KdaiP/EchoBot

What it solves

EchoBot addresses the common trade-off in AI agents where loading extensive toolsets and skill systems increases token usage and slows down response times. It also prevents "persona dilution," where the AI's character personality is lost when mixed with complex technical instructions.

How it works

EchoBot uses a three-layer architecture to separate intent, personality, and execution:

1. Decision Layer: Uses a hybrid engine (rules + lightweight LLM) to quickly determine if a user wants to chat or needs a task performed.
2. Roleplay Layer: A dedicated layer for emotional value and immersion. It is stripped of tool-use metadata to ensure fast, vivid responses that stay in character.
3. Agent Core: A background layer that handles complex tasks using a full toolchain, skill libraries, and long-term memory. It has system-level permissions for local file read/write operations.

Who it’s for

Users seeking an immersive AI companion with a Live2D anime avatar who can also function as a productive assistant for tasks like coding and file management.

Highlights

• Live2D Integration: Supports web-based Live2D rendering with real-time voice interaction.
• Multi-Platform Support: Works via a built-in WebUI, QQ, and Telegram.
• Productivity Tools: Capable of local file manipulation, scheduled (Cron) tasks, and periodic (Heartbeat) tasks.
• Flexible Voice Options: Built-in TTS (edge-tts, kokoro) and ASR (Sensevoice) with support for custom OpenAI-compatible voice APIs.
• Visual Capabilities: Supports image upload and understanding if the connected LLM supports vision.