Antonlytics Agent — Demo

A quick demo showing how to use the antonlytics Python SDK to give any AI model long-term, queryable memory backed by a knowledge graph.

What it does

1. Ingest natural language — the agent extracts entities and relationships automatically.
2. Chat — ask questions; the agent retrieves relevant context and responds.
3. Get memory — pull raw graph data to inject into your own model.

Quickstart

1. Install dependencies

pip install antonlytics django-environ

2. Set up environment variables

Create a .env file in the project root:

ANTONLYTICS_API=your_api_key_here
ANTONLYTICS_PROJECT=your_project_id_here

3. Run the notebook

Open notebook.ipynb and run all cells.

Example

from antonlytics import Agent
import environ

env = environ.Env()
environ.Env.read_env(".env")

agent = Agent(
    api_key=env("ANTONLYTICS_API"),
    project_id=env("ANTONLYTICS_PROJECT")
)

# Teach the agent something
agent.ingest("""
Customer Alice from USA purchased Laptop Pro for $999.
She's interested in our Enterprise plan.
""")

# Set a system prompt
agent.set_system_prompt(
    "You are a helpful assistant to the store's owner. "
    "Respond directly and efficiently."
)

# Ask a question — the agent uses its memory
response = agent.chat("Who purchased laptops?")
print(response["response"])
# => "Alice from USA purchased a Laptop Pro for $999"

# Or pull raw memory for use with your own model
memory = agent.get_memory("laptop purchases")
print(memory)

Sample get_memory response

{
  "entities": [
    { "type": "Person", "properties": { "name": "Alice", "country": "USA" } },
    { "type": "Product", "properties": { "name": "Laptop Pro", "price": "$999" } },
    { "type": "Product", "properties": { "name": "Enterprise plan" } }
  ],
  "relationships": [
    { "type": "PURCHASED", "source": "Alice", "target": "Laptop Pro" },
    { "type": "INTERESTED_IN", "source": "Alice", "target": "Enterprise plan" }
  ]
}

How it works

• Ingest parses free-form text into graph nodes (entities) and edges (relationships).
• Chat retrieves the most relevant subgraph and passes it as context to the underlying model.
• Get memory exposes the raw graph so you can use it with any AI model of your choice.