Quick Start EN

中文 | English

Prerequisite: installation is complete. See Installation Guide.

Configure an AI Provider

1. Open the "AI Settings" page → Add provider
2. Fill in four fields:
   • Protocol: choose one of OpenAI Compatible / Gemini / Claude (DeepSeek, Kimi, Qwen, Ollama, one-api and other relays all use OpenAI Compatible)
   • Base URL: e.g. https://api.deepseek.com/v1 (local models such as http://127.0.0.1:11434/v1)
   • API Key
   • Model: click "Fetch from API" to pull the model list from the endpoint directly, no manual typing needed
3. Click "Test" to verify connectivity, and save only after it passes
4. Click the play button on the entry to activate it

Rule: you can configure multiple providers, but only one is active at any time (mutually exclusive).

Start Chatting

Go back to the "Chat" page and type a message. Replies stream end to end (SSE + Tauri Event) and are displayed character by character, with Markdown rendering (bold, lists, tables, code blocks).

The AI can call tools out of the box: asking it to "list the files in the current directory" or "save this summary as a memory" executes directly. See Tool System for details.

UI Overview

Page	Purpose
Chat	Streaming chat, message queue, multimodal input
AI Settings	Provider management, connectivity testing
Tools	Tool registration, enable/disable, manual invocation
Skills	Reusable skills accumulated by the AI
Memory	View and manage the AI's long-term memory
Audit	Trail of all tool calls and key operations
Remote	HTTP API toggle and credentials

Recommended Settings

• Tool approval: sensitive tools pop up a confirmation card by default; if there is no response within 120 seconds the request is auto-rejected. The auto-approval scope can be adjusted in settings
• Message queue: messages typed in quick succession are queued automatically and sent one by one; drag with the mouse to reorder, double-click to remove
• Terminal mode: if you prefer the command line, use bit tui, see TUI Terminal Mode