NotEMD: An Intelligent Knowledge Base Construction Tool Based on Large Language Models

NotEMD is an open-source Obsidian plugin that integrates multiple large language models such as OpenAI GPT and Anthropic Claude. It enables automatic note processing, concept link generation, thematic note creation, and intelligent web research. Its goal is to address the pain points of manual organization in traditional knowledge management, transforming knowledge management from manual labor into intelligent collaboration, and providing users with an intelligent personal knowledge management solution.

In the era of information explosion, knowledge workers face challenges in effectively collecting and organizing knowledge: traditional notes either pile up information or require a lot of manual organization. Obsidian is favored for its local-first approach and bidirectional links, but building a high-quality knowledge base still requires manual work such as identifying concepts and establishing links. The emergence of NotEMD uses the capabilities of large language models to automate this process and achieve intelligent collaboration.

NotEMD deeply integrates the Obsidian API, monitors editing events, and generates native wiki links to ensure compatibility. It uses chunk processing, caching mechanisms, incremental updates, and batch processing to optimize LLM call costs and response times. Adhering to the local-first concept, it only sends fragments that need processing, supports locally deployed models, and allows users to control the scope of automatic processing to protect privacy.

1. Automatic Concept Recognition and Linking: Identifies entities in notes and suggests or automatically adds bidirectional links; 2. Automatic Concept Note Generation: Generates notes for new concepts that include definitions, related concepts, and citation sources; 3. Intelligent Web Research: Generates search queries, retrieves resources, summarizes information, and provides citation suggestions; 4. Note Enhancement and Reconstruction: Summary generation, title optimization, structural reorganization, tag suggestions, etc.

Applicable to multiple scenarios: academic research (automatically identifies terminology in literature and builds association networks), technical learning (constructs concept maps and organizes knowledge systems), project management (unifies terminology and precipitates knowledge assets), creative writing (manages creative elements and assists in data organization), helping users improve knowledge management efficiency.

Comparison: Compared with traditional note-taking software (such as Evernote), it is more flexible with local storage, open Markdown format, bidirectional links plus AI assistance; compared with Notion AI, it deeply integrates the Obsidian ecosystem, is open-source and free, and supports multiple LLMs to avoid vendor lock-in. Limitations: The accuracy of concept recognition needs improvement; Chinese processing is not as mature as English; performance for large-scale knowledge bases needs optimization; there is a certain learning curve.

Future Directions: More intelligent concept disambiguation, multimodal support, collaboration features, and personalized learning optimization. Suggestions: Users already using Obsidian can try this plugin; users looking for knowledge management tools can consider the Obsidian + NotEMD combination, using AI to enhance cognitive abilities rather than replace thinking.