CubeLoop — a Pythonic, async-native agent harness core designed for high performance, readability, and production-grade persistence.
Python agent-frameworkagent-harnessagent-runtimeagentic-aiai-agentsai-frameworkanthropicharnessharness-engineeringharness-frameworklanggraphlanggraph-alternativellmmcpopenaipi-agentpi-agent-corepythonpython-agent
"""Persistent chat — recipe example.
A REPL chat that survives restarts. Conversation history is kept in a SQLite
file; pass a thread_id as the first argument to identify the user/session.
uv run python examples/persistent_chat.py alice
# Have a chat, then Ctrl-D.
uv run python examples/persistent_chat.py alice
# History is restored. Ask "what did I just tell you?" — the model remembers.
uv run python examples/persistent_chat.py bob
# Different thread, clean slate.
Requires: cubepi[sqlite]
Set ANTHROPIC_API_KEY or OPENAI_API_KEY before running (see _provider.py).
"""
import asyncio
import sys
from cubeloop import Agent
from cubeloop.checkpointer import SQLiteCheckpointer
from _provider import MODEL_ID, provider
async def main(thread_id: str) -> None:
async with SQLiteCheckpointer("chat.db") as cp:
agent = Agent(
model=provider.model(MODEL_ID),
system_prompt="You are a concise, friendly assistant.",
checkpointer=cp,
thread_id=thread_id,
)
def on_event(event, signal=None):
if event.type == "message_update" and event.stream_event.type == "text_delta":
print(event.stream_event.delta, end="", flush=True)
elif event.type == "agent_end":
print()
agent.subscribe(on_event)
print(f"Chatting on thread {thread_id!r}. Ctrl-D to quit.\n")
loop = asyncio.get_event_loop()
while True:
try:
user_input = await loop.run_in_executor(None, input, "you> ")
except EOFError:
print()
return
if not user_input.strip():
continue
print("ai > ", end="", flush=True)
await agent.prompt(user_input)
if __name__ == "__main__":
thread_id = sys.argv[1] if len(sys.argv) > 1 else "default"
asyncio.run(main(thread_id))