zeno-channel-cli

Terminal channel for Zeno. Reads lines from stdin into IncomingMessages and renders the agent's OutgoingMessages to stdout. Doubles as a manual smoke-test harness and as the simplest reference implementation of the Channel protocol.

Install

uv add 'zeno-framework[cli]'
# or, without the AI package:
uv add zeno-channel-cli

Minimal usage

import asyncio

from zeno.agent import Agent
from zeno.app import ZenoApp
from zeno.channels.cli import CliChannel
from zeno.testing import FakeProvider

app = ZenoApp(
    agent=Agent(name="root", instructions="You are Zeno, a helpful assistant."),
    channels=[CliChannel()],
    provider=FakeProvider(),  # swap for OpenAIProvider / ClaudeSDKProvider in production
)

asyncio.run(app.run())

Then run the script. Type a line, hit return, see the reply. Ctrl-D or Ctrl-C exits cleanly. user_id defaults to $USER (or "local" if unset) so a developer running the snippet above gets a sensible tenant scope without configuration. Override explicitly in production:

CliChannel(user_id="alice", prompt=">>> ")

Behavior

• Capabilities: supports_streaming=False, supports_threading=False. The core StreamBuffer consolidates chunked replies into one send() at finalize-time, which the renderer emits as a single line.
• Stdin: read via asyncio.to_thread(stream.readline) so the CLI channel stays dependency-free (no aioconsole). At most one line read is in flight at a time, matching REPL semantics.
• Inbound queue: exposes inbound_put(message) so zeno-scheduler can inject proactive messages without going through stdin.
• Shutdown: stop() is idempotent and closes the renderer; calling send() after stop() raises ChannelError.

Testing

uv run pytest packages/zeno-channel-cli

Tests use io.StringIO for stdin/stdout — no terminal needed.

See also

• zeno-channel-email, zeno-channel-sendblue — other first-party channels.
• docs/adapters.md — writing your own Channel.

Part of the Zeno framework.