AMD395 × HelloAgents Hands-on: Smart Travel Planner

This tutorial demonstrates how to build a local smart travel-planning agent on AMD AI PC using:

• Local LLM inference (Ollama / LM Studio / Linglong AI app)
• MCP tool-calling protocol
• AMap APIs for attractions and weather

Goals

1. Run an LLM locally on AMD hardware
2. Connect MCP tools to map/weather APIs
3. Generate practical travel itinerary in Markdown

Prerequisites

• Python 3.10+
• AMD ROCm-compatible runtime environment
• A local model endpoint
• AMap API key

Setup

python -m pip install --upgrade pip
pip install hello-agents requests python-dotenv uv

Then either:

1. set amap_api_key directly in travel_planner_mcp.py, or
2. export the API key through an environment variable:

setx AMAP_MAPS_API_KEY "your_amap_api_key_here"

Run

python travel_planner_mcp.py

The sample script currently runs with a built-in Hangzhou example. Update the values in main() if you want a different destination, number of days, budget, or preferences. The agent will:

1. query POIs and weather
2. reason over budget constraints
3. produce a day-by-day travel plan

Output

The generated result is saved as Markdown (example: hangzhou-3-day-mcp.md).

Architecture Notes

• Agent core: HelloAgents
• Tool bridge: uvx amap-mcp-server
• External service: AMap API
• Runtime: local OpenAI-compatible endpoint

Notes

The Chinese tutorial is available as amd395-helloagents-smart-travel-planner-zh.md.