Custom Session Summary

"""
Custom Session Summary
=====================

Demonstrates configuring a custom session summary manager and reusing summaries in
context.
"""

from agno.agent import Agent
from agno.db.sqlite import SqliteDb
from agno.models.openai import OpenAIResponses
from agno.session import SessionSummaryManager
from agno.team import Team

# ---------------------------------------------------------------------------
# Setup
# ---------------------------------------------------------------------------
db = SqliteDb(
    db_file="tmp/team_session_summary.db",
    session_table="team_summary_sessions",
)
summary_manager = SessionSummaryManager(model=OpenAIResponses(id="gpt-5-mini"))


# ---------------------------------------------------------------------------
# Create Members
# ---------------------------------------------------------------------------
planner = Agent(
    name="Sprint Planner",
    model=OpenAIResponses(id="gpt-5-mini"),
    instructions=[
        "Build concise, sequenced plan summaries.",
        "Keep recommendations practical.",
    ],
)


# ---------------------------------------------------------------------------
# Create Team
# ---------------------------------------------------------------------------
sprint_team = Team(
    name="Sprint Team",
    model=OpenAIResponses(id="gpt-5-mini"),
    members=[planner],
    db=db,
    session_summary_manager=summary_manager,
    add_session_summary_to_context=True,
)


# ---------------------------------------------------------------------------
# Run Team
# ---------------------------------------------------------------------------
if __name__ == "__main__":
    session_id = "sprint-planning-session"

    sprint_team.print_response(
        "Plan a two-week sprint for a small team shipping a documentation portal.",
        stream=True,
        session_id=session_id,
    )

    sprint_team.print_response(
        "Now add testing and rollout milestones to that plan.",
        stream=True,
        session_id=session_id,
    )

    summary = sprint_team.get_session_summary(session_id=session_id)
    if summary is not None:
        print(f"\nSession summary: {summary.summary}")
        if summary.topics:
            print(f"Topics: {', '.join(summary.topics)}")

    sprint_team.print_response(
        "Using what we discussed, suggest the most important next action.",
        stream=True,
        session_id=session_id,
    )