open-swe/agent/middleware/timeout_wrapup.py

65 lines
2.3 KiB
Python
Raw Normal View History

feat: port durable dispatch hardening and startup latency improvements (#160) * feat: port plan-review & workflow-approval UX (#135) Port six upstream commits onto dev: - c03a6be7 (already ported): keep plan guidance high-level - 546042a4: add workflow approval UI with diff preview, approval URLs, web review links, and polling for approval status during active runs - 216cf181: remove workflow token elevation; approved pushes pass through directly without proxy token rewriting - 3dbc0282: preserve plan redirects after login by accepting relative same-origin redirect_to values and rejecting blocked paths - bb104d93: submit plan comments with cmd+enter - 90cb6caa: terse Slack replies, shared content via save_plan outside plan mode (PLAN_STATUS_SHARED), reject shared-content mutations Refs: #135 * feat: port durable dispatch hardening and startup latency improvements Port five upstream PRs onto dev: - #1621 / #1658: durable dispatch with loopback webhook defense, create_durable_run helper, _config_with_prepare_run_id, degradation to None for relative/loopback completion webhook URLs - #1696: run-level completion webhook deduplication (replace claim-then-post with post-then-flag per run_id), DeferredErrorModel for graph-factory resilience, ToolRetryMiddleware for task subagents, TimeoutWrapupMiddleware for all three graphs - #1697: lazy-load __init__.py for agent.middleware, agent.tools, agent.dashboard (PEP 562); defer heavy imports (exa_py in web_search, agent.webapp in request_pr_review, deepagents in sandbox.py); add ttl_cache.py with stale-while-revalidate for tool loaders Refs: #137 * fix: restore login page render and clear CI lint/format The plan-review port removed the authRedirectUrl import from login.tsx but left its call site, crashing the login page at runtime (blank page, no 'Sign in to open-swe'). Pass the relative path straight to loginUrl, matching the plan route and the backend relative-redirect handling. Also drop an unused os import in the guard test and reformat workflow_push_guard.py to satisfy ruff. * fix: restore RepairOrphaned middleware export and repoint model fake to deferred_model boundary * fix: restore RepairOrphanedToolCallsMiddleware, fix E2E model-fake patch, drop dead ttl_cache - Re-add RepairOrphanedToolCallsMiddleware to the lazy middleware __init__ (_MIDDLEWARE_MODULES, __all__, TYPE_CHECKING) so agent.reviewer can import it. - Reroute E2E model patching to deferred_model.make_model so make_model_or_defer (used by all three graph factories) returns the scripted fake instead of building a real model with fake credentials. - Drop unused agent/utils/ttl_cache.py — no agent module imports it. - Fix import ordering in agent/reviewer.py and agent/analyzer.py (ruff I001). - Format tests/test_dispatch.py. * fix: claim-then-post run-level failure dedup; stop permanent suppression --------- Co-authored-by: amoussa1229 <166072409+amoussa1229@users.noreply.github.com> Co-authored-by: Adam Moussa <adam@seahavenind.com>
2026-07-09 17:11:25 -04:00
from __future__ import annotations
import os
import time
from collections.abc import Awaitable, Callable
from langchain.agents.middleware.types import AgentMiddleware, ModelRequest, ModelResponse
from langchain_core.messages import BaseMessage, SystemMessage
_DEFAULT_TIMEOUT_SECONDS = 45 * 60
_WRAPUP_INSTRUCTION = """
<time_limit_warning>
You have been running for a long time. Wrap up immediately: finish the current
step, save or report useful state, avoid starting new investigations, and end
your turn with the best available result.
</time_limit_warning>
"""
def _configured_timeout_seconds() -> int:
raw = os.environ.get("OPEN_SWE_WRAPUP_TIMEOUT_SECONDS")
if not raw:
return _DEFAULT_TIMEOUT_SECONDS
try:
value = int(raw)
except ValueError:
return _DEFAULT_TIMEOUT_SECONDS
return value if value > 0 else _DEFAULT_TIMEOUT_SECONDS
def _content_with_instruction(message: BaseMessage | None, instruction: str) -> str | list[object]:
if message is None:
return instruction
content = message.content
if isinstance(content, list):
return [*content, {"type": "text", "text": instruction}]
return f"{content}\n\n{instruction}" if content else instruction
class TimeoutWrapupMiddleware(AgentMiddleware):
def __init__(self, timeout_seconds: int | None = None) -> None:
super().__init__()
self._timeout_seconds = timeout_seconds or _configured_timeout_seconds()
# Graph construction should create one middleware instance per run; start
# lazily so construction-time caching cannot age the run clock.
self._start: float | None = None
def _should_wrapup(self) -> bool:
if self._start is None:
self._start = time.monotonic()
return (time.monotonic() - self._start) >= self._timeout_seconds
def _apply(self, request: ModelRequest) -> ModelRequest:
if not self._should_wrapup():
return request
content = _content_with_instruction(request.system_message, _WRAPUP_INSTRUCTION)
return request.override(system_message=SystemMessage(content=content))
async def awrap_model_call(
self,
request: ModelRequest,
handler: Callable[[ModelRequest], Awaitable[ModelResponse]],
) -> ModelResponse:
return await handler(self._apply(request))