Session Resume

Resume interrupted sessions to continue where you left off.

Interrupted streamed runs now resume with their partial assistant turn (PraisonAI PR #5413). The streaming path persists the user turn and flushes the growing assistant answer as tokens arrive, so Ctrl-C mid-generation no longer loses the in-progress turn. See Streaming — Resuming after an interrupt.

from praisonaiagents import Agent, Session

session = Session(session_id="task-123", persistence="sqlite")
agent = Agent(name="Worker", session=session)

agent.resume()

The user reopens the same session ID after a crash or deploy and the agent continues the in-flight task.

Quick Start

Simple Usage

Reopen a session by its ID and call resume() to continue the in-flight task.

from praisonaiagents import Agent, Session

session = Session(session_id="task-123", persistence="sqlite")
agent = Agent(name="Worker", session=session)
agent.resume()

With Configuration

Add automatic checkpoints so long tasks resume from the latest step, not the start.

from praisonaiagents import Agent, Session

session = Session(
    session_id="task-123",
    persistence="sqlite",
    checkpoint_interval=5,
)
agent = Agent(name="Worker", session=session)
agent.start("Process 1000 items...")

How It Works

Basic Resume

from praisonaiagents import Agent, Session

# Original session
session = Session(session_id="task-123", persistence="sqlite")
agent = Agent(name="Worker", session=session)
agent.start("Start a long task...")

# ... application crashes or restarts ...

# Resume session
session = Session(session_id="task-123", persistence="sqlite")
agent = Agent(name="Worker", session=session)
agent.resume()  # Continues from last checkpoint

With Checkpoints

from praisonaiagents import Agent, Session

session = Session(
    session_id="task-123",
    persistence="sqlite",
    checkpoint_interval=5  # Checkpoint every 5 steps
)

agent = Agent(name="Worker", session=session)

# Long-running task with automatic checkpoints
agent.start("Process 1000 items...")

Manual Checkpoints

from praisonaiagents import Agent, Session

session = Session(session_id="task-123", persistence="sqlite")
agent = Agent(name="Worker", session=session)

# Create checkpoint manually
session.checkpoint(
    state={"processed": 500, "remaining": 500},
    message="Halfway done"
)

# Resume from checkpoint
session.resume_from_checkpoint()

List Sessions

from praisonaiagents import Session

# List all sessions
sessions = Session.list_all(persistence="sqlite")
for s in sessions:
    print(f"{s.session_id}: {s.status}")

Best Practices

Keep the session_id stable across restarts

resume() finds prior state by session_id. Store the ID (env var, queue message, or DB row) so a restarted process reopens the right session.

Tune checkpoint_interval to task length

Frequent checkpoints cost writes but lose less on failure. For step-heavy tasks, a small interval like checkpoint_interval=5 balances safety and overhead.

Save state at natural boundaries

Call session.save_state({...}) after a meaningful unit of work so session.restore_state() reopens at a clean point when the same session_id resumes.

List sessions to reconcile state

Session.list_all(persistence=...) shows every stored session and its status — useful for dashboards and cleanup jobs.

Session Module

Session API reference

Checkpoints

Checkpoint API

Getting Started

Install

Core Concepts

Guides

Features

Models

Databases

Observability

Memory

Telemetry & Analytics

Knowledge

RAG

Persistence

Tools

Other Features

Developers

Configuration

Best Practices

Getting Started (No Code)

Quick Start

How It Works

Basic Resume

With Checkpoints

Manual Checkpoints

List Sessions

Best Practices

Session Module

Checkpoints

​Quick Start

​How It Works

​Basic Resume

​With Checkpoints

​Manual Checkpoints

​List Sessions

​Best Practices

​Related

Session Module

Checkpoints

Quick Start

How It Works

Basic Resume

With Checkpoints

Manual Checkpoints

List Sessions

Best Practices

Related