Send collected outputs for generators

python/langsmith/run_helpers.py

@@ -1,4 +1,4 @@
 """Decorator for creating a run tree from functions."""
 
 from __future__ import annotations
 
@@ -570,22 +570,18 @@
                     yield item
             except BaseException as e:
                 await asyncio.shield(
-                    aitertools.aio_to_thread(_on_run_end, run_container, error=e)
+                    aitertools.aio_to_thread(
+                        _on_run_end,
+                        run_container,
+                        error=e,
+                        outputs=_get_function_result(results, reduce_fn),
+                    )
                 )
                 raise e
-            if results:
-                if reduce_fn:
-                    try:
-                        function_result = reduce_fn(results)
-                    except BaseException as e:
-                        LOGGER.error(e)
-                        function_result = results
-                else:
-                    function_result = results
-            else:
-                function_result = None
             await aitertools.aio_to_thread(
-                _on_run_end, run_container, outputs=function_result
+                _on_run_end,
+                run_container,
+                outputs=_get_function_result(results, reduce_fn),
             )
 
         @functools.wraps(func)
@@ -652,21 +648,13 @@
                     results.append(function_return)
 
             except BaseException as e:
-                _on_run_end(run_container, error=e)
+                _on_run_end(
+                    run_container,
+                    error=e,
+                    outputs=_get_function_result(results, reduce_fn),
+                )
                 raise e
-
-            if results:
-                if reduce_fn:
-                    try:
-                        function_result = reduce_fn(results)
-                    except BaseException as e:
-                        LOGGER.error(e)
-                        function_result = results
-                else:
-                    function_result = results
-            else:
-                function_result = None
-            _on_run_end(run_container, outputs=function_result)
+            _on_run_end(run_container, outputs=_get_function_result(results, reduce_fn))
             return function_return
 
         # "Stream" functions (used in methods like OpenAI/Anthropic's SDKs)
@@ -1709,3 +1697,15 @@
             return await self.__ls_stream__.__aexit__(exc_type, exc_val, exc_tb)
         finally:
             await self._aend_trace()
+
+
+def _get_function_result(results: list, reduce_fn: Callable) -> Any:
+    if results:
+        if reduce_fn is not None:
+            try:
+                return reduce_fn(results)
+            except BaseException as e:
+                LOGGER.error(e)
+                return results
+        else:
+            return results

python/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "langsmith"
-version = "0.1.132"
+version = "0.1.133"
 description = "Client library to connect to the LangSmith LLM Tracing and Evaluation Platform."
 authors = ["LangChain <[email protected]>"]
 license = "MIT"

python/tests/unit_tests/test_run_helpers.py

@@ -7,7 +7,7 @@
 import time
 import uuid
 import warnings
-from typing import Any, AsyncGenerator, Generator, Optional, Set, cast
+from typing import Any, AsyncGenerator, Generator, List, Optional, Set, Tuple, cast
 from unittest.mock import MagicMock, patch
 
 import pytest
@@ -50,6 +50,17 @@ def _get_calls(
     return calls
 
 
+def _get_datas(mock_calls: List[Any]) -> List[Tuple[str, dict]]:
+    datas = []
+    for call_ in mock_calls:
+        data = json.loads(call_.kwargs["data"])
+        for verb in ("post", "patch"):
+            for payload in data.get(verb) or []:
+                datas.append((verb, payload))
+
+    return datas
+
+
 def test__get_inputs_with_no_args() -> None:
     def foo() -> None:
         pass
@@ -1466,7 +1477,53 @@ async def my_function(a: int) -> AsyncGenerator[int, None]:
     mock_calls = _get_calls(
         mock_client, verbs={"POST", "PATCH", "GET"}, minimum=num_calls
     )
+
+    assert len(mock_calls) == num_calls
+    if auto_batch_tracing:
+        datas = _get_datas(mock_calls)
+        outputs = [p["outputs"] for _, p in datas if p.get("outputs")]
+        assert len(outputs) == 1
+        assert outputs[0]["output"] == list(range(5))
+
+
+@pytest.mark.parametrize("auto_batch_tracing", [True, False])
+async def test_traceable_gen_exception(auto_batch_tracing: bool):
+    mock_client = _get_mock_client(
+        auto_batch_tracing=auto_batch_tracing,
+        info=ls_schemas.LangSmithInfo(
+            batch_ingest_config=ls_schemas.BatchIngestConfig(
+                size_limit_bytes=None,  # Note this field is not used here
+                size_limit=100,
+                scale_up_nthreads_limit=16,
+                scale_up_qsize_trigger=1000,
+                scale_down_nempty_trigger=4,
+            )
+        ),
+    )
+
+    @traceable
+    def my_function(a: int) -> Generator[int, None, None]:
+        for i in range(5):
+            yield i
+        raise ValueError("foo")
+
+    with tracing_context(enabled=True):
+        with pytest.raises(ValueError, match="foo"):
+            for _ in my_function(1, langsmith_extra={"client": mock_client}):
+                pass
+
+    # Get ALL the call args for the mock_client
+    num_calls = 1 if auto_batch_tracing else 2
+    mock_calls = _get_calls(
+        mock_client, verbs={"POST", "PATCH", "GET"}, minimum=num_calls
+    )
+
     assert len(mock_calls) == num_calls
+    if auto_batch_tracing:
+        datas = _get_datas(mock_calls)
+        outputs = [p["outputs"] for _, p in datas if p.get("outputs")]
+        assert len(outputs) == 1
+        assert outputs[0]["output"] == list(range(5))
 
 
 @pytest.mark.parametrize("env_var", [True, False])