=full_text, metadata=out_meta) encode_images(file.filename, docling_result.document, result) return JSONResponse(content=result.model_dump()) +finally: + if os.path.exists(input_path): + os.remove(input_path) + if output_dir and os.path.exists(output_dir): + for f in os.listdir(output_dir): + os.remove(os.path.join(output_dir, f)) + os.rmdir(output_dir) - - os.remove(input_path) - - result = responseDocument(text=full_text, metadata=out_meta) - encode_images(file.filename, docling_result.document, result) - - return JSONResponse(content=result.model_dump())omniparse/__init__.py (1)
50-78: Add error handling for model loadingModel loading can fail for various reasons (network issues, insufficient memory, etc.). Add proper error handling.
Wrap each model loading section in try-except blocks:
def load_omnimodel(load_documents: bool, load_media: bool, load_web: bool): global shared_state print_omniparse_text_art() device = torch.device("cuda" if torch.cuda.is_available() else "cpu") + if load_documents: - print("[LOG] ✅ Loading OCR Model") - download_models() - shared_state.docling_converter = DocumentConverter( - format_options={ - InputFormat.PDF: PdfFormatOption(pipeline_options=pipeline_options) - } - ) + try: + print("[LOG] ✅ Loading OCR Model") + download_models() + shared_state.docling_converter = DocumentConverter( + format_options={ + InputFormat.PDF: PdfFormatOption(pipeline_options=pipeline_options) + } + ) + except Exception as e: + print(f"[ERROR] ❌ Failed to load OCR Model: {e}") + raise + - print("[LOG] ✅ Loading Vision Model") - shared_state.vision_model = AutoModelForCausalLM.from_pretrained( - "microsoft/Florence-2-base", torch_dtype=torch.float32, trust_remote_code=True - ).to(device) - shared_state.vision_processor = AutoProcessor.from_pretrained( - "microsoft/Florence-2-base", trust_remote_code=True - ) + try: + print("[LOG] ✅ Loading Vision Model") + shared_state.vision_model = AutoModelForCausalLM.from_pretrained( + "microsoft/Florence-2-base", torch_dtype=torch.float32, trust_remote_code=True + ).to(device) + shared_state.vision_processor = AutoProcessor.from_pretrained( + "microsoft/Florence-2-base", trust_remote_code=True + ) + except Exception as e: + print(f"[ERROR] ❌ Failed to load Vision Model: {e}") + raise if load_media: - print("[LOG] ✅ Loading Audio Model") - shared_state.whisper_model = whisper.load_model("small") + try: + print("[LOG] ✅ Loading Audio Model") + shared_state.whisper_model = whisper.load_model("small") + except Exception as e: + print(f"[ERROR] ❌ Failed to load Audio Model: {e}") + raise if load_web: - print("[LOG] ✅ Loading Web Crawler") - shared_state.crawler = WebCrawler(verbose=True) + try: + print("[LOG] ✅ Loading Web Crawler") + shared_state.crawler = WebCrawler(verbose=True) + except Exception as e: + print(f"[ERROR] ❌ Failed to load Web Crawler: {e}") + raise
🧹 Nitpick comments (4)
omniparse/media/__init__.py (1)
74-79: Use context managers to guarantee clip closure and reduce boilerplate
VideoFileClipand itsaudioclip both support context-manager semantics, eliminating manualclose()calls.- video_clip = VideoFileClip(video_path) - audio_clip = video_clip.audio - audio_clip.write_audiofile(audio_path) - audio_clip.close() - video_clip.close() + with VideoFileClip(video_path) as video_clip: + with video_clip.audio as audio_clip: + audio_clip.write_audiofile(audio_path)docs/README.md (1)
303-305: Fix markdown list formatting for consistency.The static analysis tool flagged inconsistent list formatting. The project appears to use asterisks for unordered lists elsewhere.
Apply this diff to fix the formatting:
-- Docling IBM models -- Florence-2 base -- Whisper Small +* Docling IBM models +* Florence-2 base +* Whisper Smallomniparse/image/__init__.py (1)
13-14: Fix module description to reflect image parsing functionalityThe description mentions "pdf/word/ppt parsing" but this module is specifically for image parsing.
Description: -This section of the code was adapted from the Docling repository to enhance text pdf/word/ppt parsing. +This section of the code was adapted from the Docling repository to enhance image parsing and text extraction. All credits for the original implementation go to Docling.omniparse/__init__.py (1)
33-39: Improve type annotations for SharedStateUsing
Anytype reduces type safety and IDE support. Consider using proper type annotations.+from typing import Optional +from docling.document_converter import DocumentConverter +from transformers import PreTrainedModel, PreTrainedTokenizer class SharedState(BaseModel): - docling_converter: Any = None - vision_model: Any = None - vision_processor: Any = None - whisper_model: Any = None - crawler: Any = None + docling_converter: Optional[DocumentConverter] = None + vision_model: Optional[PreTrainedModel] = None + vision_processor: Optional[PreTrainedTokenizer] = None + whisper_model: Optional[Any] = None # whisper doesn't provide type hints + crawler: Optional[WebCrawler] = None
📜 Review details
Configuration used: CodeRabbit UI
Review profile: CHILL
Plan: Pro
⛔ Files ignored due to path filters (1)
uv.lockis excluded by!**/*.lock
📒 Files selected for processing (17)
.gitignore(1 hunks).python-version(1 hunks)Dockerfile(2 hunks)README.md(6 hunks)docs/README.md(2 hunks)docs/deployment.md(1 hunks)docs/installation.md(1 hunks)examples/OmniParse_GoogleColab.ipynb(5 hunks)omniparse/__init__.py(3 hunks)omniparse/demo.py(1 hunks)omniparse/documents/__init__.py(0 hunks)omniparse/documents/router.py(5 hunks)omniparse/image/__init__.py(3 hunks)omniparse/image/router.py(1 hunks)omniparse/media/__init__.py(1 hunks)omniparse/utils.py(1 hunks)pyproject.toml(1 hunks)
💤 Files with no reviewable changes (1)
- omniparse/documents/init.py
🧰 Additional context used
🧬 Code Graph Analysis (3)
omniparse/utils.py (1)
omniparse/models/__init__.py (2)
responseDocument(15-64)add_image(21-49)
omniparse/documents/router.py (3)
omniparse/__init__.py (1)
get_shared_state(80-81)omniparse/utils.py (1)
encode_images(8-21)omniparse/models/__init__.py (1)
responseDocument(15-64)
omniparse/__init__.py (1)
omniparse/utils.py (1)
print_omniparse_text_art(24-34)
🪛 markdownlint-cli2 (0.17.2)
docs/README.md
303-303: Unordered list style
Expected: asterisk; Actual: dash
(MD004, ul-style)
304-304: Unordered list style
Expected: asterisk; Actual: dash
(MD004, ul-style)
305-305: Unordered list style
Expected: asterisk; Actual: dash
(MD004, ul-style)
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (1)
- GitHub Check: build-and-push
🔇 Additional comments (17)
docs/installation.md (1)
35-35: Bullet wording update looks correct
The switch to “Docling models and Florence-2” is accurate and consistent with the rest of the PR. No further action needed.pyproject.toml (1)
12-33: Unreleased / future package versions will break installation
torch>=2.6.0,html2text>=2025.4.15,openai-whisper>=20240930, etc. do not exist on PyPI today.pipwill fail immediately.Recommend pinning to the latest released versions or using “~=latest_minor” constraints.
.gitignore (1)
165-166: ✅ Good addition
Ignoring.vscode/keeps editor artefacts out of the repo..python-version (1)
1-1: Python version pin acknowledgedIncluding
.python-versionaligns local tooling with therequires-pythonfield.omniparse/demo.py (1)
136-136: Description updated successfullyThe flag description now reflects the Docling migration. No further changes required.
docs/deployment.md (1)
31-31: LGTM! Documentation cleanup aligns with migration.The removal of Skypilot deployment instructions and retention of Docker-based deployment documentation is consistent with the project's modernization efforts moving to Docling models.
Dockerfile (1)
58-58: LGTM! Transformers version updated appropriately.The upgrade from transformers 4.41.2 to 4.50.3 aligns with the migration to Docling models and ensures compatibility with newer features.
omniparse/image/router.py (1)
15-15: Verifiedparse_imagesignature accepts filename parameter.The function in
omniparse/image/__init__.pyis declared as:def parse_image(image_name, input_data, model_state) -> dict:which directly matches the call in
omniparse/image/router.py:result: responseDocument = parse_image(file.filename, file_bytes, model_state)No further changes needed.
docs/README.md (2)
108-108: LGTM! Documentation updated for Docling migration.The update correctly reflects the change from "Surya OCR series of models" to "Docling models and Florence-2" in the server documentation.
299-299: LGTM! Acknowledgements updated appropriately.The acknowledgements section correctly credits the Docling project instead of the previous Marker project, reflecting the migration.
examples/OmniParse_GoogleColab.ipynb (4)
46-54: LGTM! Added uv installation for modern dependency management.The addition of uv installation aligns with the project's migration to modern Python tooling and dependency management.
84-84: LGTM! Dependency installation updated to use uv.The change from
pip install -e .touv -q syncis consistent with the migration to uv for dependency management.
93-96: LGTM! Proactive fix for matplotlib backend issues.The addition of matplotlib-inline installation with a helpful comment addresses a common issue in Colab environments.
202-202: LGTM! Server launch commands updated for uv.The server launch commands are correctly updated to use
uv run server.pyinstead of direct Python execution, maintaining consistency with the uv migration.Also applies to: 250-250
README.md (2)
96-109: Documentation updates look good!The model references have been consistently updated to reflect the migration from Surya OCR to Docling models.
296-319: Limitations and acknowledgements properly updatedThe documentation correctly reflects the new Docling-based implementation and gives appropriate credit.
omniparse/__init__.py (1)
65-66: Verify torch.float32 usage for vision modelThe vision model is explicitly set to use float32, which uses more memory than float16. Was this intentional?
If memory usage is a concern and the model supports it, consider using float16:
shared_state.vision_model = AutoModelForCausalLM.from_pretrained( - "microsoft/Florence-2-base", torch_dtype=torch.float32, trust_remote_code=True + "microsoft/Florence-2-base", torch_dtype=torch.float16 if device.type == "cuda" else torch.float32, trust_remote_code=True ).to(device)This would use float16 on GPU (where it's typically supported) and float32 on CPU.
| @@ -163,36 +186,22 @@ async def parse_any_endpoint(file: UploadFile = File(...)): | |||
| ) | |||
| input_path = output_pdf_path | |||
|
|
|||
| # Common parsing logic | |||
| full_text, images, out_meta = convert_single_pdf(input_path, model_state.model_list) | |||
| docling_result = model_state.docling_converter.convert(Path(input_path)) | |||
There was a problem hiding this comment.
🛠️ Refactor suggestion
Use DocumentStream consistently across all endpoints
This endpoint uses Path while other endpoints use DocumentStream. This inconsistency could lead to different behavior.
- docling_result = model_state.docling_converter.convert(Path(input_path))
+ with open(input_path, "rb") as f:
+ file_bytes = f.read()
+ source = DocumentStream(name=file.filename, stream=BytesIO(file_bytes))
+ docling_result = model_state.docling_converter.convert(source)Committable suggestion skipped: line range outside the PR's diff.
🤖 Prompt for AI Agents
In omniparse/documents/router.py at line 189, replace the use of
Path(input_path) with a DocumentStream instance to maintain consistency across
all endpoints. Modify the code so that docling_converter.convert receives a
DocumentStream object instead of a Path, ensuring uniform behavior and interface
usage throughout the codebase.
| from fastapi import UploadFile | ||
| from fastapi.responses import JSONResponse | ||
| from moviepy.editor import VideoFileClip | ||
| from moviepy import VideoFileClip |
There was a problem hiding this comment.
Importing VideoFileClip from the top-level moviepy module will raise ImportError on most installs
VideoFileClip is re-exported only by moviepy.editor; the root package does not expose it. Importing from the wrong path will break parse_video at runtime.
-from moviepy import VideoFileClip
+from moviepy.editor import VideoFileClip🤖 Prompt for AI Agents
In omniparse/media/__init__.py at line 22, change the import of VideoFileClip
from the top-level moviepy module to import it from moviepy.editor instead,
because VideoFileClip is only re-exported by moviepy.editor and importing from
moviepy directly causes ImportError at runtime.
| def encode_images(document_name, document, inputDocument: responseDocument): | ||
| file_name = os.path.splitext(document_name)[0] | ||
| for idx, image in enumerate(document.pictures): | ||
| with tempfile.NamedTemporaryFile(delete=True, suffix=".png") as tmp_file: | ||
| filename = tmp_file.name | ||
| image.get_image(document).save(filename, "PNG") | ||
| with open(filename, "rb") as f: | ||
| image_bytes = f.read() | ||
|
|
||
| inputDocument.add_image(image_name=filename, image_data=image_base64) | ||
|
|
||
| # Remove the temporary image file | ||
| os.remove(filename) | ||
| # Convert image to base64 | ||
| image_base64 = base64.b64encode(image_bytes).decode("utf-8") | ||
| inputDocument.add_image( | ||
| image_name=f"{file_name}_image_{idx}", image_data=image_base64 | ||
| ) |
There was a problem hiding this comment.
🛠️ Refactor suggestion
Add error handling and optimize file operations
The function should handle potential errors during image extraction and can be optimized to avoid redundant file operations.
Apply this diff to improve error handling and efficiency:
def encode_images(document_name, document, inputDocument: responseDocument):
file_name = os.path.splitext(document_name)[0]
+ if not hasattr(document, 'pictures'):
+ return
+
for idx, image in enumerate(document.pictures):
- with tempfile.NamedTemporaryFile(delete=True, suffix=".png") as tmp_file:
- filename = tmp_file.name
- image.get_image(document).save(filename, "PNG")
- with open(filename, "rb") as f:
- image_bytes = f.read()
-
- # Convert image to base64
- image_base64 = base64.b64encode(image_bytes).decode("utf-8")
- inputDocument.add_image(
- image_name=f"{file_name}_image_{idx}", image_data=image_base64
- )
+ try:
+ # Get PIL image and convert directly to bytes
+ pil_image = image.get_image(document)
+ buffer = io.BytesIO()
+ pil_image.save(buffer, format="PNG")
+ image_bytes = buffer.getvalue()
+
+ # Convert image to base64
+ image_base64 = base64.b64encode(image_bytes).decode("utf-8")
+ inputDocument.add_image(
+ image_name=f"{file_name}_image_{idx}", image_data=image_base64
+ )
+ except Exception as e:
+ print(f"Warning: Failed to extract image {idx} from {document_name}: {e}")
+ continueDon't forget to add the import:
import io🤖 Prompt for AI Agents
In omniparse/utils.py around lines 8 to 21, the encode_images function lacks
error handling and inefficiently writes images to temporary files before reading
them back. To fix this, wrap the image processing code in a try-except block to
catch and handle exceptions during image extraction. Replace the temporary file
usage by saving the image directly into an in-memory bytes buffer using
io.BytesIO, then encode the image bytes from this buffer to base64. Also, add
the import statement "import io" at the top of the file.
| [project] | ||
| name = "omniparse" | ||
| version = "0.0.1" | ||
| version = "0.0.2" | ||
| description = "API to convert Documents | Media | Webpage into Structured Markdown for LLM usecases" | ||
| authors = ["Adithya S K <[email protected]>"] | ||
| license = "Apache" | ||
| authors = [ | ||
| {name = "Adithya S K", email = "[email protected]"} | ||
| ] | ||
| license = {file = "LICENSE"} | ||
| readme = "README.md" |
There was a problem hiding this comment.
[build-system] table was removed – wheels can no longer be built
PEP 517 builds require a [build-system] section. Without it, cloning users (and CI/CD) will be unable to pip install . or publish to PyPI.
+[build-system]
+requires = ["setuptools>=69", "wheel"]
+build-backend = "setuptools.build_meta"Add this (or an equivalent backend such as uv) to restore buildability.
📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
| [project] | |
| name = "omniparse" | |
| version = "0.0.1" | |
| version = "0.0.2" | |
| description = "API to convert Documents | Media | Webpage into Structured Markdown for LLM usecases" | |
| authors = ["Adithya S K <[email protected]>"] | |
| license = "Apache" | |
| authors = [ | |
| {name = "Adithya S K", email = "[email protected]"} | |
| ] | |
| license = {file = "LICENSE"} | |
| readme = "README.md" | |
| [project] | |
| name = "omniparse" | |
| version = "0.0.2" | |
| description = "API to convert Documents | Media | Webpage into Structured Markdown for LLM usecases" | |
| authors = [ | |
| {name = "Adithya S K", email = "[email protected]"} | |
| ] | |
| license = {file = "LICENSE"} | |
| readme = "README.md" | |
| [build-system] | |
| requires = ["setuptools>=69", "wheel"] | |
| build-backend = "setuptools.build_meta" |
🤖 Prompt for AI Agents
In pyproject.toml lines 1 to 9, the [build-system] table is missing, which is
required by PEP 517 for building wheels and installing the package. Add a
[build-system] section specifying the build backend (e.g.,
"setuptools.build_meta") and the required build dependencies (e.g., "setuptools"
and "wheel") to enable pip install and PyPI publishing.
| @@ -50,13 +50,15 @@ conda activate omniparse-venv | |||
| Install Dependencies: | |||
|
|
|||
| ```bash | |||
| poetry install | |||
| uv sync | |||
| # or | |||
| pip install -e . | |||
| # or | |||
| pip install -r pyproject.toml | |||
There was a problem hiding this comment.
Fix incorrect pip install command
The command pip install -r pyproject.toml is incorrect. The -r flag expects a requirements.txt file, not a pyproject.toml file.
Apply this diff to fix the installation command:
-pip install -r pyproject.toml
+pip install .Alternatively, if you want to list all installation methods:
-uv sync
-# or
-pip install -e .
-# or
-pip install -r pyproject.toml
+uv sync # Using uv package manager
+# or
+pip install -e . # Editable install with pip
+# or
+pip install . # Regular install with pip📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
| pip install -r pyproject.toml | |
| pip install . |
🤖 Prompt for AI Agents
In README.md at line 57, the pip install command incorrectly uses the -r flag
with pyproject.toml, which is invalid because -r expects a requirements.txt
file. Replace the command with the correct installation method for
pyproject.toml, such as using a tool like poetry or pip install with the
appropriate options, or remove the -r flag if installing from a requirements
file. Update the command to reflect the proper way to install dependencies from
pyproject.toml.
Summary by CodeRabbit
New Features
.python-versionfile for consistent Python 3.10 environments.Bug Fixes
Documentation
Refactor
pyproject.toml.Chores
.gitignoreto exclude.vscode/directory.