Conversation
- Introduced methods to check if models are cached locally in MLX and PyTorch backends. - Enhanced progress tracking during model loading to filter out non-download progress when models are cached. - Updated HFProgressTracker to conditionally report progress based on download status. - Added test scripts for monitoring SSE events during model downloads and verifying progress tracking functionality. - Improved overall error handling and logging for better debugging during model download processes.
- Introduced a new directory for manual test scripts aimed at debugging and validating backend functionality. - Added README.md detailing the purpose and usage of various test scripts, including tests for TTS generation, model downloads, and progress tracking. - Included an __init__.py file to define the test suite structure and provide context for the tests.
![cursor[bot]](https://avatars.githubusercontent.com/in/1210556?s=60&v=4){kind=small}
There was a problem hiding this comment.
Cursor Bugbot has reviewed your changes and found 2 potential issues.
Bugbot Autofix is OFF. To automatically fix reported issues with Cloud Agents, enable Autofix in the Cursor dashboard.
| if not is_cached: | ||
| progress_manager.mark_complete(model_name) | ||
| task_manager.complete_download(model_name) | ||
|
|
There was a problem hiding this comment.
Error handlers unconditionally track errors for untracked downloads
Medium Severity
The exception handlers call mark_error and error_download unconditionally, even when the model was cached and download tracking was never started. When is_cached is True, start_download and update_progress are skipped, but on error, mark_error still creates a new error entry in progress_manager and notifies SSE listeners. This creates inconsistent state between progress_manager (has error) and task_manager (no record). The error handling paths need the same if not is_cached conditional check that was added to the success path.
Additional Locations (2)
backend/backends/mlx_backend.py
Outdated
| from huggingface_hub import constants as hf_constants | ||
| model_path = self._get_model_path(model_size) | ||
| repo_cache = Path(hf_constants.HF_HUB_CACHE) / ("models--" + model_path.replace("/", "--")) | ||
| return repo_cache.exists() |
There was a problem hiding this comment.
Cache check doesn't verify model files exist inside directory
Medium Severity
The _is_model_cached methods only check if the cache directory exists via repo_cache.exists(), but don't verify that model files are actually present inside. This differs from the more robust check already in main.py which uses rglob to confirm files like *.safetensors or *.bin exist. If a download was interrupted, the directory may exist but be empty or incomplete. With the current check, is_cached returns True, causing filter_non_downloads=True, which suppresses download progress notifications even though a real download is occurring.
Additional Locations (1)
- Bumped version numbers for @voicebox/app, @voicebox/landing, @voicebox/tauri, and @voicebox/web to 0.1.11. - Added a new `useAutoUpdater` hook to check for app updates on startup and notify users with toast messages. - Enhanced `UpdateStatus` component to handle version retrieval errors more gracefully. - Updated dependencies in `package.json` for Tauri plugins to support new update functionalities.
- Added a step to install PyTorch with CUDA for Windows in the release workflow. - Updated model references in backend/main.py to use openai/whisper models instead of mlx-community for the MLX backend.
- Rearranged imports for consistency across components. - Enhanced the ModelManagement component to include detailed logging for download actions and errors. - Updated the ModelProgress component to connect to SSE only when actively downloading, preventing connection exhaustion. - Added a downloading state to the model status to indicate ongoing downloads. - Improved toast notifications for model downloads with completion and error callbacks. - Refactored the useModelDownloadToast hook to support new callbacks for download completion and error handling. - Updated backend model status to reflect downloading state during active downloads.
- Rearranged imports for consistency in useModelDownloadToast hook. - Improved logging in useModelDownloadToast for better debugging during download events. - Updated progress calculation to handle cases where progress exceeds 100%. - Enhanced toast notifications to reflect download completion and error states. - Introduced throttling in ProgressManager to optimize SSE updates and prevent overwhelming clients. - Added new test scripts for monitoring SSE events during model downloads, ensuring accurate progress reporting.
- Updated caching methods in MLX, PyTorch, and backend to ensure models are fully downloaded before being marked as cached. - Improved progress tracking to filter out non-download progress and provide accurate feedback during model downloads. - Enhanced HFProgressTracker to skip non-byte progress bars and ensure meaningful progress reporting. - Refactored progress initialization to provide immediate feedback while fetching metadata from HuggingFace. - Added error handling and logging for better debugging during cache checks and download processes.
Enhance model caching checks and progress tracking for downloads
1 participant
Note
Medium Risk
Touches model loading and progress/task tracking paths for both TTS and STT backends; incorrect cache detection or filtering could suppress real download progress or leave tasks untracked.
Overview
Prevents misleading "model downloading" UX when running generation against already-cached HuggingFace models.
Both
pytorch_backend.pyandmlx_backend.pynow detect whether the target TTS/Whisper model exists in the local HF cache and, if so, skip creating download tasks and skip emitting initial/complete download progress events; when not cached, behavior remains the same.HFProgressTrackergained an optionalfilter_non_downloadsmode to ignore non-download tqdm progress bars (e.g., internal generation "Segment 1/1"), and a set of manual debug scripts were added underbackend/tests/to validate SSE progress behavior and inspect progress/task state.Written by Cursor Bugbot for commit 0b17073. Configure here.