Implement Gemini Multimodal Interaction

Extend your Google ADK agent to use Gemini 1.5 Pro for processing user utterances and generating responses. Your agent should be able to interpret natural language commands related to navigation and safety. Show how you would configure the ADK to use Gemini as the primary LLM for reasoning and how to pass a user's voice input to Gemini and then use Gemini's response to drive the `speak` tool. Also, outline how you would pass environmental context (like current location and activity) to Gemini.