v0.1.0 — First public release

First public release of archiver-rag — agent-agnostic semantic memory for your Obsidian vault.

Install

pipx install archiver-rag
archiver-rag init

📦 PyPI: https://pypi.org/project/archiver-rag/
📖 Docs: https://fernandojrr.github.io/archiver-rag/

What's included

Retrieval

• Semantic search with three-layer pipeline: contextual prefix injection, metadata filtering, and graph-aware reranking
• Wikilink-aware: results boosted by graph proximity and hub importance

Knowledge management

• Auto-linking — appends ## Related wikilinks after every ingest
• Knowledge logging — create dated, typed notes from any agent
• Label-propagation clustering — organize the vault by link structure
• Wikilink-safe file moves — links rewritten automatically

Integration

• 7 MCP tools, works with any MCP-compatible agent
• Per-agent skill files for Claude Code, OpenCode, Codex CLI, and Copilot
• Background watcher keeps the index in sync automatically

Requirements

• Python ≥ 3.10
• An Obsidian vault (local .md files)
• An MCP-compatible agent

Known limitations

• Markdown-only ingestion (PDF/Office support on the roadmap)
• Link-map rebuild scans the full vault on each search — fine for personal vaults, optimization planned for larger ones