Oboyu (覚ゆ)

What is Oboyu (覚ゆ)?

Oboyu (覚ゆ) is a local semantic search engine that indexes documents and enables natural language queries. It is designed for individuals and AI assistants, with best-in-class Japanese language support, and runs entirely on the user’s machine.

How to use Oboyu (覚ゆ)?

Install with pip install oboyu or uv tool install oboyu. Index documents with oboyu index ~/Documents, then search interactively with oboyu query --interactive. For AI assistant integration, start the MCP server with oboyu mcp and configure it in Claude Desktop’s settings.

Key features of Oboyu (覚ゆ)

• Hybrid search combining semantic understanding and keyword matching.
• Supports plain text, Markdown, PDFs, code files, and Jupyter notebooks.
• Automatic Japanese encoding detection (Shift-JIS, EUC-JP, UTF-8).
• ONNX acceleration for 2–4x faster performance.
• Built-in MCP server for Claude Desktop and AI coding assistants.
• Smart reranking and incremental indexing for speed.

Use cases of Oboyu (覚ゆ)

• Index and search academic research papers and notes.
• Search through project code documentation and comments.
• Build a personal knowledge base from local notes.
• Query mixed Japanese‑English documents seamlessly.

FAQ from Oboyu (覚ゆ)

What are the system requirements?

Python 3.10 or higher, macOS or Linux (Windows via WSL), 2GB RAM minimum, and 1GB free storage for models and index.

How does Oboyu differ from keyword search?

It uses semantic search to understand the meaning behind questions, not just exact terms, and offers hybrid modes for best results.

Where are documents and indexes stored?

Everything runs locally on your machine; documents never leave your device. Models are downloaded automatically on first use (~90MB).

Does Oboyu support MCP transport?

Yes, it runs an MCP server via oboyu mcp and integrates with Claude Desktop, Cursor, and other AI assistants.

What file formats are supported?

Plain text, Markdown, code files, PDF, Jupyter notebooks, and more, with automatic encoding handling.