This Obsidian plugin enhances knowledge management by integrating a privacy-focused AI assistant, allowing users to chat with their notes using local or cloud-based LLMs. It targets Obsidian users seeking to leverage AI for deeper insights and efficient information retrieval from their personal knowledge base.

How It Works

The plugin employs a Retrieval Augmented Generation (RAG) pipeline. User notes are embedded into vectors and stored locally. When a query is made, relevant notes are retrieved based on vector similarity, and these notes are used to inform the LLM's response. This approach ensures data privacy by enabling offline operation with local models via Ollama and provides traceable answers by referencing the source notes.

Quick Start & Requirements

• Install via Obsidian's community plugin browser.
• Requires Ollama for local LLM execution or API keys for cloud models (e.g., OpenAI).
• Recommended: Python 3.10+ for backend components if self-hosting.
• See Architecture Wiki for more details.

Highlighted Details

• Supports RAG with local LLMs via Ollama and cloud models like ChatGPT.
• Provides direct reference links to source notes within AI-generated answers.
• Offers customizable chat views ('comfy' and 'compact') and chat saving.
• Utilizes Langchain and Orama for its RAG pipeline, including hierarchical tree summarization.

Maintenance & Community

The project originated from a university project but remains under active development by the original team in their spare time. Suggestions and contributions are welcomed. Issue reporting and feature requests can be made via GitHub.

Licensing & Compatibility

The plugin is free and open-source. The README does not explicitly state a license, but the open-source nature suggests permissive usage. Compatibility with commercial or closed-source applications is not specified.

Limitations & Caveats

Performance and response quality are dependent on the chosen LLM and the user's note organization. AI may generate incorrect or irrelevant answers, requiring query rephrasing or more detailed context. Obsidian Sync users should exclude the vector store directory to avoid large sync sizes.