A repository for implementing and testing an autonomous agentic pipeline on a real robotic environment using an N9 robotic station (https://www.northrobotics.com/robots).

Installation

1. Clone the repository:

git clone https://github.com/katerinavr/SDL-Agents.git
cd SDL-Agents

2. Install dependencies:

pip install -r requirements.txt

3. Add the API keys (config/settings.py):

OPENAI_API_KEY = ""
anthropic_api_key = ""

4. Run the Gradio app:

python app.py

Components

Core Files

• app.py: Gradio app integrating an option for using human feedback, and an option for displaying live video of the robotic station when a USB camera is connected and has an associated URL address.
• n9_robot_operation_commands.py: Defines the set of available robot operation commands
• S26_commandline_full.py : Defines the set of allowed function to operate the X-ray nanoprobe
• s26_general_info.py : Description of how to interprete the nano-fluoresence and nano-diffraction X-ray images
• params.py: Contains configuration parameters and settings for the system
• sdl_agents.py: Main implementation of SDL agents
• beamline_agents.py: Main implementation of the agentic workflow in the X-ray nanoprobe

Teachability Databases

• teachability_db_gpt4o/: Contains the ChromaDB with the saved input-output pairs after the human teachings using as a base model GPT-4o
• teachability_db_gpt4o-mini/: Contains the ChromaDB with the saved input-output pairs after the human teachings using as a base model GPT-4o-mini

Examples

• notebooks: Contain examples of using the agentic pipeline to operate the N9 robot will tasks of increased complexity.

• videos: A video showing the agentic interface to operate the robotic equipment using a simple prompt can be found under assets\sdl_agents_calms_task1.mp4.