A lightweight Python-based weather assistant built with the agents library, designed to provide weather information using local context (city and country). This project demonstrates a minimal, context-aware AI agent with a single-tool architecture and basic memory to enhance user interactions.
-
Local Context Handling: Uses a Pydantic LocationContext model to personalize weather responses based on the user's city and country (default: Lahore, Pakistan).
-
Simplified Tool: Implements a get_weather tool with string-based arguments to fetch weather data (currently using dummy data for simplicity).
-
Minimal Memory: Stores the last query-response pair in a dictionary to enable context-aware responses, such as acknowledging repeated weather queries.
-
Greeting Support: Handles greetings like "hello" with friendly, conversational responses.
-
Error Handling: Robust try-catch blocks to manage API or tool-related errors gracefully.
-
Integrates with the gemini-2.0-flash model via a custom API client for natural language processing.