Obsidian Local Brain

A fully local RAG pipeline that turns an Obsidian vault into daily spaced-repetition quizzes with a local vision LLM.

Status
Active
Date
Context
Personal
Language
Python
Stars
0
Last commit
2025-11-28
  • Code assisted
  • Writing assisted
What do these mean?

Turn your passive notes into an active study partner. A fully local, privacy-first automation pipeline that connects your Obsidian Vault to a local LLM (Llama 3.2 Vision) via Docker.

The Concept

Working through security material, I take hundreds of notes (Penetration Testing, Network Forensics, etc.). However, taking notes is passive. Remembering them is active.

I built this tool to automate Spaced Repetition and Active Recall without sending my private data to the cloud.

How it works

  • Productivity Analysis: The script scans my vault daily to check how many words I wrote (Productivity Score).
  • Weighted Selection: It selects 3 notes from “Today” and 2 notes from “Last Month” to force spaced repetition.
  • Local RAG: It feeds these notes into a local Dockerized LLM (Ollama).
  • Exam Generation: The AI acts as a strict examiner, generating a Markdown-formatted quiz with Hidden Answers and Backlinks to the original source.
  • Feedback Loop: If I tag a question as #missed, the system detects my weakness and re-tests me on that topic later.

Features

  • 100% Privacy: No API keys, no data leaves the localhost.
  • Multimodal Ready: Supports extracting text from images using Llama 3.2 Vision.
  • Dynamic Difficulty:
    • High Activity Day (>500 words): Generates a “Deep Work” exam (10 questions).
    • Low Activity Day: Generates a “Maintenance” quiz (5 questions).
  • Interactive UI: Uses Obsidian Callouts and Collapsible Admonitions for a clean “App-like” experience inside Markdown.