AI + Obsidian + n8n Workflow for Organizing Study Materials

Hi everyone,

following up on the 4/29 call, I wanted to share my project in a bit more detail and get some feedback.

I’m building a personal AI system to manage my school/university workflow. The idea is to automatically organize everything I send — files, notes, audio, etc. — and make it usable later through an AI assistant in a RAG-style setup.

Long-term goal: turn this into a fully self-organizing knowledge system with an AI assistant on top.

Current setup:

- Telegram as the input layer

- n8n as the automation layer

- an AI “distributor” that decides whether to create a new file or update an existing one

- Obsidian as the main knowledge structure

- Google Drive as the synced storage layer

- folder IDs planned to be mapped through a Google Sheet for routing and scalability

Simplified vault structure:

OS-Data/

Subjects/ # Main subject-based knowledge structure

Lit/ # Literature / German language & literature

_subject.md # Subject metadata: teacher, grading, requirements, rules

12_H2/ # Grade 12, second semester

Assignments/ # Tasks that need to be completed

active/ # Open / unfinished assignments

done/ # Completed / submitted assignments

Materials/ # Teacher materials, PDFs, worksheets, sources

Notes/ # Lesson notes, summaries, personal notes

Works/ # Student-created work: essays, presentations, projects

Exams/ # Actual course exams, corrections, grades, feedback

MAT/ # Mathematics

Eng/ # English

INF/ # Computer Science / Informatik

Che/ # Chemistry

Pol/ # Politics

Spa/ # Spanish

Spo/ # Sports

Global/ # General context for the AI system

Me.md # Personal learning profile / preferences

School.md # School context, grading, general rules

Writing.md # Writing style rules and expectations

Templates/ # Templates for creating new notes/files

Unsorted/ # Fallback area for unclear or unsorted files

Each subject also has a _subject.md file with metadata like teacher info, requirements, grading rules, etc.

What works so far:

- automatic classification of incoming content

- routing into subject/category folders

- safe update logic with needs_review fallback

- clean file naming and structure normalization

- updating existing Markdown files in the vault

Current issue:

After syncing Obsidian with Google Drive, I ran into a problem: Claude/Cowork no longer has proper direct access to the actual vault content.

Because the source of truth is now split between Obsidian and Google Drive, the files are accessed through a sync/link layer rather than as one consistent, queryable knowledge base.

So classification, storage and updating files work, but the RAG/querying part is still limited.

Questions:

1. Has anyone worked with an Obsidian + Google Drive + AI setup, especially with Claude/Cowork?

2. How would you give the AI reliable access to the actual content?

3. Would you keep this architecture, or move to something more AI-native like a vector database, Notion, etc.?

I attached the n8n workflow JSON and a manually described vault structure.

Would really appreciate any thoughts, especially on the system design.

0 comments