SK7 AtmosVision Agent

🧩 Features

• Basic LLM agent: Responds to general prompts.
• Custom Weather Agent: Provides city-specific weather, local time, and a generated AI portrait.
• Tool-assisted reasoning: Extracts city info and calls external APIs automatically.
• Interactive UI: Uses Gradio for simple text-based interaction - exclusive for weather agent alone.

---

⚡ Custom Weather Agent Workflow

1. User Input:

   • User types a query (e.g., What's the weather in Paris?).

2. City Extraction:

   • The agent extracts the city name from the input.

3. Tool Calls:

   • get_coordinates(city) → fetches latitude & longitude.
   • get_weather(city, lat, lon) → retrieves current temperature.
   • get_time(city, lat, lon) → calculates local time.
   • generate_portrait(city, temp, time) → creates an AI-generated image of the city reflecting weather/time.

4. Agent Reasoning:

   • The LLM combines all tool results to generate a final response.

5. Output:

   • Displays weather, local time, and portrait image.
   • If running via Gradio, output is interactive and visual.

---

🛠 Models & APIs Used

• LLM: Kimi 2.5 — for reasoning and generating text responses.
• Weather & Geocoding: Open-Meteo — for fetching city coordinates, weather, and forecasts.
• AI Portraits: FLUX.1 Schnell — for generating high-quality city visuals.

---

🔑 Configuration

Before running the agent, create a config.ini file in the project root with your Hugging Face token:

[keys]
hf_token = [YOUR_HUGGINGFACE_TOKEN]