AutoGen Integration

CortexDB provides persistent memory for Microsoft AutoGen agents, enabling multi-agent conversations that remember context across sessions.

Installation

pip install cortexdb[autogen]

Setup

from autogen import AssistantAgent, UserProxyAgent
from cortexdb.integrations.autogen import CortexMemoryAgent

# Create a memory-augmented assistant
assistant = CortexMemoryAgent(
    name="assistant",
    system_message="You are a helpful AI assistant with long-term memory.",
    llm_config={"model": "gpt-4o"},
    cortex_config={
        "api_key": "your-cortex-api-key",
        "tenant_id": "my-app",
        "namespace": "autogen-assistant",
        "top_k": 10,
    },
)

user = UserProxyAgent(
    name="user",
    human_input_mode="ALWAYS",
)

# Conversations are automatically stored and recalled
user.initiate_chat(assistant, message="What do you remember about our project?")

Teachable Agent Pattern

from cortexdb.integrations.autogen import CortexTeachableAgent

agent = CortexTeachableAgent(
    name="teachable_assistant",
    llm_config={"model": "gpt-4o"},
    cortex_config={
        "api_key": "your-cortex-api-key",
        "tenant_id": "my-app",
    },
)

# Teach the agent
user.initiate_chat(agent, message="Remember: our deploy window is Tuesdays 2-4pm EST.")
# Later sessions...
user.initiate_chat(agent, message="When can we deploy?")
# "Based on what you told me, your deploy window is Tuesdays 2-4pm EST."

Configuration

| Parameter | Default | Description | |---|---|---| | api_key | $CORTEX_API_KEY | CortexDB API key | | tenant_id | Required | Tenant identifier | | namespace | None | Memory namespace | | top_k | 10 | Results per recall | | auto_remember | True | Auto-store conversation turns |