The last load attempt is not spent while folder scans are still pending, since a smaller candidate can still enter the pool; earlier attempts keep flowing incrementally. Resolution workers stop claiming scans once autoload reaches a terminal result, so a successful early load no longer leaves background scans contending with inference. Local rows now apply the picker's chat-only format gate (GGUF on any host, MLX on Mac) and the cascade skips cached non-GGUF repos on chat-only installs, matching what the picker offers; the remembered path stays ungated since a recorded load proves the model runs on this install. |
||
|---|---|---|
| .. | ||
| 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 | ||