mirror of
https://github.com/usestrix/strix.git
synced 2026-08-25 20:32:38 +02:00
fix: recoverable guardrail blocks and decoupled crash-notify (#919)
Co-authored-by: Ahmed Allam <ahmed39652003@gmail.com>
This commit is contained in:
co-authored by
Ahmed Allam
parent
fade37025d
commit
e037d8d727
+30
-1
@@ -21,6 +21,7 @@ from openai import (
|
||||
RateLimitError,
|
||||
)
|
||||
|
||||
from strix.config import codex
|
||||
from strix.core.hooks import (
|
||||
BudgetExceededError,
|
||||
BudgetPausedError,
|
||||
@@ -88,6 +89,11 @@ async def _compact_session(
|
||||
)
|
||||
|
||||
|
||||
_GUARDRAIL_PARK_ERROR = (
|
||||
"Blocked by the model's content guardrail (flagged as a possible cybersecurity risk). "
|
||||
"Set STRIX_LLM to a model that isn't blocked and resume the scan to continue."
|
||||
)
|
||||
|
||||
_TRANSIENT_MODEL_STATUS_CODES = frozenset({408, 500, 502, 503, 504})
|
||||
_MAX_TRANSIENT_MODEL_RETRIES = 4
|
||||
_TRANSIENT_MODEL_RETRY_BASE_DELAY_S = 2.0
|
||||
@@ -304,6 +310,7 @@ async def respawn_subagents(
|
||||
if coordinator.parent_of.get(aid) is None or aid == root_id:
|
||||
continue
|
||||
md["_restored_status"] = status
|
||||
md["_restored_error"] = coordinator.errors.get(aid)
|
||||
candidates.append(
|
||||
(
|
||||
aid,
|
||||
@@ -316,7 +323,8 @@ async def respawn_subagents(
|
||||
for child_id, name, parent_id, md in candidates:
|
||||
try:
|
||||
restored_status = str(md.get("_restored_status") or "running")
|
||||
start_parked = interactive and restored_status != "running"
|
||||
recoverable_park = restored_status == "waiting" and bool(md.get("_restored_error"))
|
||||
start_parked = interactive and restored_status != "running" and not recoverable_park
|
||||
|
||||
if start_parked:
|
||||
logger.warning(
|
||||
@@ -572,6 +580,10 @@ async def _run_cycle( # noqa: PLR0912, PLR0915
|
||||
if session is not None:
|
||||
input_data = []
|
||||
continue
|
||||
if codex.is_content_guardrail_error(exc):
|
||||
return await _handle_content_guardrail(
|
||||
coordinator, agent_id, exc, interactive=interactive
|
||||
)
|
||||
if not interactive:
|
||||
raise
|
||||
if isinstance(exc, MaxTurnsExceeded):
|
||||
@@ -589,6 +601,22 @@ async def _run_cycle( # noqa: PLR0912, PLR0915
|
||||
return stream
|
||||
|
||||
|
||||
async def _handle_content_guardrail(
|
||||
coordinator: AgentCoordinator,
|
||||
agent_id: str,
|
||||
exc: BaseException,
|
||||
*,
|
||||
interactive: bool,
|
||||
) -> RunResultBase | None:
|
||||
logger.warning("agent %s blocked by the model's content guardrail: %s", agent_id, exc)
|
||||
if interactive:
|
||||
await coordinator.set_status(agent_id, "waiting", error=_GUARDRAIL_PARK_ERROR)
|
||||
return None
|
||||
await coordinator.set_status(agent_id, "failed", error=_GUARDRAIL_PARK_ERROR)
|
||||
await _notify_parent_on_terminal(coordinator, agent_id, "failed")
|
||||
return None
|
||||
|
||||
|
||||
async def _settle_run_result(
|
||||
coordinator: AgentCoordinator,
|
||||
agent_id: str,
|
||||
@@ -685,6 +713,7 @@ async def _notify_parent_on_terminal(
|
||||
"priority": "high",
|
||||
"content": template.format(name=name, agent_id=agent_id),
|
||||
},
|
||||
interrupt=False,
|
||||
)
|
||||
|
||||
|
||||
|
||||
Reference in New Issue
Block a user