Reference

Configuration

Three layers, highest first: .env (secrets only), config/config.yaml (everything else, hot-reloaded on save), config/settings.py (fallbacks). The values below are what 0.1.4 ships with.

Character and her rest

character.name
Nexus — the one place her name lives
character.creator
how she refers to you; empty = "your creator"
rest.enabled
true
rest.minutes
20 — every entrance closes, what people say is not remembered
rest.press.needed
2 reviewers, within rest.press.window_s = 300
rest.forget.window_minutes
30 — the lines she may ask to forget

The model

ai.model.endpoint
http://127.0.0.1:11434/v1 (Ollama)
ai.model.model_name
gemma4:12b
ai.model.max_tokens
1024 for the visible answer
ai.model.temperature / top_p
0.8 / 0.95
ai.model.thinking_budget_tokens
1024 — a cap while thinking; local servers only
ai.profiles
named models you can switch to mid-conversation: /model use NAME
ai.keep_warm_s
240 — keeps a local server from unloading the model

Voice

voice.tts.provider
gpu-or-cpu — the GPU voice when its server is installed, else CPU
voice.tts.remote.endpoint
http://127.0.0.1:9820
voice.stt.provider
sherpa — words appear while a person is still talking
voice.stt.whisper.model
base — the fallback recogniser

Memory and privacy

memory.enabled
true
memory.dream
true — the nightly pass: writes the day, keeps, lets go, leaves a note
embedding.provider / model
local / all-MiniLM-L6-v2
privacy.forget_mode
her_choice | long_term | purge

Discord, the Hub, the panel

discord.enabled
true — false runs her without Discord at all
hub.host / hub.port
127.0.0.1 / 8100 — any other address needs AURA_HUB_KEY (8+)
web.host / web.port
127.0.0.1 / 8081 — 0.0.0.0 needs AURA_API_KEY (16+)
monitor.host / monitor.port
127.0.0.1 / 9876 — local only, no key
activity.ws_host / ws_port
127.0.0.1 / 9700 — another machine needs AURA_ACTIVITY_KEY

Ports

9700
activity SDK, WebSocket (games and voice clients)
9876
monitor API, HTTP, local only
9877
activity health
8081
web panel / chat entrance
8100
the Hub, in a browser
9820
the GPU voice server (tools/tts_mlx)
8001
VTube Studio API

Values taken from aura/config/config.yaml at the 0.1.4-preview commit. The long version, with every option and its comment, is CONFIG_GUIDE.md in the repository.