4791969bc9
Cached GGUF repos now order by the quant that will actually load instead of the row size_bytes, which sums every downloaded quant and pushed a repo holding one small quant behind larger models. Cached entries apply the same seen gate local rows use and skip keys are scoped to the backend load target, so a cached repo and an indexed local row aliasing the same files no longer spend two attempt slots on one failure. The fallback resolves candidates through the bounded worker pool and consumes them incrementally from a size-ordered pool after a short settle grace, so one slow folder scan no longer stalls the send path behind the transport timeout. |
||
|---|---|---|
| .. | ||
| backend | ||
| frontend | ||
| src-tauri | ||
| __init__.py | ||
| install_llama_prebuilt.py | ||
| install_node_prebuilt.py | ||
| install_python_stack.py | ||
| install_whisper_prebuilt.py | ||
| LICENSE.AGPL-3.0 | ||
| MCP.md | ||
| node_prebuilt_pins.json | ||
| package-lock.json | ||
| package.json | ||
| prebuilt_core.py | ||
| setup.bat | ||
| setup.ps1 | ||
| setup.sh | ||
| Unsloth_Studio_Colab.ipynb | ||