Fix/pyannote ollama

[Gambitnl]

No description provided.

[gemini-code-assist]

Summary of Changes

Hello @Gambitnl, 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 application's stability and user experience by addressing core issues in its processing pipeline. It introduces robust memory management for Ollama-based classification and fixes a critical embedding extraction bug in PyAnnote diarization, making these components more resilient. Furthermore, the user interface is improved with a live transcription preview and a fully functional, modern 'Characters' tab, providing users with better real-time feedback and comprehensive character management capabilities.

Highlights

• Robust Ollama Memory Handling: Implemented a retry mechanism for the Ollama classifier to gracefully handle memory allocation errors. This includes an initial retry with low-VRAM settings and an optional fallback model, significantly improving stability on systems with limited RAM. Preflight checks now also warn users about potential memory issues based on their selected model.
• PyAnnote Diarization Embedding Fix: Resolved a critical issue in PyAnnote 3.x diarization where embedding extraction would crash due to changes in output types (numpy arrays instead of torch tensors). A new helper function normalizes embedding outputs, and error handling ensures that individual embedding failures do not halt the entire diarization process.
• Live Transcription Preview in UI: Introduced a live-updating preview of chunk transcriptions in the Gradio UI. During Stage 3 processing, users can now see the latest transcribed text and timing metadata, providing real-time feedback and improving visibility into the transcription progress.
• Modern UI Character Tab Parity: The 'Characters' tab in the modern UI has been brought to feature parity with its legacy counterpart. It now includes a campaign-filtered table, character overview, import/export functionalities, and an automatic transcript extraction workflow, all integrated seamlessly with the campaign management system.

Using Gemini Code Assist

The 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 /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands on the current page.

Feature	Command	Description
Code Review	/gemini review	Performs a code review for the current pull request in its current state.
Pull Request Summary	/gemini summary	Provides a summary of the current pull request in its current state.
Comment	@gemini-code-assist	Responds in comments when explicitly tagged, both in pull request comments and review comments.
Help	/gemini help	Displays a list of available commands.

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 .gemini/ folder in the base of the repository. Detailed instructions can be found here.

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

1. Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩

[gemini-code-assist]

Code Review

This pull request introduces several significant improvements and bug fixes. Key changes include making PyAnnote embedding extraction more resilient to prevent crashes, adding robust memory-error handling to the Ollama classifier with low-VRAM and fallback model retries, and implementing a live transcription preview in the UI. The changes are well-tested and documented in the implementation plans. My review includes a few suggestions to improve the maintainability and robustness of the new classifier logic.

[gemini-code-assist]

The series of if statements to determine the required memory based on model size is not easily maintainable. If new model sizes are introduced, this logic will need to be modified. A data-driven approach would be more robust and easier to extend.

    def _estimate_required_memory_gb(self, model_name: str) -> Optional[int]:
        model_lower = model_name.lower()
        match = re.search(r"(\d+)\s*b", model_lower)
        if not match:
            return None
        try:
            size = int(match.group(1))
        except (TypeError, ValueError):
            return None

        # (size_in_b, required_ram_gb)
        ram_map = [
            (20, 16),
            (14, 12),
            (10, 10),
            (7, 8),
            (5, 6),
        ]
        for size_threshold, ram_required in ram_map:
            if size >= size_threshold:
                return ram_required
        return None

[gemini-code-assist]

The _estimate_total_memory_gb function uses a broad except Exception: pass block when trying to estimate memory with psutil. Silently ignoring all exceptions can hide underlying issues and make debugging difficult. It would be better to log the exception at a DEBUG level to aid in troubleshooting while still allowing the function to proceed to the next estimation method.

Suggested change

	except Exception:
	pass
	except Exception as exc:
	self.logger.debug("Failed to estimate memory with psutil: %s", exc)
	pass