vaultspec-rag 0.2.23

GPU-accelerated semantic search for your docs and source code: hybrid dense + sparse RAG on a local Qdrant backend, served to Claude Code and other MCP clients.

vaultspec-rag - semantic search for a vaultspec-core workspace

vaultspec-rag is the semantic search companion to vaultspec-core. The RAG in the name stands for retrieval-augmented generation. That's the pattern of pulling relevant snippets out of your own files so an agent can answer with grounded context. It indexes the markdown documents in your vault and the source code that sits beside them, then lets you search them by meaning rather than by exact keyword match. Search by meaning closes the vocabulary gap. A query for "how do we handle file locks during indexing" finds a decision record about concurrent writes and per-root locks, even though it never uses the word "indexing."

Requirements

• Python 3.13 or newer
• uv as the package manager
• An NVIDIA GPU with CUDA support
• About 3 GB of free GPU memory
• Linux or Windows

macOS, AMD GPUs, and Apple Silicon are not supported. For the reasoning behind the hardware floor, see the architecture overview.

Quickstart

uv add vaultspec-rag
uv run vaultspec-rag install
uv sync
uv run vaultspec-rag index
uv run vaultspec-rag search "your question here"

install configures the GPU PyTorch build and provisions the search models and the managed Qdrant server. uv sync then fetches the GPU PyTorch build. The first run is slower because of one-time model downloads.

For repeat use, start the server-backed service first - see the getting started guide for the full walkthrough.

What's a vault?

A vault is a .vault/ directory of markdown files - research notes, architecture decision records, plans, and execution logs - that vaultspec-core creates and manages. If you don't have one yet, set one up there first.

vaultspec-rag adds exactly one capability on top of that: semantic search over the vault and the source code beside it. Vault creation, document templates, frontmatter validation, and the spec-driven workflow all stay in vaultspec-core.

Both packages live side by side in the same project. You can use vaultspec-core on its own without ever installing vaultspec-rag. vaultspec-rag without vaultspec-core has nothing to search.

Documentation

Getting started

• Getting started - install, index, and run your first query end to end.
• Installation - the GPU build, dependency provisioning, and recovery steps.

Daily use

• Search and index - run searches and refresh the index.
• Service mode - keep models warm in a background service for faster queries.
• Backends - the managed Qdrant server versus local-only mode.
• MCP integration - wire search into Claude Code and other MCP clients.
• Automation - JSON output and scripting.
• Preprocessing hooks - index PDFs, spreadsheets, and other formats.

Reference

• CLI reference - every command and flag.
• Configuration - settings, environment variables, and defaults.
• Glossary - terms used across the docs.

Concepts

• Architecture - how it works, why a GPU is required, and why the service is server-first.
• Indexing - indexing and retrieval internals.

Support and help

File bugs and ask questions on the GitHub issue tracker.

A useful bug report includes your vaultspec-rag version, your operating system, your GPU model, the exact command you ran, and the full stderr output.

What changed

See the changelog for release notes and version history.

License

vaultspec-rag is released under the MIT License. See LICENSE for the full text.