I Built An Obsidian AI Research Assistant with Oz...

AI Research Assistant with Obsidian & Oz: A Detailed Summary

Key Concepts:

• Obsidian: Offline-first, cloud-syncable note-taking application, favored for its flexibility and linking capabilities.
• Oz (by Warp): Orchestration platform for cloud agents, enabling scheduled tasks and SDK-driven agent control.
• Warp: Agentic development environment in the terminal, providing a foundation for Oz.
• Cloud Agents: Autonomous programs running in the cloud, performing tasks like web research and data summarization.
• Scheduled Agents: Agents triggered to run tasks at predefined intervals.
• SDK (Software Development Kit): Tools and libraries for developers to interact with Oz programmatically.
• Hacker News API: Used for fetching news articles, specifically from Y Combinator.
• Git & GitHub: Version control system and platform used for collaborative code and note management.
• UV: Package manager used as an alternative to pip for Python dependencies.

1. Introduction & Project Overview

The video demonstrates building an AI research assistant integrated with an Obsidian vault using Oz, a cloud agent orchestration platform by Warp. The goal is to automate research on the AI/ML space, summarize articles, and create Obsidian notes, all while maintaining version control via GitHub. Two primary approaches are explored: scheduled agents for recurring tasks and the SDK for on-demand research. The presenter highlights Oz’s capabilities for running agents asynchronously and in parallel. Warp is positioned as the underlying agentic development environment.

2. Setting Up the Environment: Warp & Oz Installation

The initial step involves installing the Warp terminal. Installation instructions are provided for macOS, Windows, and Linux (including using an AppImage and an outdated AUR package for Arch Linux). Once installed, Warp provides a terminal interface capable of both standard commands (like lspwd) and agent interaction (via /agent or natural language input). Oz is accessed through a web interface at o.warp.dev or directly through the Warp CLI. The presenter prefers using the CLI for greater control.

3. GitHub Repository Creation & Cloning

A dedicated GitHub repository ("Obsidian-Oz-tutorial") is created to store the Obsidian vault and facilitate collaborative work. This repository is then cloned locally using git clone. The agent will require write access to this repository.

4. Defining the Agent Task & Initial Setup

The core task is defined as: a scheduled agent that scrapes Hacker News for AI/ML-related articles every 5 minutes, summarizes them, creates Obsidian notes, tags them appropriately, and pushes changes to the GitHub repository. The agent is also instructed to tag manually added notes. This task is communicated to Oz via the Warp terminal. Oz automatically creates an environment named "Obsidian Hacker News" using a Docker image (defbase:latest) and connects it to the newly created GitHub repository.

5. Authorization & Scheduled Agent Configuration

Oz requires authorization to access the GitHub repository for write access. This is achieved through a link provided by Oz, leading to a GitHub authorization page. The presenter restricts access to only the "Obsidian-Oz-tutorial" repository for security. Oz then creates a scheduled task to run the defined research and summarization process every 5 minutes (though the presenter acknowledges this is frequent for demonstration purposes). The prompt driving the agent is visible, detailing the Hacker News API usage.

6. Monitoring Agent Execution & Results

The progress of the scheduled agent can be monitored through the Oz web interface, showing the next run time, last run status, and resource usage. After a run completes, new Markdown files representing summarized articles are created in the local Git repository. A git pull command confirms the updates. Opening the repository as an Obsidian vault reveals the newly created notes, properly tagged and formatted.

7. Proactive Research with the Oz SDK

The presenter demonstrates triggering research on demand using the Oz SDK. A Python script is generated by Oz (via the Warp terminal) to manually initiate research on a specified topic. The script utilizes the Oz API. The script initially contains an import error ("cannot import name 'OAgent' from 'oz'") which is corrected by changing 'OAgent' to 'OApi'. The script requires a Warp API key, which is obtained from the Warp settings (Settings -> Platform -> API Keys). The API key is stored in a .env file for security. Running the script with the topic "open source" initiates a new cloud agent run, visible in the Oz web interface.

8. Verification & Customization

After the on-demand research completes, a git pull reveals new notes in the Obsidian vault related to the "open source" topic. The presenter demonstrates modifying the scheduled agent's interval from 5 minutes to 5 hours for practical use. The ability to pause, delete, or edit schedules is highlighted.

9. Conclusion & Call to Action

The video concludes by emphasizing Oz’s capabilities for orchestrating AI agents in the cloud, enabling automated research, and collaborative workflows. The presenter encourages viewers to explore Oz using a provided link ([link in description]) offering 1,000 free credits. He also encourages likes, comments, subscriptions, and hitting the notification bell for future content.

Notable Quotes:

• “In my opinion, it’s better than Notion. It’s offline first, but you can also sync it via the cloud or via a version control system like Git.” – Regarding Obsidian’s advantages.
• “This is a full automated cloud environment for orchestrating AI agents.” – Summarizing Oz’s core functionality.

Technical Terms & Explanations:

• API (Application Programming Interface): A set of rules and specifications that allow different software applications to communicate with each other.
• Docker Image: A lightweight, standalone, executable package of software that includes everything needed to run an application.
• Markdown: A lightweight markup language with plain text formatting syntax.
• UV: A modern Python package manager aiming to be faster and more reliable than pip.
• AppImage: A portable application format for Linux.
• AUR (Arch User Repository): A community-driven repository for Arch Linux packages.
• Cron Job: A time-based job scheduler in Unix-like operating systems.