Configuration
Configure dialogue models, task models, voice providers, proxy settings, and personality profiles.
Most day-to-day settings are managed from the desktop app settings panel. Those settings persist model and provider choices, while environment files provide defaults and secrets for local development.
Environment Resolution
The Electron main process can load environment files from common project locations:
apps/desktop/.env
.env
<current-working-directory>/.envLater settings from the app UI should override defaults where the runtime supports persisted configuration.
Model Groups
Noema separates conversational replies from task execution. Keep these model groups independent so the dialogue layer can stay fast while the task runtime uses a stronger model when needed.
| Group | Used By | Notes |
|---|---|---|
LLM_* | Dialogue layer | Optimized for natural replies and relationship continuity. |
TASK_* | Task runtime | Optimized for planning, tool calls, and verification. |
ASR_* | Voice input | Converts microphone input to text. |
TTS_* | Voice output | Synthesizes assistant speech. |
Personality Profiles
Personality profiles define the character, relationship, speaking style, values, and behavior rules used by the dialogue layer.
{
"name": "ECHO",
"style": "concise, emotionally coherent",
"memoryPolicy": "use relevant long-term memory only",
"taskPolicy": "delegate concrete work to the runtime"
}Proxy
Set PROXY_URL when provider traffic needs to route through a local or remote proxy:
PROXY_URL=http://127.0.0.1:7890