Luna's documentation

Everything the README only touches on, one subject per page.

• Install: the installer, LM Studio or llama.cpp, the voice server, first run
• Configuration: config.json, changing settings live, logging
• Controls and voice: keys, voice modes, speech detection, barge-in, wake word
• Tool calling: every tool, turning them off, MCP servers
• Files and commands: reading, writing with approval, allow-all, run_command
• Web: search, research, reading pages
• Email: checking, reading and searching your inbox, sending with your OK
• Vision: looking at your screen, the webcam
• Reminders and alarms
• Knowledge: your notes, manuals and PDFs as something she can look things up in
• Memory: facts, recall by meaning, the memory manager, conversation history
• Her reasoning: thought log, token confidence, adaptive thinking, introspection
• Learning from herself: lessons, episodes, fact cleanup, calibration
• Moods: the moods file, intensity, quick feelings she shows for a moment
• Portrait: Live2D and VRM, looking and thinking
• Textures: new clothes, irises and skin for her 3D model from UV templates, through ComfyUI
• Gestures: how her body moves, and writing your own
• Home automation: Home Assistant on your own network, setup, what always asks, local devices
• Sysadmin: looking after your machines, scheduled checks, other boxes over SSH, from your phone
• Plugins: how they work, writing your own, scheduled jobs and Discord
• Stream chat: pomf.tv, Twitch, YouTube and IRC, and what viewers can and can't reach
• Games: Minecraft, Zork, Vibe City, playing on her own
• Home and devices: Home Assistant, Lovense, toys of most brands through Buttplug.io
• Development: project structure, extras

---

See inside her

Live at 127.0.0.1:8792: the turn as it happens, newest on top, coloured by confidence.

Thoughts at 127.0.0.1:8792/thoughts/: every turn kept, searchable, filterable by flags.

Memory at 127.0.0.1:8792/memory/: the facts she keeps about you, to edit, retire or delete.

---

Everything she does

• Local speech recognition with Silero voice detection
• Local language model with tool calling
• Local text-to-speech via a shared Kokoro server
• Streaming replies: she talks while she's still thinking
• Barge-in: interrupt her mid-sentence
• Optional wake word, so open mode needs no keypress
• Optional vision: she can read what's on your screen, and see through your webcam when you switch it on
• Three push-to-talk modes, including hands free
• Configurable AI personality, and moods that shift with the clock and the session
• Long-term memory the model writes, corrects and forgets
• Permanent SQLite fact store that retires what stops being true, with a browser editor
• Reminders in plain language, with repeats, retry and DST-safe schedules
• Alarms that ring, nag and snooze until you're actually up
• Writes and edits your files, with every change approved on screen first
• Runs shell commands and opens apps, each one approved first
• Read, write and execute permissions you set from the tools pane: allow, ask or off
• A tools pane showing what each schema costs, and switches to turn them off
• Web search the model reaches for on its own
• Scheduled jobs she runs on her own, approved once, with writes fenced to folders you pick
• Remote access from Discord for the owner only, off until you switch it on, with approvals answered right in the chat
• Plays Minecraft: joins your world, works towards a goal, asks when something's unclear
• Plays Zork I, the classic text adventure, on her own or with you calling the moves
• Plays mayor in Vibe City, building the town in your open game window while you watch
• Runs your smart home through a Home Assistant on your own network, locks and garage doors always asking first
• Searchable archive of every conversation, which pruning never deletes
• Her reasoning kept per turn, with token confidence, review flags and a browser viewer
• Adaptive thinking: when an answer comes out shaky, she asks herself again and goes with what most answers agree on
• A grounding check: specifics she states with no tool, memory or message behind them get flagged, and said with a hedge
• Lessons from her own mistakes, written while she's idle and followed from then on
• Notes on every conversation, and a greeting that picks up where you left off
• Recipes: the steps of jobs that went right, saved and brought back when a similar job comes up
• Ongoing projects that carry across sessions: what was tried, what was found, what's next
• A live situation block each turn: your focused window, music, machine load, what's running
• Duplicate and contradicting facts merged or retired, never deleted
• Says when she's unsure, measured from her own token probabilities
• Checks every script she writes for syntax errors, without running it
• Researches properly: reads the top pages in one go and keeps the parts that answer the question
• Follow-through: when she says she'll check something and stops, she's made to actually do it
• An animated portrait that looks at your terminal and visibly thinks
• Rotating debug log that records decisions, not just errors
• Every setting changeable from the terminal and saved, most without a restart
• Diagnostics for the things that fail quietly: tool selection, prompt size, scrolling
• Themeable full-screen terminal interface
• Cross-platform architecture