feat: allow syncing indexed dataset (#1714)

Fixes LS-3592

Author: samnoyes

python/langsmith/__init__.py

@@ -20,7 +20,7 @@
     from langsmith.utils import ContextThreadPoolExecutor
 
 # Avoid calling into importlib on every call to __version__
-__version__ = "0.3.41"
+__version__ = "0.3.42"
 version = __version__  # for backwards compatibility
 
 

python/langsmith/async_client.py

@@ -1108,6 +1108,32 @@ async def index_dataset(
         )
         ls_utils.raise_for_status_with_text(resp)
 
+    @ls_beta.warn_beta
+    async def sync_indexed_dataset(
+        self,
+        *,
+        dataset_id: ls_client.ID_TYPE,
+        **kwargs: Any,
+    ) -> None:
+        """Sync dataset index. This already happens automatically every 5 minutes, but you can call this to force a sync.
+
+        Args:
+            dataset_id (UUID): The ID of the dataset to sync.
+
+        Returns:
+            None
+
+        Raises:
+            requests.HTTPError
+        """  # noqa: E501
+        dataset_id = ls_client._as_uuid(dataset_id, "dataset_id")
+        resp = await self._arequest_with_retries(
+            "POST",
+            f"/datasets/{dataset_id}/index/sync",
+            content=ls_client._dumps_json({**kwargs}),
+        )
+        ls_utils.raise_for_status_with_text(resp)
+
     @ls_beta.warn_beta
     async def similar_examples(
         self,

python/langsmith/client.py

@@ -4920,6 +4920,30 @@ def index_dataset(
         )
         ls_utils.raise_for_status_with_text(resp)
 
+    @warn_beta
+    def sync_indexed_dataset(
+        self,
+        *,
+        dataset_id: ID_TYPE,
+        **kwargs: Any,
+    ) -> None:
+        """Sync dataset index. This already happens automatically every 5 minutes, but you can call this to force a sync.
+
+        Args:
+            dataset_id (Union[UUID, str]): The ID of the dataset to sync.
+
+        Returns:
+            None
+        """  # noqa: E501
+        dataset_id = _as_uuid(dataset_id, "dataset_id")
+        resp = self.request_with_retries(
+            "POST",
+            f"/datasets/{dataset_id}/index/sync",
+            headers=self._headers,
+            data=json.dumps({**kwargs}),
+        )
+        ls_utils.raise_for_status_with_text(resp)
+
     # NOTE: dataset_name arg explicitly not supported to avoid extra API calls.
     @warn_beta
     def similar_examples(

python/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "langsmith"
-version = "0.3.41"
+version = "0.3.42"
 description = "Client library to connect to the LangSmith LLM Tracing and Evaluation Platform."
 authors = ["LangChain <[email protected]>"]
 license = "MIT"

python/tests/integration_tests/test_async_client.py

@@ -44,6 +44,27 @@ async def check_similar_examples():
                 {"name": "Alice", "age": 30}, dataset_id=dataset.id, limit=1
             )
             assert examples[0].id == example.id
+
+            example2 = await client.create_example(
+                inputs={"name": "Bobby", "age": 30},
+                outputs={"hi": "there"},
+                dataset_id=dataset.id,
+            )
+
+            await client.sync_indexed_dataset(dataset_id=dataset.id)
+
+            async def check_similar_examples():
+                examples = await client.similar_examples(
+                    {"name": "Bobby", "age": 30}, dataset_id=dataset.id, limit=2
+                )
+                return len(examples) == 2
+
+            await wait_for(check_similar_examples, timeout=20)
+            examples = await client.similar_examples(
+                {"name": "Bobby", "age": 30}, dataset_id=dataset.id, limit=2
+            )
+            assert examples[0].id == example2.id
+            assert examples[1].id == example.id
         finally:
             if dataset:
                 await client.delete_dataset(dataset_id=dataset.id)

python/tests/integration_tests/test_client.py

@@ -389,6 +389,20 @@ def test_similar_examples(langchain_client: Client) -> None:
     )
     assert len(similar_list) == 2
 
+    langchain_client.create_example(
+        inputs={"text": "howdy"},
+        outputs={"response": "howdy"},
+        dataset_id=dataset.id,
+    )
+
+    langchain_client.sync_indexed_dataset(dataset_id=dataset.id)
+    time.sleep(5)
+
+    similar_list = langchain_client.similar_examples(
+        {"text": "howdy"}, limit=5, dataset_id=dataset.id
+    )
+    assert len(similar_list) == 4
+
     langchain_client.delete_dataset(dataset_id=dataset.id)