Player not loading? Watch on YouTube
This tutorial walks through the full Hermes Agent setup, using Ollama to run models locally and Firecrawl for web search. The speaker recommends a separate device because the agent's permissions could expose personal files and accounts if it were compromised.
The model setup uses a custom endpoint connected to Ollama and selects Gemma 4 E4B. In the speaker's tests, E2B, described as a 7 GB model, failed basic web-search instructions in Hermes despite working in OpenClaw. E4B, described as 9.6 GB, handled those requests. These are reported results for his setup; he advises testing a model that fits the device and intended tasks.
The walkthrough selects local NEUTTS speech output and a local terminal backend. It also covers tool-call limits, context compression and session resets before creating a Telegram bot and allowing access through a specific user ID. The speaker warns that leaving the allowed-user field blank permits anyone who finds the bot to chat with it.
For self-hosted Firecrawl, Docker must be installed and running. The speaker encounters repeated setup screens and a launch crash, then opens Hermes manually. He reports successful search and Telegram responses. Although he calls the setup private, the demonstration establishes local model and search-service hosting rather than verifying privacy across the entire workflow. It ends with configuration options and a full uninstall.
