Hermes

Overview

Hermes is my long-term engineering project focused on building an AI teammate rather than another chatbot.

Instead of relying on individual prompts, Nova combines persistent memory, document retrieval, scheduled automation, and infrastructure integrations into a single platform capable of assisting with real operational work.

Every capability is designed around one principle:

Reliable systems are more valuable than impressive demos.

The project continues to evolve through real-world usage as I work as a Systems Administrator while expanding into cloud engineering and AI automation.

What I Built

• Memory

  Built a persistent memory system that allows Nova to retain context across sessions.

• Automation

  Created agents capable of completing scheduled and event-driven workflows.

• Knowledge

  Integrated document retrieval from my Obsidian knowledge base.

• Communication

  Connected Telegram to interact with Nova remotely.

• Infrastructure

  Hosted the platform on a dedicated Mac mini with Docker services.

• Architecture

  Designed the platform to be modular so additional agents can be added over time.

What I Learned

Building an AI assistant is primarily a systems engineering problem.

Some of the biggest lessons so far:

• Retrieval matters more than larger prompts.

• Reliable automation beats impressive demos.

• Memory should improve future work, not simply store conversations.

• Every workflow should have a clear failure path.

• Users trust systems they can verify.

Challenges

Current engineering challenges include:

• Scaling the architecture while keeping modules independent.

• Deciding what Nova should remember.

• Preventing automation from becoming overly complex.

• Balancing autonomy with human oversight.

• Maintaining reliability as capabilities grow.