Full Hermes Agent Tutorial (Desktop) 🧠 A Useful Agentic AI Workflow
This tutorial demonstrates how to set up and use Hermes, an open-source AI agent by Nous Research that learns and evolves over time through persistent memory and automated skill generation. The video covers local model setup with Ollama, cloud model integration with OpenAI, messaging gateway configuration via Telegram, and creating a self-improving daily AI briefing automation.
Summary
Callum (Waterloos) presents a comprehensive guide to Hermes, an AI agent framework that differentiates itself by learning user workflows and preferences over time rather than remaining static. Unlike most AI tools that have fixed capabilities, Hermes automatically generates reusable skills from recurring workflows and loads them when similar tasks are recognized.
The tutorial begins with installation of the Hermes desktop application and covers two primary setup approaches: local model deployment using Ollama (with models like Gemma 4) and cloud model integration using OpenAI. For local models, the presenter demonstrates how to increase context windows from the default 2048 tokens to 65,536 tokens, which is necessary for Hermes to function properly with its full toolset. The platform allows users to create separate profiles for different models and configurations, enabling privacy-focused local operations alongside more powerful cloud-based models.
A significant portion covers Hermes' architecture and settings, including its 71 built-in skills (reusable workflows), tool sets that can be enabled/disabled for security and control, and persistent memory features that create user.md and memory.md files. These memory files are updated automatically as the agent learns about user preferences and projects. The platform also offers execution sandboxing options through Docker containers, remote execution via VPS or SSH, and voice capabilities that run locally.
The practical demonstration shows setting up Telegram as a messaging gateway, allowing the user to interact with Hermes from their phone. The presenter walks through creating a bot using BotFather and configuring Telegram user ID authentication. The main use case presented is building a self-improving daily briefing automation that uses a cron job to run at scheduled times, pull AI news from the web, message it to Telegram, and then update its internal skill based on user feedback.
The self-learning cycle is the core innovation: the user receives a daily briefing, provides feedback on what they want more or less of, and the agent updates the underlying skill that gets used for the next day's report. This requires creating multiple cron jobs—one for the daily report and another that runs beforehand to update the skill based on previous feedback. The presenter demonstrates iterating through several attempts to get the automation working correctly, highlighting the practical reality of tuning autonomous systems.
Key Insights
- Hermes differentiates itself from other AI tools by automatically learning from recurring workflows and generating reusable skills, whereas most AI tools have static capabilities regardless of how frequently they're used.
- Nous Research built Hermes with an explicit mission to advance human rights and freedoms by making open-source language models freely available and unrestricted, reflecting their commitment to accessible AI.
- Hermes requires a minimum context window of 64,000 tokens to provide agents with tools; many default Ollama models use only 2,048 tokens, necessitating manual context window expansion.
- The presenter argues that users should ask Hermes to set up its own configurations rather than doing manual setup, as the agent learns to handle these tasks better over time through experimentation.
- The daily briefing automation creates a feedback loop where user comments on Telegram trigger memory updates and skill updates, enabling the system to become progressively more personalized without explicit retraining.
Topics
Transcript
[0:00] Okay, there we go. We can now launch the Hermes desktop. Most AI tools are as capable at day one as they are at 100. They don't necessarily learn and evolve over time. You learn it, but it doesn't learn you. Hermes is different. The more you use it, the better it gets at your specific work, your task, what you want it to do automatically. Not because you trained it or told it to do that, but because it trains itself. It compounds and learns from its behavior over time. It's built into the Hermes architecture. It's by design. It's built by an actual research company that specializes in [0:32] training open models. Hermes by Nous Research. Nous…
Full transcript available for MurmurCast members
Sign Up to AccessMore from Wanderloots
100% Free & Private AI 🦙 Build & Run Local AI Agents #ai #agenticai #ollama #localllm
The video introduces local AI models as a privacy-preserving alternative to cloud-based AI services. By running AI assistants locally on your computer, users can maintain complete control over their data while leveraging AI capabilities for personal tasks like analyzing notes and answering work-related questions.
Information ➡️ Knowledge: How To Build An LLM Wiki In Obsidian 🧠 #obsidian #ai #agenticai
This video introduces the concept of an 'LLM Wiki' built in Obsidian — a structured knowledge system where AI agents automatically extract and organize concepts from raw information. The presenter explains that the goal is to transform information into persistent, accessible knowledge. This episode focuses on the practical setup, following a prior video that covered the rationale.
Obsidian Smart Plugin Workflow 📝 New Smart Connections + Context AI
Callum (Waterloo Loot) demonstrates a three-plugin 'smart loop' workflow in Obsidian using Smart Connections, Smart Context, and Smart Chat to solve the problem of losing notes and context over time. The workflow follows a discovery-preparation-use cycle that helps users surface relevant notes, bundle them into reusable context, and optionally pass that context to AI while keeping chat threads linked to specific projects. The entire system is built on a shared 'smart environment' vector database layer that automatically updates context bundles as notes evolve.
Why LLM Wiki? 🧠 Future Of Knowledge For Agentic AI & Humans
Callum, a former IP lawyer, explains the concept of knowledge graphs and introduces the 'LLM Wiki' — a separate, AI-maintained structured knowledge base that allows multiple AI tools to share the same persistent, interlinked information. He contrasts standard RAG retrieval with graph RAG, arguing that a structured wiki layer dramatically improves how AI handles complex, multi-source knowledge.
Academic Search Engine + Agentic AI 📚 Consensus & Zotero Deep Research
The video demonstrates Consensus AI, an academic search platform that searches over 220 million peer-reviewed papers using an agentic 'Scholar AI' system. The presenter walks through deep research workflows, library management, Zotero integration, and a citation graph feature that visually maps relationships between papers across time.