Boris 1.2.0
Stable Boris 1.2 for Windows 10 and 11 x64. This release promotes beta.3
with a numeric version and adds an MSI installer. The app features are the
same as beta.3.
What's new in 1.2
- Teach Boris your wake word with four samples. The optional speaker filter
helps reject wakes from a TV, translation audio, or Boris's own speech.
Audio cleanup improves the signal used for wake detection and transcription. - Interrupt Boris by saying the wake word while it speaks, thinks, or asks for
confirmation. Confirmation prompts recover from device changes and re-ask
when a reply is unclear. - A local SQLite memory store keeps evidence for facts, preferences, and
projects. Memory search and explicit forgetting use those records. - Type exact values, masked secrets, large pastes, or numbered choices into
Home or the overlay when speech is a poor fit. - Follow a voice turn through the presence orb, live transcription previews,
tool progress, and streamed captions. Reduced-motion settings keep the orb
static. - Agent playbooks cover coding, debugging, explanations, design, review,
writing, mentoring, and skill creation. Context handling preserves recent
turns and separates trusted instructions from retrieved data.
Install and update
- Download
Boris_1.2.0_x64-setup.exeor
Boris_1.2.0_x64_en-US.msifrom this release. - Run the installer over your existing Boris installation. Your
%USERPROFILE%\.borisdata stays outside the installation directory. - In Settings > General > Updates, select Stable to follow future
stable releases. First-time installs also need the model setup and an
OpenRouter API key.
The in-app Stable updater uses the signed NSIS installer. The MSI is available
for manual installation.
Memory migration
When upgrading from beta.1 or an older version, Boris migrates legacy memory
to its local SQLite store. The one-time refinement sends historical memory to
the LLM provider already configured for Boris. Boris retires original files
only after refinement and verification succeed. Set BORIS_MEMORY=0 before
the first launch to disable canonical memory and skip migration.
Upgrades from beta.2 or beta.3 keep the existing local store.
Configuration
Freeform voice capture defaults to 30 seconds and can be set from 15 to 180
seconds with BORIS_MAX_UTTERANCE_SECS. Set
BORIS_STT_PARTIALS=0 to disable live transcription previews.
Full changes: CHANGELOG.md.