Fix Linux GPU auto-detect fallback and avoid bad hipblas default#354
Open
SyedAmerGilani wants to merge 2 commits intoZackriya-Solutions:devtestfrom
Open
Fix Linux GPU auto-detect fallback and avoid bad hipblas default#354SyedAmerGilani wants to merge 2 commits intoZackriya-Solutions:devtestfrom
SyedAmerGilani wants to merge 2 commits intoZackriya-Solutions:devtestfrom
Conversation
## Description This PR delivers a focused update improving provider model management, fixing API key persistence issues, enhancing Whisper model support, and bundling FFmpeg at build time for a smoother first-run experience. Key highlights include: * Dynamic real-time model fetching from OpenAI, Anthropic, and Groq (when API key is provided) * Build-time FFmpeg bundling with 3-tier fallback (bundled → dev → system) * Support for Q5_1 quantized Whisper models (tiny, base, small) * Centralized provider API key management * Proper model and endpoint state preservation when switching providers * Multiple Whisper metadata corrections * Windows system audio device reset fix * Transcript copy newline preservation * UI improvements to model selection and notifications --- ## Related Issue Fixes Zackriya-Solutions#307 Fixes Zackriya-Solutions#346 Fixes Zackriya-Solutions#322 Fixes Zackriya-Solutions#323 (Also includes fixes reported by community members in related provider/model issues.) --- ## Type of Change * [x] Bug fix * [x] New feature * [ ] Documentation update * [x] Performance improvement * [x] Code refactoring * [ ] Other (please describe) --- ## Testing * [x] Unit tests added/updated * [x] Manual testing performed * [x] All tests pass Tested scenarios include: * Provider switching with API key persistence * Model selection preservation per provider * Dynamic model fetching with valid API keys * Custom OpenAI-compatible endpoint connection testing * Whisper model downloads and metadata validation * Q5_1 model loading and transcription * FFmpeg resolution (bundled/dev/system fallback) * Windows system audio device persistence * Transcript copy with preserved newlines --- ## Documentation * [x] Documentation updated * [ ] No documentation needed Updated: * Release notes --- ## Checklist * [x] Code follows project style * [x] Self-reviewed the code * [x] Added comments for complex code * [x] Updated README if needed * [x] Branch is up to date with devtest * [x] No merge conflicts --- ## Additional Notes * FFmpeg download source switched to Zackriya’s GitHub release for improved Windows reliability. * Model fetching now occurs dynamically when opening the model selector and an API key is present. * Provider-specific model selection and endpoints are now cached and automatically restored. * Includes contributions from: * @matbe (Windows audio fix, Whisper metadata corrections) * @5m4u66y (OpenAI-compatible API connection test fix)
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Description
This fixes Linux GPU detection order so AMD can still be detected when NVIDIA is present without CUDA.
It also avoids passing unsupported helper features to llama-helper.
For ROCm 6+ environments, auto-detect now skips hipblas because build will fail with it.
Manual override still works via TAURI_GPU_FEATURE=hipblas for local testing.
Related Issue
[Link to the issue this PR addresses (e.g., "Fixes #123")]
Type of Change
Testing
Documentation
Checklist
Screenshots (if applicable)
[Add screenshots here if your changes affect the UI]
Additional Notes
[Add any additional information that might be helpful for reviewers]