This project provides a prototype retrieval-augmented generation (RAG) assistant for Obsidian users, designed to streamline writing and thinking by automatically drafting sections based on relevant notes. It targets users who leverage Obsidian for knowledge management and seek to enhance their content creation workflow.

How It Works

The system retrieves relevant documents from an Obsidian vault using both keyword search via OpenSearch and semantic search. This retrieved context is then fed into a language model to generate paragraphs for the specified section, improving content relevance and reducing manual research.

Quick Start & Requirements

• Install via git clone and setting OBSIDIAN_PATH and TRANSFORMER_CACHE environment variables.
• Requires Docker for OpenSearch, Python, and Hugging Face Hub cache.
• Build artifacts and run the retrieval app using make build-artifacts and make run.
• Install the Obsidian copilot plugin and configure API keys.

Highlighted Details

• Supports drafting sections based on notes and reflecting on weekly journals.
• Utilizes both keyword (OpenSearch) and semantic search for document retrieval.
• Integrates with Obsidian via a community plugin.

Maintenance & Community

• The project is a prototype with a single primary contributor.
• Contribution guidelines include setting up pre-commit hooks.
• TODOs include adding support for Anthropic Claude and assessing full document context.

Licensing & Compatibility

• The repository does not explicitly state a license in the provided README.

Limitations & Caveats

• This is a prototype implementation, and stability or feature completeness is not guaranteed.
• Future support for models like Anthropic Claude is listed as a TODO.