DocsPlugin System

Plugins

Plugins are reusable event handlers. Package capabilities and reuse them across agents.

Quick Start

main.py
python
from connectonion import Agent
from connectonion.useful_plugins import re_act

agent = Agent("assistant", tools=[search], plugins=[re_act])
agent.input("Search for Python")
# 💭 Planning: Will search for Python info first.
# ... tool executes ...
# 🤔 Reflecting: Found Python basics, task complete.

What is a Plugin?

A plugin is a list of event handlers:

main.py
python
from connectonion import after_user_input, after_tools

# Define handlers
def plan(agent):
    print("Planning...")

def reflect(agent):
    print("Reflecting...")

# Plugin = list of event handlers
re_act = [after_user_input(plan), after_tools(reflect)]

# Use it
agent = Agent("assistant", tools=[search], plugins=[re_act])

Plugin vs on_events

main.py
python
# on_events: one list (custom for this agent)
agent = Agent("a", on_events=[after_llm(handler)])

# plugins: list of lists (reusable across agents)
agent = Agent("a", plugins=[re_act, logger])

Built-in Plugins

Plugin	Purpose	Docs
re_act	Planning + reflection (ReAct pattern)	docs
auto_compact	Auto-compress context at 90% — prevents hitting token limits	docs
subagents	Load sub-agents from AGENT.md files via task() tool	docs
ulw	Ultra Light Work — autonomous mode from web UI	docs
ui_stream	WebSocket streaming of completion events to UI	docs
tool_approval	Require user approval before each tool call	docs
shell_approval	Approve shell commands before execution	docs
prefer_write_tool	Encourage agent to use FileTools.write() over bash	docs
system_reminder	Inject system reminders into LLM context	docs
image_result_formatter	Format images for vision models	docs
eval	Task evaluation for debugging	docs

main.py
python
from connectonion.useful_plugins import re_act, eval, image_result_formatter, shell_approval

# Combine plugins
agent = Agent("assistant", plugins=[re_act, image_result_formatter])

Writing Custom Plugins

Simple example - log each tool execution:

main.py
python
from connectonion import Agent, after_each_tool

def log_tool(agent):
    trace = agent.current_session['trace'][-1]
    print(f"✓ {trace['tool_name']} completed")

# Plugin = list of handlers
logger = [after_each_tool(log_tool)]

# Use it
agent = Agent("assistant", tools=[search], plugins=[logger])

For more complex plugins, see Events for available event hooks.

Reusing Plugins

main.py
python
# Define once
logger = [after_each_tool(log_tool)]

# Use across agents
researcher = Agent("researcher", tools=[search], plugins=[logger])
writer = Agent("writer", tools=[generate], plugins=[logger])