Shared Executor (#11028)

libs/langchain/langchain/callbacks/tracers/evaluation.py

@@ -2,20 +2,33 @@
 from __future__ import annotations
 
 import logging
-from concurrent.futures import Future, ThreadPoolExecutor
-from typing import Any, Dict, List, Optional, Sequence, Set, Union
+import weakref
+from concurrent.futures import Future, wait
+from typing import Any, Dict, List, Optional, Sequence, Union
 from uuid import UUID
 
 import langsmith
+from langsmith import schemas as langsmith_schemas
 from langsmith.evaluation.evaluator import EvaluationResult
 
 from langchain.callbacks import manager
 from langchain.callbacks.tracers import langchain as langchain_tracer
 from langchain.callbacks.tracers.base import BaseTracer
+from langchain.callbacks.tracers.langchain import _get_executor
 from langchain.callbacks.tracers.schemas import Run
 
 logger = logging.getLogger(__name__)
 
+_TRACERS: weakref.WeakSet[EvaluatorCallbackHandler] = weakref.WeakSet()
+
+
+def wait_for_all_evaluators() -> None:
+    """Wait for all tracers to finish."""
+    global _TRACERS
+    for tracer in list(_TRACERS):
+        if tracer is not None:
+            tracer.wait_for_futures()
+
 
 class EvaluatorCallbackHandler(BaseTracer):
     """A tracer that runs a run evaluator whenever a run is persisted.
@@ -24,9 +37,6 @@ class EvaluatorCallbackHandler(BaseTracer):
     ----------
     evaluators : Sequence[RunEvaluator]
         The run evaluators to apply to all top level runs.
-    max_workers : int, optional
-        The maximum number of worker threads to use for running the evaluators.
-        If not specified, it will default to the number of evaluators.
     client : LangSmith Client, optional
         The LangSmith client instance to use for evaluating the runs.
         If not specified, a new instance will be created.
@@ -59,7 +69,6 @@ class EvaluatorCallbackHandler(BaseTracer):
     def __init__(
         self,
         evaluators: Sequence[langsmith.RunEvaluator],
-        max_workers: Optional[int] = None,
         client: Optional[langsmith.Client] = None,
         example_id: Optional[Union[UUID, str]] = None,
         skip_unfinished: bool = True,
@@ -72,11 +81,14 @@ class EvaluatorCallbackHandler(BaseTracer):
         )
         self.client = client or langchain_tracer.get_client()
         self.evaluators = evaluators
-        self.max_workers = max_workers or len(evaluators)
-        self.futures: Set[Future] = set()
+        self.executor = _get_executor()
+        self.futures: weakref.WeakSet[Future] = weakref.WeakSet()
         self.skip_unfinished = skip_unfinished
         self.project_name = project_name
+        self.logged_feedback: Dict[str, List[langsmith_schemas.Feedback]] = {}
         self.logged_eval_results: Dict[str, List[EvaluationResult]] = {}
+        global _TRACERS
+        _TRACERS.add(self)
 
     def _evaluate_in_project(self, run: Run, evaluator: langsmith.RunEvaluator) -> None:
         """Evaluate the run in the project.
@@ -120,15 +132,11 @@ class EvaluatorCallbackHandler(BaseTracer):
             return
         run_ = run.copy()
         run_.reference_example_id = self.example_id
-        if self.max_workers > 0:
-            with ThreadPoolExecutor(max_workers=self.max_workers) as executor:
-                list(
-                    executor.map(
-                        self._evaluate_in_project,
-                        [run_ for _ in range(len(self.evaluators))],
-                        self.evaluators,
-                    )
-                )
-        else:
-            for evaluator in self.evaluators:
-                self._evaluate_in_project(run_, evaluator)
+        for evaluator in self.evaluators:
+            self.futures.add(
+                self.executor.submit(self._evaluate_in_project, run_, evaluator)
+            )
+
+    def wait_for_futures(self) -> None:
+        """Wait for all futures to complete."""
+        wait(self.futures)

libs/langchain/langchain/callbacks/tracers/langchain.py

@@ -6,7 +6,7 @@ import os
 import weakref
 from concurrent.futures import Future, ThreadPoolExecutor, wait
 from datetime import datetime
-from typing import Any, Callable, Dict, List, Optional, Set, Union
+from typing import Any, Callable, Dict, List, Optional, Union
 from uuid import UUID
 
 from langsmith import Client
@@ -21,8 +21,7 @@ logger = logging.getLogger(__name__)
 _LOGGED = set()
 _TRACERS: weakref.WeakSet[LangChainTracer] = weakref.WeakSet()
 _CLIENT: Optional[Client] = None
-_MAX_EXECUTORS = 10  # TODO: Remove once write queue is implemented
-_EXECUTORS: List[ThreadPoolExecutor] = []
+_EXECUTOR: Optional[ThreadPoolExecutor] = None
 
 
 def log_error_once(method: str, exception: Exception) -> None:
@@ -50,6 +49,14 @@ def get_client() -> Client:
     return _CLIENT
 
 
+def _get_executor() -> ThreadPoolExecutor:
+    """Get the executor."""
+    global _EXECUTOR
+    if _EXECUTOR is None:
+        _EXECUTOR = ThreadPoolExecutor()
+    return _EXECUTOR
+
+
 class LangChainTracer(BaseTracer):
     """An implementation of the SharedTracer that POSTS to the langchain endpoint."""
 
@@ -71,21 +78,10 @@ class LangChainTracer(BaseTracer):
         self.project_name = project_name or os.getenv(
             "LANGCHAIN_PROJECT", os.getenv("LANGCHAIN_SESSION", "default")
         )
-        if use_threading:
-            global _MAX_EXECUTORS
-            if len(_EXECUTORS) < _MAX_EXECUTORS:
-                self.executor: Optional[ThreadPoolExecutor] = ThreadPoolExecutor(
-                    max_workers=1
-                )
-                _EXECUTORS.append(self.executor)
-            else:
-                self.executor = _EXECUTORS.pop(0)
-                _EXECUTORS.append(self.executor)
-        else:
-            self.executor = None
         self.client = client or get_client()
-        self._futures: Set[Future] = set()
+        self._futures: weakref.WeakSet[Future] = weakref.WeakSet()
         self.tags = tags or []
+        self.executor = _get_executor() if use_threading else None
         global _TRACERS
         _TRACERS.add(self)
 
@@ -229,7 +225,4 @@ class LangChainTracer(BaseTracer):
 
     def wait_for_futures(self) -> None:
         """Wait for the given futures to complete."""
-        futures = list(self._futures)
-        wait(futures)
-        for future in futures:
-            self._futures.remove(future)
+        wait(self._futures)

libs/langchain/langchain/smith/evaluation/runner_utils.py

@@ -24,8 +24,11 @@ from langsmith import Client, RunEvaluator
 from langsmith.schemas import Dataset, DataType, Example
 
 from langchain.callbacks.manager import Callbacks
-from langchain.callbacks.tracers.evaluation import EvaluatorCallbackHandler
-from langchain.callbacks.tracers.langchain import LangChainTracer, wait_for_all_tracers
+from langchain.callbacks.tracers.evaluation import (
+    EvaluatorCallbackHandler,
+    wait_for_all_evaluators,
+)
+from langchain.callbacks.tracers.langchain import LangChainTracer
 from langchain.chains.base import Chain
 from langchain.evaluation.loading import load_evaluator
 from langchain.evaluation.schema import (
@@ -915,7 +918,6 @@ def _prepare_run_on_dataset(
                 EvaluatorCallbackHandler(
                     evaluators=run_evaluators or [],
                     client=client,
-                    max_workers=0,
                     example_id=example.id,
                 ),
                 progress_bar,
@@ -934,7 +936,7 @@ def _collect_test_results(
     configs: List[RunnableConfig],
     project_name: str,
 ) -> TestResult:
-    wait_for_all_tracers()
+    wait_for_all_evaluators()
     all_eval_results = {}
     for c in configs:
         for callback in cast(list, c["callbacks"]):