swarmvault

Compile sources, code, and research into a local markdown knowledge vault with graph and search.

30.5k|3.5k|Updated Jul 4, 2025
One-click install
npx skills add https://github.com/davila7/claude-code-templates --skill swarmvault
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: swarmvault
Source: https://github.com/davila7/claude-code-templates/tree/main/cli-tool/components/skills/development/swarmvault
Command: npx skills add https://github.com/davila7/claude-code-templates --skill swarmvault

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires @swarmvaultai/cli, and includes references (resource) components.

What problem does it solve?

Knowledge work scattered across books, transcripts, exports, datasets, and codebases often disappears into chat history. SwarmVault keeps a durable, local-first wiki on disk—built on the LLM Wiki three-layer pattern (raw sources, wiki, schema)—so compiled pages, graphs, dashboards, and review artifacts stay inspectable and reusable.

Core Features & Use Cases

  • Broad Ingest: Parse code via tree-sitter (JS/TS, Python, Go, Rust, and more) plus non-code formats like PDF, DOCX, EPUB, XLSX, PPTX, SRT/VTT transcripts, Slack exports, email, calendars, audio, YouTube URLs, and images.
  • Compile, Query, Review: Generate wiki pages, graph artifacts, and local search with compile, query, explore, and staged approval flows via review and candidate commands.
  • Graph & Sharing: Run reverse-import impact analysis with graph blast, serve a live workspace, export HTML/Obsidian/JSON reports, or expose the vault over MCP.
  • Use Case: Point it at a repository and a folder of research PDFs, run swarmvault compile, then ask "What is the auth flow?" and get a saved, durable answer page alongside a navigable graph.

Quick Start

Ask the assistant to initialize a SwarmVault workspace in the current project, ingest your sources, compile the vault, and answer your first question from the generated wiki.

Frequently Asked Questions about swarmvault

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I set up a local knowledge vault for my repo?▼

Install the CLI with npm install -g @swarmvaultai/cli, then run swarmvault init in your project and swarmvault ingest ./src --repo-root . followed by swarmvault compile. The generated wiki/, state/graph.json, and wiki/graph/report.md give you module pages and a graph overview.

What file formats can SwarmVault ingest?▼

It ingests code in over 30 languages via tree-sitter, plus PDF, Word, Excel, PowerPoint, EPUB, CSV/TSV, Jupyter notebooks, BibTeX, Org-mode, SRT/VTT transcripts, Slack exports, email, calendar files, images, audio files, and YouTube transcript URLs.

Does SwarmVault work offline without an API key?▼

Yes, the default heuristic provider is a valid local/offline starting point, and code parsing always runs locally via tree-sitter. For richer synthesis, embeddings, or audio transcription, configure a provider such as Ollama with a local model like gemma4.

Why is audio transcription not appearing after ingest?▼

Audio ingest requires tasks.audioProvider in swarmvault.config.json to point at a provider with audio capability. Without it, the source is still ingested but an extraction warning is recorded instead of a transcript. YouTube transcript ingest does not need a provider.

Can I expose the vault to other AI agents or tools?▼

Yes, run swarmvault mcp to expose tools and resources for page search, reads, source listing, query, ingest, compile, and lint over MCP. You can also install agent-specific rules with swarmvault install --agent for Claude, Codex, Cursor, Aider, and others.