python-tty 0.2.0

A multi-console TTY framework for complex CLI/TTY apps

Command Line Framework (TTY + Executor + RPC/Web)

中文

This project focuses on the TTY core while providing a unified executor, runtime events, RPC/Web frontends, and session management for complex CLI/TTY apps and lightweight service endpoints.

Concepts

Console layer:

• python_tty/consoles/core.py: BaseConsole, MainConsole, SubConsole
• python_tty/consoles/manager.py and python_tty/consoles/registry.py for lifecycle and registration

Commands layer:

• python_tty/commands/core.py: BaseCommands, CommandValidator
• python_tty/commands/registry.py: CommandRegistry, ArgSpec
• python_tty/commands/general.py: GeneralValidator, GeneralCompleter
• python_tty/commands/mixins.py: CommandMixin and built-in mixins

Execution and runtime:

• python_tty/executor/executor.py: CommandExecutor (unified execution entry)
• python_tty/runtime/jobs.py: JobStore (RunState/Invocation/event history)
• python_tty/runtime/events.py: RuntimeEvent, UIEvent, UIEventLevel
• python_tty/runtime/router.py: proxy_print and output routing

Session and callbacks:

• python_tty/session/manager.py: SessionManager (lifecycle + submit + constraints)
• python_tty/session/store.py: SessionStore (in-memory session state)
• python_tty/session/callbacks.py: callback subscriptions (thread-safe cancel)

RPC/Web:

• python_tty/frontends/rpc: gRPC Invoke + event streaming
• python_tty/frontends/web: FastAPI + WS snapshot (Meta)

Table rendering:

• python_tty/utils/table.py: table rendering (auto wrap supported)

TTY / RPC scheduling logic in Executor:

• TTY: builds Invocation, submits via executor.submit_threadsafe(), workers emit RuntimeEvent.
• RPC: builds Invocation, enforces exposure/allowlist/audit, StreamEvents subscribes to event queues.

Configuration

Config entry is python_tty/config/config.py.

ConsoleFactoryConfig:

• run_mode: "tty" or "concurrent"; decides whether the main thread runs the loop + TTY thread.
• start_executor: auto-start executor on factory start.
• executor_in_thread: start executor in a background thread in TTY mode.
• executor_thread_name: executor loop thread name.
• tty_thread_name: TTY thread name in concurrent mode.
• shutdown_executor: shutdown executor when factory stops.

ExecutorConfig:

• workers: number of workers (execution concurrency).
• retain_last_n: keep last N completed runs in memory.
• ttl_seconds: TTL for completed runs.
• pop_on_wait: drop run state after wait_result.
• exempt_exceptions: treat these as cancellation.
• emit_run_events: emit start/success/failure state events.
• event_history_max: max events per run.
• event_history_ttl: TTL for per-run history.
• sync_in_threadpool: run sync handlers in a thread pool.
• threadpool_workers: max thread pool workers.
• audit: AuditConfig for audit sink.

AuditConfig:

• enabled: enable audit.
• file_path: JSONL file output.
• stream: stream output (exclusive with file_path).
• async_mode: async writer mode.
• flush_interval: async flush interval.
• keep_in_memory: keep records in memory (tests).
• sink: custom AuditSink instance.

RPCConfig:

• enabled: start RPC server.
• bind_host / port: listen address and port.
• max_message_bytes: max gRPC message size.
• keepalive_time_ms / keepalive_timeout_ms / keepalive_permit_without_calls: keepalive options.
• max_concurrent_rpcs: max RPC concurrency.
• max_streams_per_client: max concurrent streams per client.
• stream_backpressure_queue_size: per-stream queue limit.
• default_deny: deny when exposure is missing.
• require_rpc_exposed: require exposure.rpc=True.
• allowed_principals: principal allowlist.
• admin_principals: admin principals (bypass allowlist only).
• require_audit: RPC requires audit sink.
• trust_client_principal: trust request.principal (default False).
• mtls: MTLSServerConfig.

MTLSServerConfig:

• enabled: enable mTLS.
• server_cert_file / server_key_file: server cert/key.
• client_ca_file: client CA bundle.
• require_client_cert: require client certs.
• principal_keys: auth_context keys for principal extraction.

WebConfig:

• enabled: start Web server.
• bind_host / port: listen address and port.
• root_path: reverse-proxy root path.
• cors_*: CORS options.
• meta_enabled: enable /meta.
• meta_cache_control_max_age: cache max-age for /meta.
• ws_snapshot_enabled: enable /meta/snapshot websocket.
• ws_snapshot_include_jobs: include running job summary.
• ws_max_connections: max WS connections.
• ws_heartbeat_interval: WS heartbeat seconds.
• ws_send_queue_size: WS send queue size.

Factory impact:

• run_mode="tty": TTY runs on main thread; executor can be started in main or background thread.
• run_mode="concurrent": main thread runs asyncio loop; TTY runs in background thread.
• rpc.enabled=True / web.enabled=True: services are attached during factory startup.
• start_executor=False: RPC/Web/Session submit will be unavailable or raise (no executor loop).

Examples

Quick start (TTY core):

1. Define your consoles and commands (see examples in python_tty/consoles/examples and python_tty/commands/examples).
2. Ensure console modules are imported so decorators can register them: update DEFAULT_CONSOLE_MODULES in python_tty/consoles/loader.py or call load_consoles([...]).
3. Start the factory:

from python_tty.console_factory import ConsoleFactory

factory = ConsoleFactory(service=my_business_core)
factory.start()

Job/Session basic usage:

from python_tty.console_factory import ConsoleFactory

factory = ConsoleFactory(service=my_service)
factory.start_executor()
sm = factory.session_manager

sid = sm.open_session(principal="local")
run_id = sm.submit_command(sid, "cmd:root:help")
result = await sm.result(run_id)

Mode A: synchronous orchestration (outer awaits inner)

async def outer():
    inner_id = sm.submit_command(sid, "cmd:root:long_task", await_result=True)
    inner_result = await sm.result(inner_id)
    return inner_result

Mode B: async orchestration (callbacks)

def on_event(evt):
    pass

def on_done(evt):
    pass

inner_id = sm.submit_command(sid, "cmd:root:long_task")
sub_id = sm.register_callback(inner_id, on_event=on_event, on_done=on_done)

Table rendering:

from python_tty.utils import Table

header = ["name", "status", "detail"]
data = [
    ["job-1", "running", "very long text ..."],
    ["job-2", "done", "short"],
]

table = Table(header, data, title="jobs", wrap=True)
print(table)

Single row (debug):

print(table.print_line(["job-3", "queued", "pending"]))

展望

待定