Nemotron Farm Planner

This application demonstrates a multi-agent system for farm task planning using NVIDIA Nemotron. It allows users to input farm goals via text or audio, which are then processed and executed by a chain of specialized AI agents to generate a validated and summarized farm plan.

---

🚀 Features

• Multi-Agent Architecture: A sophisticated system comprising several specialized AI agents working in conjunction.
  • Interface Agent: Processes human input (text or audio), augments it for better prompting, and clarifies ambiguities for the planning agent.
  • Planner Agent: Parses natural-language farm goals into structured tasks, utilizing function calls for reasoning about constraints like distance, time-windows, and skill checks.
  • Dispatcher Agent: Assigns tasks to workers based on proximity, skill, and shift availability, outputting a per-worker task order.
  • Safety Agent: Automatically checks for and inserts safety-related tasks, such as hydration breaks, considering factors like re-entry intervals (REI), personal protective equipment (PPE), and heat index.
  • Validator Agent: Critically evaluates the outputs from the Planner and Dispatcher agents, ensuring the generated plan is feasible, realistic, and free from overlaps or timing conflicts.
  • Summarizer Agent: Generates a concise, human-readable summary of the final validated farm plan.
• Flexible Input: Provide instructions via text input or voice recording using st.audio_input.
• Dynamic Configuration: Edit system prompts for each agent directly within the Streamlit UI.
• Interactive Data Editors: Modify worker and task data using st.data_editor components.
• Visual Task & Worker Locations: Side-by-side maps display the geographical distribution of workers and tasks.
• Detailed Output: View structured tasks, assignments with integrated water breaks, and notes from the planning process.

---

⚙️ Setup and Run

1. Requirements

• Python 3.10+
• An NVIDIA API Key

2. Environment Variables

Set your NVIDIA API key in a .env file or as an environment variable:

NIM_KEY="<your_api_key>"

3. Installation

python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

4. Run the Application

streamlit run app.py

---

📁 Project Structure

File	Description
app.py	The main Streamlit application, implementing the multi-agent system and UI.
requirements.txt	Python dependencies for the project.
.env	Environment variables, including your NVIDIA API key.
README.md	This comprehensive guide to the Nemotron Farm Planner.

---

© 2025 | Built for demonstration purposes