Browse / LLM wiki / Personal-LLM-Wiki

Personal-LLM-Wiki

Vncntvx

基于 LLM 的可自我维护个人知识库的设计与工作流

Open Vncntvx/Personal-LLM-Wiki → Checked against the four rules.

Works with any agent

ships AGENTS.md, the cross-tool instruction file.

In its own words

README.mdVncntvx/Personal-LLM-Wiki

Personal knowledge-base system that encapsulates core workflows as Skills to automatically transform reusable source materials into a structured Wiki. Paired with the qmd MCP search engine, it enables automated ingestion, intelligent querying, and health maintenance for long-term knowledge preservation and efficient retrieval.

Overview

Core capabilities
  • 🔄 Automatic sync detection — detect new, modified and deleted files under the ObsidianRaw/03_Resources/ directory
  • 📥 Intelligent ingestion — extract concepts and entities from source materials, produce concise summaries, and build indexes
  • 🔍 Hybrid retrieval — combine lexical (keyword) search and vector (semantic) search for accurate recall
  • 📊 Health checks — detect orphan pages, concept gaps, contradiction annotations, cognitive quality, and sync issues

Architecture

Ingestion scope

**Only files under ObsidianRaw/03_Resources/ are ingested.**

Exclusions

The ingestion pipeline skips:

  • The Personal/ folder (sensitive data)
  • Files that contain only images (no text)
  • Shopping lists or configuration dumps (no knowledge value)

Quick start

1. Full sync (recommended)

Runs the complete workflow in one pass: detect → confirm → process → verify → log → report.

2. Detect new/changed files

Produces a sync report comparing ObsidianRaw/03_Resources/ and Wiki/Sources/:

Quoting 50% of the README. Read the rest on GitHub ↗

Read from the repository's git tree on 2026-08-17.