Conversation
This commit introduces support for using Hugging Face's hosted Inference API to anonymize documents. Key changes: - Added `huggingface-hub` as a dependency. - Updated `ModelProvider` and `ModelName` enums to include Hugging Face models. - Implemented the logic to call the Hugging Face API using `InferenceClient`. - Updated the CLI's `README.md` to include instructions for the `HUGGING_FACE_TOKEN` environment variable and list the new models.
This commit introduces support for using Hugging Face's hosted Inference API to anonymize documents. Key changes: - Added `huggingface-hub` as a dependency. - Updated `ModelProvider` and `ModelName` enums to include Hugging Face models. - Implemented the logic to call the Hugging Face API using `InferenceClient`. - Updated the CLI's `README.md` to include instructions for the `HUGGING_FACE_TOKEN` environment variable, a link to obtain the token, and a list of the new models.
Summary of ChangesHello @leo-gan, I'm Gemini Code Assist1! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed! This pull request significantly enhances the PDF Anonymizer application by integrating Hugging Face models as a new provider for Large Language Models. This allows users to leverage a broader range of LLMs for anonymization tasks, providing greater flexibility and choice. The changes include updates to core logic for model interaction, new configuration requirements, and comprehensive documentation to guide users through the setup and usage of these new models. Highlights
Using Gemini Code AssistThe full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips. Invoking Gemini You can request assistance from Gemini at any point by creating a comment using either
Customization To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a Limitations & Feedback Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for Github and other Google products, sign up here. You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension. Footnotes
|
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
There was a problem hiding this comment.
Code Review
This pull request adds support for using Hugging Face models for anonymization. The changes include updating dependencies, adding configuration for Hugging Face models, and implementing the client logic to call the Hugging Face Inference API. The documentation and environment variable templates have also been updated accordingly.
My review focuses on improving the robustness, performance, and maintainability of the new Hugging Face integration. I've suggested adding validation for the Hugging Face token, optimizing client instantiation to avoid re-creation in retry loops, and refactoring the response parsing logic for better readability. I also pointed out a potential fragility in how model providers are determined.
| elif model_enum.provider == ModelProvider.HUGGINGFACE: | ||
| client = InferenceClient( | ||
| model=model_name, token=os.getenv("HUGGING_FACE_TOKEN") | ||
| ) | ||
| response = client.chat_completion( | ||
| messages=[{"role": "user", "content": prompt}], | ||
| ) |
There was a problem hiding this comment.
It's good practice to validate the presence of the HUGGING_FACE_TOKEN before attempting to use it. This provides a clearer error message to the user if the token is missing. The current implementation would rely on the InferenceClient to fail, which might result in a less specific error message and unnecessary retries.
elif model_enum.provider == ModelProvider.HUGGINGFACE:
token = os.getenv("HUGGING_FACE_TOKEN")
if not token:
raise ValueError(
"HUGGING_FACE_TOKEN environment variable not set. Please set it in your .env file."
)
client = InferenceClient(model=model_name, token=token)
response = client.chat_completion(
messages=[{"role": "user", "content": prompt}],
)| client = InferenceClient( | ||
| model=model_name, token=os.getenv("HUGGING_FACE_TOKEN") | ||
| ) |
There was a problem hiding this comment.
For performance optimization, consider instantiating the InferenceClient outside the for loop. Creating a new client on every retry attempt is inefficient. This observation also applies to the genai.Client() instantiation in the else block.
| elif model_enum.provider == ModelProvider.HUGGINGFACE: | ||
| if ( | ||
| response | ||
| and hasattr(response, "choices") | ||
| and response.choices | ||
| and hasattr(response.choices[0], "message") | ||
| and hasattr(response.choices[0].message, "content") | ||
| ): | ||
| return response.choices[0].message.content or "" |
There was a problem hiding this comment.
This series of checks can be simplified using a try...except block, which is more idiomatic in Python (EAFP - Easier to Ask for Forgiveness than Permission). This makes the code cleaner and easier to read. The try...except block will gracefully handle cases where response is None, or any of the nested attributes (choices, message, content) or list indices are missing.
elif model_enum.provider == ModelProvider.HUGGINGFACE:
try:
return response.choices[0].message.content or ""
except (AttributeError, IndexError, TypeError):
return ""| if "/" in self.value: | ||
| return ModelProvider.HUGGINGFACE |
There was a problem hiding this comment.
Relying on the presence of a '/' in the model name to determine the provider is a bit fragile. While it works for the current Hugging Face models, this heuristic might break in the future if other providers adopt a similar naming scheme or if Hugging Face model names change. A more robust approach would be to explicitly map each model to its provider, for example, using a dictionary. This would make the code more maintainable and less prone to errors.
1 participant
No description provided.