Author Builds 4-Agent Local Travel Planner With Hermes 0.21.1 and Ollama
Updated
Updated · KDnuggets · Nov 5
Author Builds 4-Agent Local Travel Planner With Hermes 0.21.1 and Ollama
3 articles · Updated · KDnuggets · Nov 5
Summary
A 4-agent setup called Wanderlist runs entirely on local hardware, using one coordinator plus three specialists to research destinations, suggest stays and generate clickable Google Flights search links.
Hermes Agent 0.21.1 provides the orchestration, memory, file access and web-browsing tools, while Ollama serves a local OpenAI-compatible model endpoint so requests stay on the user's machine.
The build recommends gemma4:31b and as much as 24 GB RAM for reliable tool calling, with the coordinator delegating fresh web searches for entry rules, accommodation guidance and fare context.
A custom Python tool creates Google Flights URLs instead of using a paid real-time flight API, validating YYYY-MM-DD dates and returning a working search link without an API key.
The project is pitched as a privacy-first, subscription-free travel assistant whose main trade-off is slower, weaker performance than frontier cloud models, especially on consumer hardware.