There is a lot of content here. You can bookmark it and copy the full text to your Claude Code or Cursor to have it help you with the operations. You can also read it slowly when you have time to learn something new!
You might have encountered these problems:
You see a great post on Twitter and copy-paste it. When you want to find it later to study, you have to scroll manually for ages.
You have hundreds of notes stored in Notion, but they are isolated from each other. You have no idea if a certain concept has appeared elsewhere.
Every time you ask an AI a question, it starts searching from scratch and pieces together a temporary answer. There is no accumulation, no memory, and half your tokens are wasted.
More importantly, your notes are stored on someone else's server. If the service shuts down one day, your data is gone.
Okay, the system I built solves these problems.
Four core advantages:
- Fully Automated: You don't need to manually organize notes; the AI does it for you.
- Local Storage: The data always belongs to you and will not be uploaded to any server.
- Persistent Accumulation: Knowledge continues to accumulate rather than starting from zero every time.
- Just Ask: You only need to ask questions and explore; leave the rest to the AI.
Simply put, you give documents to the system, and it automatically organizes them into a structured knowledge network. You can browse freely using bidirectional links.
The entire chain is: Document Import → AI Organization → Wiki Generation → Bidirectional Linking.
The whole process requires no manual operation of any graphical interface; just drop the files in, and they automatically form a network.
Three Tools, Each with a Role
This system consists of two main tools.
Obsidian: The Note Display Layer
Obsidian is a local bidirectional link note-taking tool. It is completely free and cross-platform for Windows, Mac, and Linux. Its core feature is bidirectional linking.
What is a bidirectional link?
By typing double brackets in a note, such as [[Claude-Code-Notes]], Obsidian automatically turns it into a purple link. If the note "Claude-Code-Notes" exists, clicking it will jump there. If it doesn't exist, clicking it will create it. This is how bidirectional links work—it's very simple. You don't need to maintain it manually; Obsidian builds the relationships for you.
The problem with traditional note-taking software is that notes are isolated. You wouldn't know if the word "Apple" appeared anywhere else. Obsidian's Graph View can visualize all notes and their link relationships into a map. You can see the knowledge structure at a glance, identifying which nodes are islands and which are hubs.
Additionally, Obsidian is completely free for personal use with no restrictions. All data is local.
Hermes Agent: The Automated Execution Engine
Hermes Agent is an autonomous AI agent developed by Nous Research. Its biggest feature is a built-in learning loop that can create and improve skills from experience. In this knowledge management workflow, Hermes acts as the automated execution engine. It has a built-in llm-wiki skill that can operate the knowledge base directly according to LLM Wiki file structure specifications.
What does this mean? You don't need to manually create folders, organize notes, or add links. You just tell Hermes to "write this article into the knowledge base," and Hermes will automatically:
- Extract key entities (people, tools, projects)
- Extract core concepts (methodologies, technical principles)
- Create structured Markdown files
- Add bidirectional links to related concepts
- Update the knowledge base index
One important rule: Hermes only operates on the knowledge base when you explicitly ask it to.
- When you say "write to knowledge base" or "import to knowledge base," Hermes executes the import.
- When you say "combine with knowledge base" or "search the knowledge base," Hermes retrieves information.
In ordinary daily conversations, Hermes will not actively touch your knowledge base, ensuring it isn't polluted by irrelevant chats.
LLM Wiki: The Knowledge Base Standard
LLM Wiki is not a standalone app but a set of knowledge base file structure specifications. It defines how to organize knowledge. The core idea is to let the AI incrementally build a persistent Wiki.
What does persistent mean? When you import a document, the system doesn't just index it and finish. It truly understands the document, extracts key entities, concepts, and relationships, and then generates or updates corresponding Wiki pages. These pages are saved locally. As you import more documents, the Wiki becomes richer. Pages form references and associations, and contradictions are flagged. When you ask questions later, the AI answers based on the structured Wiki rather than piecing together raw documents, and it will cite sources.
Full Workflow Demonstration
Let's look at the tools together. The entry point for the entire workflow is Hermes Agent.
Step 1: Give Instructions
For example, you say "Write this article about AI novel writing into the knowledge base."
Step 2: Hermes Automatically Organizes
Hermes uses the llm-wiki skill to read the content, extract entities and concepts, create Markdown files, add links, and update the index.
Step 3: File Structure Generation
Hermes creates files according to LLM Wiki specs, including metadata, core content, and links.
Step 4: Obsidian Displays the Knowledge Network
Open the knowledge base directory and drag it into Obsidian as a Vault. You now have an AI-organized knowledge base where you can browse the network and see link strength in Graph View.
Step 5: Infinite Loop
This process repeats. Every new document updates the network, supplements existing pages, and links related concepts across different documents.
Installation Steps
Step 1: Install Obsidian
Download the macOS version from obsidian.md. Create or select a Vault (a folder for your Markdown files).
Step 2: Install LLM Wiki
Visit nashsu/llm-wiki on GitHub and download the latest release. Open the app, create a new project, and enter your API Key in the settings. It supports OpenAI, Claude, Minimax, and any OpenAI-compatible API.
Step 3: Install Hermes Agent
Open your terminal and run:
Then reload your config: source ~/.zshrc. Run hermes setup to configure your model provider.
Configure Knowledge Base Rules (Important):
Tell Hermes your rules:
"My knowledge base directory is: /Users/username/Documents/knowledge_base. Rules: 1. Only write when I say 'write to knowledge base'. 2. Only retrieve when I say 'combine with knowledge base'. 3. Use Markdown and [[bidirectional links]]. 4. Use LLM Wiki structure."
Usage Rules
- Say "Write to knowledge base" for Hermes to organize new documents.
- Say "Combine with knowledge base" for Hermes to retrieve and answer based on your data.
- Obsidian is always available to browse, edit, and visualize your Markdown files.
Summary
Hermes Agent is the execution engine, and Obsidian is the display layer. Use this workflow to let AI handle the tedious work of knowledge management so you can focus on exploration.
