How to use Agent Space

Environment Set Up

1. If you do not have Conda download it

2. Create a new conda environment in terminal using:

conda create -n <environment_name> python=3.13

3. Activate the conda environment using:

conda activate <environment_name>

4. Download dependencies by running the following command in the same directory as environment.yml. If on apple use environment.yml on windows use environment_windows.yml:

conda env update -f environment.yml

Using AgentSpace with Jupyter Notebooks

1. Navigate to the KnowledgeRetriever folder and open Neo4jGraph.ipynb

2. Follow the instructions in Neo4jGraph.ipynb to get Neo4j set up and create your graph

   Note: These two variables are set:

   INDEX_NAME = "essay_chunk_agentspace"
   CHUNK_NAME = "chunk_index"
   

   What they are set to must be passed into the agent. It is how the agent determines what documents to look at.

   agent = personAgent("essay_chunk_agentspace", "chunk_index")

3. Navigate to the Agent folder and open TestAgent.ipynb

4. Run testAgent to ask the agent questions

   Note: Agent must be made using the INDEX_NAME and CHUNK_NAME set in Neo4jGraph.ipynb

Current Repository Setup

├── Orchestrator/                 # AgentSpace Agent
│   ├── Agent.py                  # Defines the Agent, uses Knowledge and Personality Retrievers
│   └── TestAgent.ipynb           # Test an Agent
├── KnowledgeRetriever/           # Graph/retrievers for knowledge
│   ├── Retriever.py              # Holds the graph retriever
│   ├── Neo4jAdapter.py           # Wraps Retriever.py for use with Neo4j
│   └── Neo4jGraph.ipynb          # Jupyter notebook to set up Neo4j graph
├── PersonalityRetriever/         # Database/retrievers for personality
│   ├── faiss_db                  # faiss vector store
|   ├── create_profile.py         # Uses LLM API call to generate personality.json
|   ├── grant_tutor_rag           # Current personality pipeline. Creates faiss db
|   ├── personality_summary.json  # JSON profile. Serves as a general lookup of personality traits and characteristics.
|   └── personality_wrapper.py    # Wraps retrieval and loading of personality JSON
└── README.md                     # Project documentation