Advanced Settings

Advanced Settings

Advanced configuration (#/advanced) is where every tunable model, generation, and QA knob lives — reached from Admin's Advanced configuration → card or Account's own pointer of the same name. Everything here persists on disk and survives server restarts; the page itself warns you're tuning "at your own risk."

The knobs are grouped into the same collapsible, side-nav-indexed accordion used by Model Manager: LLM sampling parameters, analyzer chunking & truncation, analyzer prompts & skills, analyzer models & endpoints, voice engine & device, voice batching & throughput, per-sentence QA gates, audio loudness targets, GPU arbitration & memory, Gemini rate limits, and LAN access & device tokens. High-risk groups (marked with a small warning glyph) start collapsed; the rest start open.

Accelerator profile

The Voice engine & device group is the one most people reach for first — it pins which GPU stack each engine runs on:

Accelerator profile is the headline knob: auto (default) detects your hardware and picks NVIDIA/CUDA, AMD/ROCm-DirectML, Apple/Metal, or CPU; pinning nvidia, amd, or cpu overrides that detection. Changing it is tagged rebuilds env for a reason — it's not instant. It rebuilds the Python virtual environment with a different torch / ONNX-runtime install and restarts the sidecar. Your books, cast, and voices are untouched.

Below it sit the per-engine device pins (Coqui, Kokoro, Qwen), the Qwen attention implementation, and the Coqui/Kokoro preload-at-startup toggles — each tagged restart (needs a sidecar restart, not a full env rebuild) or left plain where the change applies live. When the local analyzer is your active analyzer engine, a read-only Analyzer (Ollama) device row also appears at the end of this group — Ollama's device isn't app-pinnable, so it just reports what the daemon is currently doing and links to the documented OS-level steps to change it.

Everything else on the page

• Reset all (top-right) and a per-section Reset section button (once that section has at least one overridden value) both revert to shipped defaults, no partial-reset trap.
• Prompt-shaped knobs (the analyzer's system prompts, under Analyzer prompts & skills) render as an Edit / Revert to default pair instead of a form control — editing forks the prompt to your own copy; a Using your fork vs. Using shipped default chip always shows which one is active.
• A Restart sidecar banner appears the moment a sidecar-scoped change is pending, and a plainer amber banner appears for changes that need a full app restart instead. If CUDA_VISIBLE_DEVICES / CUDA_DEVICE_ORDER is set in server/.env, a banner explains that env var overrides every per-engine device pin above it, with a link to switch to per-engine pins.

Next: Account & Settings.