Skip to main content

Use your main vault

Augent works best inside the vault you already use. Your personal notes, daily journals, projects, research, whatever you keep in Obsidian, that’s where Augent should be creating files too. When Augent takes notes, generates transcriptions, or builds automations, those files belong alongside the rest of your work. They link to your existing notes. They show up in your graph. They compound with everything you’ve already built. There’s no reason to separate this into a different vault. Augent is a tool for your notes, not a separate system.

Point Augent at your vault

By default, take_notes saves files to ~/Desktop. To save directly into your vault instead, set the output path when calling the tool:
take_notes content="..." path="~/Documents/MyVault/Notes/"
Or configure a default output directory in your augent settings so every tool writes to your vault automatically.

Optional: dedicated memory vault

If you want a separate view of just your transcriptions and audio memory, you can open Augent’s memory directory as its own vault:
  1. Open Obsidian
  2. Click Open another vault (bottom-left)
  3. Click Open folder as vault
  4. Select ~/.augent/memory/transcriptions/
This gives you a dedicated graph view of everything you’ve transcribed, with tags, MOCs, and wikilinks. Some people like having this as a focused workspace for audio research. But this is optional, not the primary setup.

Rebuild the graph

If you already have transcriptions in memory, run rebuild_graph to add YAML frontmatter, compute [[wikilinks]] between related files, and generate MOC hub files:
rebuild_graph
Safe to run repeatedly. It migrates old-format files, syncs tags, and recreates any missing .md files from the database.

Frontmatter format

Every file Augent creates has YAML frontmatter at the top:
---
title: "How Transformers Work"
tags:
  - AI
  - Science
source: "how_transformers_work.webm"
source_url: "https://youtube.com/watch?v=..."
duration: "47:23"
language: en
date: 2026-03-20
type: transcription
---
Obsidian reads this natively. Tags become filterable nodes in graph view, and frontmatter fields are searchable.