feat: notify_on_complete for background processes (#5779)

* feat: notify_on_complete for background processes

When terminal(background=true, notify_on_complete=true), the system
auto-triggers a new agent turn when the process exits — no polling needed.

Changes:
- ProcessSession: add notify_on_complete field
- ProcessRegistry: add completion_queue, populate on _move_to_finished()
- Terminal tool: add notify_on_complete parameter to schema + handler
- CLI: drain completion_queue after agent turn AND during idle loop
- Gateway: enhanced _run_process_watcher injects synthetic MessageEvent
  on completion, triggering a full agent turn
- Checkpoint persistence includes notify_on_complete for crash recovery
- code_execution_tool: block notify_on_complete in sandbox scripts
- 15 new tests covering queue mechanics, checkpoint round-trip, schema

* docs: update terminal tool descriptions for notify_on_complete

- background: remove 'ONLY for servers' language, describe both patterns
  (long-lived processes AND long-running tasks with notify_on_complete)
- notify_on_complete: more prescriptive about when to use it
- TERMINAL_TOOL_DESCRIPTION: remove 'Do NOT use background for builds'
  guidance that contradicted the new feature
This commit is contained in:
Teknium
2026-04-07 02:40:16 -07:00
committed by GitHub
parent 1c425f219e
commit e120d2afac
6 changed files with 398 additions and 10 deletions

View File

@@ -6048,12 +6048,13 @@ class GatewayRunner:
platform_name = watcher.get("platform", "")
chat_id = watcher.get("chat_id", "")
thread_id = watcher.get("thread_id", "")
agent_notify = watcher.get("notify_on_complete", False)
notify_mode = self._load_background_notifications_mode()
logger.debug("Process watcher started: %s (every %ss, notify=%s)",
session_id, interval, notify_mode)
logger.debug("Process watcher started: %s (every %ss, notify=%s, agent_notify=%s)",
session_id, interval, notify_mode, agent_notify)
if notify_mode == "off":
if notify_mode == "off" and not agent_notify:
# Still wait for the process to exit so we can log it, but don't
# push any messages to the user.
while True:
@@ -6077,6 +6078,47 @@ class GatewayRunner:
last_output_len = current_output_len
if session.exited:
# --- Agent-triggered completion: inject synthetic message ---
if agent_notify:
from tools.ansi_strip import strip_ansi
_out = strip_ansi(session.output_buffer[-2000:]) if session.output_buffer else ""
synth_text = (
f"[SYSTEM: Background process {session_id} completed "
f"(exit code {session.exit_code}).\n"
f"Command: {session.command}\n"
f"Output:\n{_out}]"
)
adapter = None
for p, a in self.adapters.items():
if p.value == platform_name:
adapter = a
break
if adapter and chat_id:
try:
from gateway.platforms.base import MessageEvent, MessageType
from gateway.session import SessionSource
from gateway.config import Platform
_platform_enum = Platform(platform_name)
_source = SessionSource(
platform=_platform_enum,
chat_id=chat_id,
thread_id=thread_id or None,
)
synth_event = MessageEvent(
text=synth_text,
message_type=MessageType.TEXT,
source=_source,
)
logger.info(
"Process %s finished — injecting agent notification for session %s",
session_id, session_key,
)
await adapter.handle_message(synth_event)
except Exception as e:
logger.error("Agent notify injection error: %s", e)
break
# --- Normal text-only notification ---
# Decide whether to notify based on mode
should_notify = (
notify_mode in ("all", "result")
@@ -6101,8 +6143,9 @@ class GatewayRunner:
logger.error("Watcher delivery error: %s", e)
break
elif has_new_output and notify_mode == "all":
elif has_new_output and notify_mode == "all" and not agent_notify:
# New output available -- deliver status update (only in "all" mode)
# Skip periodic updates for agent_notify watchers (they only care about completion)
new_output = session.output_buffer[-500:] if session.output_buffer else ""
message_text = (
f"[Background process {session_id} is still running~ "