AI Local Knowledge Base

中文

A RAG-based intelligent Q&A assistant powered by LangChain.js. Import Markdown documents to build a knowledge base. When users ask questions, the system automatically retrieves relevant content and generates answers via AI models.

• Default embedding model: bge-large-zh-v1.5
• Default general model: Qwen2.5-0.5B-Instruct
• When the browser has no model cache, it will download from HF-Mirror by default and cache it
• Once cached, it no longer depends on the network. You can integrate with Ollama locally deployed general models for complete local deployment

Recommended to use this extension in the EasyEDA Pro client Local AI requires good computer performance and memory. It is recommended to use 1.5B or larger models for a better experience

Features

✅ Support connecting to general model API, with local vector model enabling intelligent Q&A based on knowledge base content

✅ Support local vector model for fast keyword search of knowledge base content

✅ Support local general model + vector model for knowledge base Q&A

✅ Integrate all documents from prodocs.lceda.cn, with instant Q&A for any issues encountered

Installation

Usage

1. Import the eext-knowledge-base.eext extension file in "Advanced" → "Extension Manager".

2. Enable "Allow External Interaction" in the configuration.

3. Open a schematic or PCB view, click "Advanced" → "AI Knowledge Base" in the top navigation bar and select the desired function.

Acknowledgements

• LangChain.js — RAG framework (MIT)
• Transformers.js — Browser-side model inference (Apache-2.0)
• bge-large-zh-v1.5 — Chinese embedding model (MIT)
• HF-Mirror — Hugging Face mirror for China