ForcePilot/backend/package/yuxi/services/run_worker.py
2026-06-05 21:02:13 +08:00

530 lines
19 KiB
Python

"""ARQ worker for agent runs."""
from __future__ import annotations
import asyncio
import json
import os
import time
from dataclasses import dataclass, field
from sqlalchemy import select
from sqlalchemy.exc import OperationalError
from yuxi.agents.mcp.service import ensure_builtin_mcp_servers_in_db
from yuxi.agents.skills.service import init_builtin_skills
from yuxi.repositories.agent_run_repository import TERMINAL_RUN_STATUSES, AgentRunRepository
from yuxi.services.chat_service import stream_agent_chat, stream_agent_resume
from yuxi.services.run_queue_service import (
append_run_stream_event,
clear_cancel_signal,
has_cancel_signal,
wait_for_cancel_signal,
)
from yuxi.storage.postgres.manager import pg_manager
from yuxi.storage.postgres.models_business import User
from yuxi.utils.logging_config import logger
LOADING_FLUSH_INTERVAL_MS = 100
LOADING_FLUSH_MAX_CHARS = 512
RUN_CANCEL_POLL_SECONDS = 0.2
REDIS_URL = os.getenv("REDIS_URL", "redis://redis:6379/0")
class RetryableRunError(Exception):
"""Error type that should trigger ARQ retry."""
class NonRetryableRunError(Exception):
"""Error type that should not trigger ARQ retry."""
@dataclass
class RunContext:
run_id: str
cancel_event: asyncio.Event = field(default_factory=asyncio.Event)
_watch_task: asyncio.Task | None = None
async def start(self) -> None:
if self._watch_task is None:
self._watch_task = asyncio.create_task(self._watch_cancel_signal())
async def close(self) -> None:
if self._watch_task:
self._watch_task.cancel()
await asyncio.gather(self._watch_task, return_exceptions=True)
self._watch_task = None
async def wait_cancelled(self) -> None:
await self.cancel_event.wait()
async def is_cancelled(self) -> bool:
if self.cancel_event.is_set():
return True
if await has_cancel_signal(self.run_id):
self.cancel_event.set()
return True
return False
async def _watch_cancel_signal(self) -> None:
while not self.cancel_event.is_set():
cancelled = await wait_for_cancel_signal(
self.run_id,
poll_timeout_seconds=RUN_CANCEL_POLL_SECONDS,
)
if cancelled:
self.cancel_event.set()
return
_ALL_THREADS = object()
@dataclass
class _ThreadBuffer:
items: list[dict] = field(default_factory=list)
chars: int = 0
last_flush: float = field(default_factory=time.monotonic)
class ChunkedEventWriter:
def __init__(self, run_id: str, thread_id: str | None, interval_ms: int = 100, max_chars: int = 512):
self.run_id = run_id
self.thread_id = thread_id
self.interval_seconds = interval_ms / 1000
self.max_chars = max_chars
self.thread_buffers: dict[str | None, _ThreadBuffer] = {}
def _target_thread_id(self, thread_id: str | None = None) -> str | None:
return thread_id or self.thread_id
async def append(self, chunk: dict, *, thread_id: str | None = None):
target_thread_id = self._target_thread_id(thread_id or _thread_id_from_mapping(chunk))
buffer = self.thread_buffers.setdefault(target_thread_id, _ThreadBuffer())
buffer.items.append(chunk)
buffer.chars += _loading_chunk_size(chunk)
if _flush_loading_chunk_immediately(chunk):
await self.flush(target_thread_id)
return
if (time.monotonic() - buffer.last_flush) >= self.interval_seconds or buffer.chars >= self.max_chars:
await self.flush(target_thread_id)
async def flush(self, thread_id: str | None | object = _ALL_THREADS):
if thread_id is _ALL_THREADS:
for target_thread_id in list(self.thread_buffers):
await self.flush(target_thread_id)
return
buffer = self.thread_buffers.get(thread_id)
if not buffer or not buffer.items:
return
await append_run_event(self.run_id, "messages", {"items": buffer.items}, thread_id=thread_id)
buffer.items = []
buffer.chars = 0
buffer.last_flush = time.monotonic()
async def _get_run(run_id: str):
async with pg_manager.get_async_session_context() as db:
repo = AgentRunRepository(db)
return await repo.get_run(run_id)
async def append_run_event(run_id: str, event_type: str, payload: dict, *, thread_id: str | None = None):
await append_run_stream_event(run_id, event_type, payload, thread_id=thread_id)
async def mark_run_running(run_id: str):
async with pg_manager.get_async_session_context() as db:
repo = AgentRunRepository(db)
await repo.mark_running(run_id)
async def mark_run_terminal(run_id: str, status: str, error_type: str | None = None, error_message: str | None = None):
async with pg_manager.get_async_session_context() as db:
repo = AgentRunRepository(db)
await repo.set_terminal_status(run_id, status=status, error_type=error_type, error_message=error_message)
async def _load_user(uid: str):
async with pg_manager.get_async_session_context() as db:
result = await db.execute(select(User).where(User.uid == uid, User.is_deleted == 0))
return result.scalar_one_or_none()
async def _is_cancel_requested(run_id: str) -> bool:
run = await _get_run(run_id)
return bool(run and run.status == "cancel_requested")
def _job_try(ctx) -> int:
if isinstance(ctx, dict):
try:
return int(ctx.get("job_try") or 1)
except Exception:
return 1
return 1
def _is_last_try(ctx) -> bool:
return _job_try(ctx) >= max(1, int(getattr(WorkerSettings, "max_tries", 1)))
def _is_retryable_exception(exc: Exception) -> bool:
if isinstance(exc, NonRetryableRunError):
return False
return isinstance(exc, (RetryableRunError, OperationalError, ConnectionError, TimeoutError, asyncio.TimeoutError))
def _iter_json_chunks(chunk_bytes: bytes) -> list[dict]:
text = chunk_bytes.decode("utf-8")
chunks: list[dict] = []
for line in text.splitlines():
line = line.strip()
if not line:
continue
try:
chunks.append(json.loads(line))
except Exception:
logger.warning(f"Failed to parse run stream chunk: {line[:200]}")
return chunks
def _thread_id_from_mapping(value: object) -> str | None:
if not isinstance(value, dict):
return None
thread_id = value.get("thread_id")
if isinstance(thread_id, str) and thread_id.strip():
return thread_id.strip()
for key in ("meta", "metadata", "configurable", "stream_event"):
nested = value.get(key)
if isinstance(nested, dict):
nested_thread_id = _thread_id_from_mapping(nested)
if nested_thread_id:
return nested_thread_id
return None
def _loading_chunk_size(chunk: dict) -> int:
response = chunk.get("response")
total = len(response) if isinstance(response, str) else 0
stream_event = chunk.get("stream_event")
if not isinstance(stream_event, dict):
return total
for key in ("content", "reasoning_content", "additional_reasoning_content", "args_delta"):
value = stream_event.get(key)
if isinstance(value, str):
total += len(value)
return total
def _flush_loading_chunk_immediately(chunk: dict) -> bool:
stream_event = chunk.get("stream_event")
return isinstance(stream_event, dict) and stream_event.get("type") == "tool_call"
def _chunk_thread_id(chunk: dict, fallback: str | None) -> str | None:
return _thread_id_from_mapping(chunk) or fallback
def _map_chunk_to_run_event(chunk: dict) -> tuple[str, dict]:
status = chunk.get("status") or "event"
if status == "loading":
return "messages", {"chunk": chunk}
if status == "agent_state":
return "custom", {"name": "yuxi.agent_state", "chunk": chunk, "agent_state": chunk.get("agent_state") or {}}
if status in {"ask_user_question_required", "human_approval_required", "interrupted"}:
reason = "human_approval" if status == "human_approval_required" else status
return "interrupt", {"reason": reason, "chunk": chunk}
if status == "warning":
return "custom", {"name": "yuxi.warning", "chunk": chunk}
if status == "error":
return "error", {"chunk": chunk, "retryable": bool(chunk.get("retryable"))}
if status == "finished":
return "end", {"status": "completed", "chunk": chunk}
return "custom", {"name": f"yuxi.{status}", "chunk": chunk}
async def _append_end_event(run_id: str, status: str, *, thread_id: str | None, payload: dict | None = None):
end_payload = {"status": status}
if payload:
end_payload.update(payload)
await append_run_event(run_id, "end", end_payload, thread_id=thread_id)
async def _consume_stream_with_cancel(agen, run_ctx: RunContext):
while True:
next_task = asyncio.create_task(agen.__anext__())
cancel_task = asyncio.create_task(run_ctx.wait_cancelled())
done, _ = await asyncio.wait({next_task, cancel_task}, return_when=asyncio.FIRST_COMPLETED)
if cancel_task in done:
next_task.cancel()
await asyncio.gather(next_task, return_exceptions=True)
raise asyncio.CancelledError(f"run {run_ctx.run_id} cancelled")
cancel_task.cancel()
await asyncio.gather(cancel_task, return_exceptions=True)
try:
yield next_task.result()
except StopAsyncIteration:
return
async def process_agent_run(ctx, run_id: str):
run = await _get_run(run_id)
if not run:
logger.warning(f"Run not found: {run_id}")
return
if run.status in TERMINAL_RUN_STATUSES:
logger.info(f"Run already terminal, skip: {run_id}, status={run.status}")
return
payload = run.input_payload or {}
query = payload.get("query")
resume_input = payload.get("resume")
run_type = payload.get("run_type") or "chat"
config = payload.get("config") or {}
agent_id = payload.get("agent_id")
image_content = payload.get("image_content")
uid = payload.get("uid")
request_id = payload.get("request_id")
thread_id = config.get("thread_id") or payload.get("thread_id")
user = await _load_user(uid)
if not user:
await mark_run_terminal(run_id, "failed", "user_not_found", f"user {uid} not found")
return
if not request_id:
request_id = run.request_id
meta = {
"run_id": run_id,
"request_id": request_id,
"query": query,
"agent_id": agent_id,
"server_model_name": config.get("model", agent_id),
"thread_id": config.get("thread_id"),
"uid": user.uid,
"has_image": bool(image_content),
"attachment_file_ids": payload.get("attachment_file_ids") or [],
"model_spec": payload.get("model_spec"),
}
await mark_run_running(run_id)
run_ctx = RunContext(run_id=run_id)
writer = ChunkedEventWriter(
run_id=run_id,
thread_id=thread_id,
interval_ms=LOADING_FLUSH_INTERVAL_MS,
max_chars=LOADING_FLUSH_MAX_CHARS,
)
await run_ctx.start()
await append_run_event(
run_id,
"metadata",
{
"request_id": request_id,
"agent_id": agent_id,
"backend_id": payload.get("backend_id"),
"uid": uid,
},
thread_id=thread_id,
)
terminal_set = False
try:
async with pg_manager.get_async_session_context() as db:
if run_type == "resume":
stream = stream_agent_resume(
thread_id=thread_id,
resume_input=resume_input,
meta=meta,
current_user=user,
db=db,
)
else:
stream = stream_agent_chat(
query=query,
agent_id=config.get("agent_id") or agent_id,
thread_id=thread_id,
meta=meta,
image_content=image_content,
current_user=user,
db=db,
save_user_message=False,
)
async for chunk_bytes in _consume_stream_with_cancel(stream, run_ctx):
for chunk in _iter_json_chunks(chunk_bytes):
target_thread_id = _chunk_thread_id(chunk, thread_id)
if chunk.get("status") == "loading":
await writer.append(chunk, thread_id=target_thread_id)
continue
await writer.flush(target_thread_id)
status = chunk.get("status") or "event"
event_type, event_payload = _map_chunk_to_run_event(chunk)
if event_type != "end":
await append_run_event(run_id, event_type, event_payload, thread_id=target_thread_id)
if target_thread_id != thread_id:
if await run_ctx.is_cancelled():
raise asyncio.CancelledError(f"run {run_id} cancelled")
continue
if status == "finished":
await mark_run_terminal(run_id, "completed")
await _append_end_event(run_id, "completed", thread_id=thread_id, payload={"chunk": chunk})
terminal_set = True
elif status == "error":
await mark_run_terminal(
run_id,
"failed",
error_type=chunk.get("error_type") or "stream_error",
error_message=chunk.get("error_message") or chunk.get("message"),
)
await _append_end_event(run_id, "failed", thread_id=thread_id, payload={"chunk": chunk})
terminal_set = True
elif status == "interrupted":
status_value = "cancelled" if await _is_cancel_requested(run_id) else "interrupted"
await mark_run_terminal(
run_id,
status_value,
error_type=status_value,
error_message=chunk.get("message"),
)
await _append_end_event(run_id, status_value, thread_id=thread_id, payload={"chunk": chunk})
terminal_set = True
elif status in {"ask_user_question_required", "human_approval_required"}:
questions = chunk.get("questions") if isinstance(chunk, dict) else None
first_question = ""
if isinstance(questions, list) and questions:
first = questions[0]
if isinstance(first, dict):
first_question = str(first.get("question") or "").strip()
await mark_run_terminal(
run_id,
"interrupted",
error_type=status,
error_message=first_question or "需要用户回答问题",
)
await _append_end_event(run_id, "interrupted", thread_id=thread_id, payload={"chunk": chunk})
terminal_set = True
if await run_ctx.is_cancelled():
raise asyncio.CancelledError(f"run {run_id} cancelled")
await writer.flush()
if not terminal_set:
finished_chunk = {"status": "finished", "request_id": request_id}
await mark_run_terminal(run_id, "completed")
await _append_end_event(run_id, "completed", thread_id=thread_id, payload={"chunk": finished_chunk})
except asyncio.CancelledError:
await writer.flush()
cancel_chunk = {"status": "interrupted", "message": "对话已取消", "request_id": request_id}
await append_run_event(
run_id,
"interrupt",
{"reason": "cancelled", "chunk": cancel_chunk},
thread_id=thread_id,
)
await mark_run_terminal(run_id, "cancelled", error_type="cancelled", error_message="对话已取消")
await _append_end_event(run_id, "cancelled", thread_id=thread_id, payload={"chunk": cancel_chunk})
logger.info(f"Run cancelled: {run_id}")
except Exception as e:
await writer.flush()
if _is_retryable_exception(e):
job_try = _job_try(ctx)
logger.warning(f"Run retryable failure {run_id} (try={job_try}): {e}")
retryable_error_chunk = {
"status": "error",
"error_type": "retryable_worker_error",
"error_message": str(e),
"request_id": request_id,
"retryable": True,
"job_try": job_try,
}
await append_run_event(
run_id,
"error",
{"chunk": retryable_error_chunk, "retryable": True},
thread_id=thread_id,
)
if _is_last_try(ctx):
await mark_run_terminal(
run_id,
"failed",
error_type="retryable_worker_error",
error_message=str(e),
)
await _append_end_event(
run_id,
"failed",
thread_id=thread_id,
payload={"chunk": retryable_error_chunk},
)
logger.error(f"Run failed after retries exhausted {run_id}: {e}")
return
if isinstance(e, RetryableRunError):
raise
raise RetryableRunError(str(e)) from e
logger.error(f"Run failed {run_id}: {e}")
error_chunk = {
"status": "error",
"error_type": "worker_error",
"error_message": str(e),
"request_id": request_id,
"retryable": False,
}
await append_run_event(
run_id,
"error",
{"chunk": error_chunk, "retryable": False},
thread_id=thread_id,
)
await mark_run_terminal(run_id, "failed", error_type="worker_error", error_message=str(e))
await _append_end_event(run_id, "failed", thread_id=thread_id, payload={"chunk": error_chunk})
return
finally:
await run_ctx.close()
await clear_cancel_signal(run_id)
async def _worker_startup(ctx):
del ctx
pg_manager.initialize()
await pg_manager.create_business_tables()
await pg_manager.ensure_business_schema()
await ensure_builtin_mcp_servers_in_db()
async with pg_manager.get_async_session_context() as session:
await init_builtin_skills(session)
async def _worker_shutdown(ctx):
await pg_manager.close()
class WorkerSettings:
functions = [process_agent_run]
max_tries = 2
retry_jobs = True
job_timeout = 900
keep_result = 60
on_startup = _worker_startup
on_shutdown = _worker_shutdown
try:
from arq.connections import RedisSettings
redis_settings = RedisSettings.from_dsn(REDIS_URL)
except Exception:
redis_settings = None