mirror of
https://github.com/imartinez/privateGPT.git
synced 2026-07-17 01:48:03 +00:00
576589bdc5
* feat: add websearch in a sandbox
* fix: celery callbacks
(cherry picked from commit fe7d0ce9d5)
* fix: mypy
108 lines
3.9 KiB
Python
108 lines
3.9 KiB
Python
from __future__ import annotations
|
|
|
|
import asyncio
|
|
import logging
|
|
from collections.abc import Awaitable, Callable
|
|
|
|
logger = logging.getLogger(__name__)
|
|
|
|
BatchRunner = Callable[[list[str], int], Awaitable[list[str | Exception]]]
|
|
|
|
|
|
class _PendingScrape:
|
|
__slots__ = ("future", "url")
|
|
|
|
def __init__(self, url: str, future: asyncio.Future[str]) -> None:
|
|
self.url = url
|
|
self.future = future
|
|
|
|
|
|
class ScrapeBatcher:
|
|
"""Coalesces near-simultaneous scrape requests into one browser run.
|
|
|
|
The first request opens a batch window of ``batch_wait_seconds``; requests
|
|
arriving within it join the batch. The batch is dispatched early when it
|
|
reaches ``batch_size``, otherwise when the window closes — so N pages
|
|
requested at the same time share a single session/browser instead of
|
|
launching N in parallel. ``batch_size=1`` dispatches immediately.
|
|
"""
|
|
|
|
def __init__(
|
|
self,
|
|
*,
|
|
batch_size: int,
|
|
batch_wait_seconds: float,
|
|
run_batch: BatchRunner,
|
|
) -> None:
|
|
self._batch_size = max(1, batch_size)
|
|
self._wait = max(0.0, batch_wait_seconds)
|
|
self._run_batch = run_batch
|
|
self._pending: list[_PendingScrape] = []
|
|
self._timeout_seconds = 0
|
|
self._window: asyncio.Task[None] | None = None
|
|
self._dispatches: set[asyncio.Task[None]] = set()
|
|
self._lock = asyncio.Lock()
|
|
self._closed = False
|
|
|
|
async def submit(self, url: str, timeout_seconds: int) -> str:
|
|
future: asyncio.Future[str] = asyncio.get_running_loop().create_future()
|
|
async with self._lock:
|
|
if self._closed:
|
|
raise RuntimeError("Scrape batcher is closed")
|
|
self._pending.append(_PendingScrape(url, future))
|
|
self._timeout_seconds = max(self._timeout_seconds, timeout_seconds)
|
|
if len(self._pending) >= self._batch_size:
|
|
self._flush_locked()
|
|
elif self._window is None:
|
|
self._window = asyncio.create_task(self._close_window())
|
|
return await future
|
|
|
|
def _flush_locked(self) -> None:
|
|
"""Dispatch the pending batch. Caller must hold the lock."""
|
|
if self._window is not None:
|
|
self._window.cancel()
|
|
self._window = None
|
|
batch, self._pending = self._pending, []
|
|
timeout_seconds, self._timeout_seconds = self._timeout_seconds, 0
|
|
if batch:
|
|
task = asyncio.create_task(self._dispatch(batch, timeout_seconds))
|
|
self._dispatches.add(task)
|
|
task.add_done_callback(self._dispatches.discard)
|
|
|
|
async def _close_window(self) -> None:
|
|
await asyncio.sleep(self._wait)
|
|
async with self._lock:
|
|
self._window = None
|
|
self._flush_locked()
|
|
|
|
async def _dispatch(
|
|
self, batch: list[_PendingScrape], timeout_seconds: int
|
|
) -> None:
|
|
try:
|
|
results = await self._run_batch([p.url for p in batch], timeout_seconds)
|
|
except BaseException as exc:
|
|
for pending in batch:
|
|
if not pending.future.done():
|
|
pending.future.set_exception(exc)
|
|
return
|
|
for pending, result in zip(batch, results, strict=True):
|
|
if pending.future.done():
|
|
continue
|
|
if isinstance(result, Exception):
|
|
pending.future.set_exception(result)
|
|
else:
|
|
pending.future.set_result(result)
|
|
|
|
async def close(self) -> None:
|
|
async with self._lock:
|
|
self._closed = True
|
|
if self._window is not None:
|
|
self._window.cancel()
|
|
self._window = None
|
|
pending, self._pending = self._pending, []
|
|
for p in pending:
|
|
if not p.future.done():
|
|
p.future.set_exception(RuntimeError("Scrape batcher is closed"))
|
|
if self._dispatches:
|
|
await asyncio.gather(*self._dispatches, return_exceptions=True)
|