AUTOGEN · PYTHON

AutoGen Integration

Persistent, semantically-recalled memory for AutoGen agents. Your agents remember everything — across sessions, across restarts. Dakera handles embedding, storage, and retrieval server-side.

Package: autogen-dakera · GitHub →

Quick Start

Run Dakera

docker run -d \
  --name dakera \
  -p 3300:3300 \
  -e DAKERA_ROOT_API_KEY=dk-mykey \
  ghcr.io/dakera-ai/dakera:latest

curl http://localhost:3300/health

Install

# Core + integration
pip install autogen-dakera

# With AutoGen (if not already installed)
pip install "autogen-dakera[autogen]"

Requirements: Python ≥ 3.10, a running Dakera server.

Add memory to your agent

from autogen_agentchat.agents import AssistantAgent
from autogen_ext.models.openai import OpenAIChatCompletionClient
from autogen_dakera import DakeraMemory

memory = DakeraMemory(
    api_url="http://localhost:3300",
    api_key="dk-mykey",
    agent_id="my-agent",
)

model_client = OpenAIChatCompletionClient(model="gpt-4o")

agent = AssistantAgent(
    name="assistant",
    model_client=model_client,
    memory=[memory],
)

# Agent now persists what it learns across sessions

Configuration

Parameter	Type	Default	Description
`api_url`	str	—	Dakera server URL (e.g. `http://localhost:3300`)
`api_key`	str	`""`	API key set via `DAKERA_ROOT_API_KEY`
`agent_id`	str	—	Logical identifier for this agent's memory
`min_importance`	float	`0.0`	Minimum importance score for recalled memories
`top_k`	int	`5`	Number of memories to surface per query

Multi-agent team with shared memory

import asyncio
from autogen_agentchat.agents import AssistantAgent
from autogen_agentchat.teams import RoundRobinGroupChat
from autogen_agentchat.conditions import MaxMessageTermination
from autogen_ext.models.openai import OpenAIChatCompletionClient
from autogen_dakera import DakeraMemory

async def main():
    shared_memory = DakeraMemory(
        api_url="http://localhost:3300",
        api_key="dk-mykey",
        agent_id="research-team",
        top_k=8,
    )

    model_client = OpenAIChatCompletionClient(model="gpt-4o")

    researcher = AssistantAgent(
        name="researcher",
        model_client=model_client,
        memory=[shared_memory],
        system_message="You are a research expert. Remember key findings.",
    )

    analyst = AssistantAgent(
        name="analyst",
        model_client=model_client,
        memory=[shared_memory],
        system_message="You are a data analyst. Build on what the researcher found.",
    )

    team = RoundRobinGroupChat(
        [researcher, analyst],
        termination_condition=MaxMessageTermination(max_messages=6),
    )

    # First session — agents learn and store
    result = await team.run(task="Research AI memory architectures")
    print(result.messages[-1].content)

    # Later session — agents recall prior research
    result = await team.run(task="What do we know about transformer memory?")
    print(result.messages[-1].content)

asyncio.run(main())

How it works

During conversation, AutoGen calls DakeraMemory.add() with new messages
Dakera embeds the content server-side and stores it with a semantic vector
Before each agent response, AutoGen calls DakeraMemory.query() — Dakera performs hybrid search and returns the most relevant past memories
Memories are injected into the agent's context automatically