Local LLM for transcript cleanup, smart task extraction, and task breakdown. Runs fully offline after the model is downloaded.
-
Model Tier
+
Model Tier
{#each LLM_MODELS as model}
-
+
{#if llmModelLoaded(model.id)}
Loaded
{:else if llmModelDownloaded(model.id)}
@@ -1858,15 +1858,15 @@
{LLM_MODELS.find((model) => model.id === selectedLlmModelId())?.subtitle || "Local model"}
-
{llmStatus}
+
{llmStatus}
{#if llmTestHint}
-
{llmTestHint}
+
{llmTestHint}
{/if}
{#if llmDownloadingModel === selectedLlmModelId()}
-
+
{llmDownloadProgress}%
{#if llmDownloadTotal > 0}
· {formatBytes(llmDownloadBytes)} / {formatBytes(llmDownloadTotal)}
@@ -1922,7 +1922,7 @@
-
+
Recommended for this machine:
{LLM_MODELS.find((model) => model.id === (settings.llmModelId || "qwen3_5_4b"))?.subtitle || "Qwen3.5 4B"}
@@ -1955,12 +1955,12 @@
structure of LLM cleanup output; composes on top of the
active profile's initial prompt. -->
-
Cleanup preset
+
Cleanup preset
-
+
{#if settings.llmPromptPreset === "email"}
Formats output as an email paragraph, tight sentences, no markdown, no auto-added greeting or signoff.
{:else if settings.llmPromptPreset === "notes"}
@@ -1979,12 +1979,12 @@
other; adds reload latency between transcribe and
cleanup phases. -->
-
GPU concurrency
+
GPU concurrency
-
+
{#if settings.aiGpuConcurrency === "sequential"}
On tight-VRAM cards (≤6 GB), loading Whisper + LLM together OOMs. Sequential mode frees the other model before loading; adds a small reload pause between transcribe and cleanup.
{:else}
@@ -2026,7 +2026,7 @@
open={searchActive ? searchMatches('Rituals morning triage wind-down launch login autostart') : false}
>
-
+
All off by default. Rituals only appear when you ask for them.
@@ -2038,7 +2038,7 @@
{#if settings.ritualsMorning}
-
@@ -2093,7 +2093,7 @@
open={searchActive ? searchMatches('Tasks page gamification visuals header sparkline') : false}
>
-
+
How the Tasks page surfaces progress. Always additive.
@@ -2113,7 +2113,7 @@
open={searchActive ? searchMatches('Nudges soft notifications hourly') : false}
>
-
+
Gentle, anticipatory reminders. Capped at 3 per hour. Never fires while you're looking at Magnotia.
@@ -2158,12 +2158,12 @@
open={searchActive ? searchMatches('Read aloud TTS voice rate system') : false}
>
-
+
Uses your operating system's built-in voices. No audio leaves the machine.
-
Voice
+
Voice
-
+
Rate · {settings.ttsRate.toFixed(1)}×
-
+
Slower
Normal
Faster
@@ -2242,7 +2242,7 @@
/>
{#if settings.meetingAutoCapture}
-
Apps to watch (comma-separated)
+
Apps to watch (comma-separated)
{#if settings.soundCues}
-
Volume
+
Volume
-
{Math.round(settings.soundCueVolume * 100)}%
+
{Math.round(settings.soundCueVolume * 100)}%