Skip to main content
This example demonstrates how to configure agents to search through previous sessions and limit the number of historical sessions included in context. This helps manage context length while maintaining relevant conversation history.
1

Create a Python file

last_n_session_messages.py
import os

from agno.agent import Agent
from agno.db.sqlite import SqliteDb
from agno.models.openai import OpenAIResponses

# Remove the tmp db file before running the script
os.remove("tmp/data.db")

agent = Agent(
    model=OpenAIResponses(id="gpt-5.2"),
    user_id="user_1",
    db=SqliteDb(db_file="tmp/data.db"),
    add_history_to_context=True,
    num_history_runs=3,
    search_session_history=True,  # allow searching previous sessions
    num_history_sessions=2,  # only include the last 2 sessions in the search to avoid context length issues
)

session_1_id = "session_1_id"
session_2_id = "session_2_id"
session_3_id = "session_3_id"
session_4_id = "session_4_id"
session_5_id = "session_5_id"

agent.print_response("What is the capital of South Africa?", session_id=session_1_id)
agent.print_response("What is the capital of China?", session_id=session_2_id)
agent.print_response("What is the capital of France?", session_id=session_3_id)
agent.print_response("What is the capital of Japan?", session_id=session_4_id)
agent.print_response(
    "What did I discuss in my previous conversations?", session_id=session_5_id
)  # It should only include the last 2 sessions
2

Set up your virtual environment

uv venv --python 3.12
source .venv/bin/activate
3

Install dependencies

uv pip install -U agno openai
4

Export your OpenAI API key

​
Set OpenAI Key

Set your OPENAI_API_KEY as an environment variable. You can get one from OpenAI.
export OPENAI_API_KEY=sk-***
5

Run Agent

python last_n_session_messages.py