Dashboard

AI Status: Idle
Twitch: Disconnected
Discord: Disconnected
VTube Software: Disconnected
OBS: Disconnected
Performance: N/A

Real-time Activity

Chat Messages

  • Waiting for chat messages...

AI Responses / Thoughts

Waiting for AI activity...
Voice Activity: Idle
Live Transcription: ...

Screen Perception (Preview)

Screen preview updates periodically based on Vision Model settings.

Configuration

Personality Studio

The name your AI Vtuber will use.
Provide context for the AI's personality and knowledge.
Keywords describing the core personality.
The main instruction set for the LLM. Advanced users can use templates. Check documentation for variables.
Version: 1.0 | Manage Versions | Templating Help
Helps steer the AI away from undesirable outputs.
Few-shot examples to guide the AI's tone, style, and response format.

Model Selection & Configuration

Large Language Model (LLM)

Choose the source for the primary language model.
Specific model identifier for the selected provider/type. For local models, this might be a name (Ollama) or file path (Llama.cpp).
Required for Ollama, Llama.cpp (if server running), or Custom API.

Advanced LLM parameters (temperature, top_p, repetition penalty, quantization settings, adapter paths) can be configured under 'Advanced Settings' or via the configuration file.

Vision Model (for Screen Perception)

Select the model for understanding screen content.
Specific model identifier for the selected vision provider.

Voice & Emotion Settings

Text-to-Speech (TTS)

Choose the engine to generate the AI's voice.
Enter the specific voice identifier or path to a local voice model file. Voice selection dropdown will be populated based on the chosen engine where possible (Future Feature).
1.0
1.0
Configure how AI emotions translate to TTS prosody and avatar expressions (Future Feature).

Speech-to-Text (STT)

Choose the engine to transcribe your voice input.
Identifier or path for the selected STT engine/model.
Specify language code (like 'en', 'ja') or 'auto' if supported by the engine/model.
Helps detect when speech starts and stops to reduce processing and latency. Recommended.

Integration Settings

Twitch

Status: Disconnected

Discord

Status: Disconnected

Vtubing Software

Default: VTube Studio=8001, Warudo=19190. Adjust if changed.
Status: Disconnected

OBS Studio

Requires obs-websocket plugin v5.x or later installed in OBS.
Status: Disconnected

Alert Platforms (Webhooks)

Allows AI to react to follows, subs, etc. Requires setup in Streamlabs.
Allows AI to react to events. Requires setup in StreamElements.

API Keys & Authentication

Enter API keys required by selected models or integrations. Keys are stored securely using the OS credential manager (e.g., Windows Credential Manager, macOS Keychain) and are not saved directly in configuration files. Fields will appear blank after saving.

Required for sending messages. Generate via sites like twitchapps.com/tmi/ (ensure 'chat:read' and 'chat:edit' scopes are included). Use with caution.
Create a bot application in the Discord Developer Portal.

Note: Keys for external tools (Web Search, Weather) might be configured under the 'Actions' tab if required by the tool.

Memory System (Long-Term)

Determines how the AI remembers past interactions long-term.

Vector Database Settings (if enabled)

Path for local DB (will be created if it doesn't exist) or URL for remote.
Model used to create embeddings for memory storage/retrieval (local via Ollama/SentenceTransformers or API). Affects quality and resource usage.
How text is divided before embedding (Future Feature).
Max number of past interactions/facts to fetch for context (RAG).
Use Maximal Marginal Relevance (MMR) or similar technique to diversify retrieved memories, improving relevance.
Helps condense information and manage database size. Requires an LLM call. Leave blank to disable.

Action & Tool Configuration

Enable and configure actions the AI can perform beyond speaking. Available actions depend on enabled integrations and plugins. Permissions should be reviewed carefully.

Available Actions:

(Checking integrations and plugins)

More actions will appear here as they are developed or added via plugins. Configuration and permission settings for actions are planned.

Screen Perception Setup

Configure how the AI perceives and understands content on your screen. Requires a Vision Model to be selected in the 'Models' tab.

How often to capture the screen for analysis (higher frequency uses more resources).
Choose which part of the screen the AI should see (Future Feature).
Limit perception to specific applications (Future Feature).
Advanced configuration for the vision model (e.g., OCR language hints) (Future Feature).

Advanced Settings

Influences internal trade-offs between speed, resource usage, and response quality/complexity (e.g., model choices, queue priorities).
Controls the verbosity of logs shown in the 'Application Logs' section below and saved to file.
Display intermediate steps like retrieved memory, chosen actions, or function calls in the AI Response view (if supported by the pipeline).
-1 for auto/max (uses available VRAM), 0 for CPU only. Adjust based on your GPU VRAM. Requires GPU support compiled/installed.
Controls randomness in LLM responses. Lower values = more deterministic, higher = more creative. (Overrides default if set).
Nucleus sampling parameter. Considers only the most probable tokens with cumulative probability P. (Overrides default if set).

More advanced settings may be available via direct editing of the configuration file (e.g., `config.json` or `config.yaml`). Refer to documentation.

Application Logs

Application logs will appear here...