Add method to consistently get project (#289)

We plan to introduce an environment variable in hosted langserve that
will overwrite the project used for traces from hosted deployments. We
don't centralize how we refer to projects in any of our repos, so in
order to ensure that our tracing works properly and our project
selection is consistent, we need to introduce a shared method for
getting the project that we can use throughout our different repos

python/langsmith/client.py

@@ -661,11 +661,8 @@ def create_run(
             "project_name",
             kwargs.pop(
                 "session_name",
-                os.environ.get(
-                    # TODO: Deprecate LANGCHAIN_SESSION
-                    "LANGCHAIN_PROJECT",
-                    os.environ.get("LANGCHAIN_SESSION", "default"),
-                ),
+                # if the project is not provided, use the environment's project
+                ls_utils.get_tracer_project(),
             ),
         )
         run_create = {
@@ -925,10 +922,7 @@ def get_run_url(
         elif project_name is not None:
             session_id = self.read_project(project_name=project_name).id
         else:
-            project_name = os.environ.get(
-                "LANGCHAIN_PROJECT",
-                "default",
-            )
+            project_name = ls_utils.get_tracer_project()
             session_id = self.read_project(project_name=project_name).id
         return (
             f"{self._host_url}/o/{self._get_tenant_id()}/projects/p/{_as_uuid(session_id)}/"

python/langsmith/utils.py

@@ -4,7 +4,7 @@
 import logging
 import os
 import subprocess
-from typing import Any, Callable, Dict, List, Mapping, Tuple, Union
+from typing import Any, Callable, Dict, List, Mapping, Optional, Tuple, Union
 
 import requests
 
@@ -252,6 +252,26 @@ def is_base_message_like(obj: object) -> bool:
     )
 
 
+def get_tracer_project(return_default_value=True) -> Optional[str]:
+    """Get the project name for a LangSmith tracer."""
+    return os.environ.get(
+        # Hosted LangServe projects get precedence over all other defaults.
+        # This is to make sure that we always use the associated project
+        # for a hosted langserve deployment even if the customer sets some
+        # other project name in their environment.
+        "HOSTED_LANGSERVE_PROJECT_NAME",
+        os.environ.get(
+            "LANGCHAIN_PROJECT",
+            os.environ.get(
+                # This is the legacy name for a LANGCHAIN_PROJECT, so it
+                # has lower precedence than LANGCHAIN_PROJECT
+                "LANGCHAIN_SESSION",
+                "default" if return_default_value else None,
+            ),
+        ),
+    )
+
+
 class FilterPoolFullWarning(logging.Filter):
     """Filter urrllib3 warnings logged when the connection pool isn't reused."""
 

python/pyproject.toml

@@ -28,6 +28,7 @@ langsmith = "langsmith.cli.main:main"
 python = ">=3.8.1,<4.0"
 pydantic = ">=1,<3"
 requests = "^2"
+pytest-subtests = "^0.11.0"
 
 
 [tool.poetry.group.dev.dependencies]

python/tests/unit_tests/test_utils.py

@@ -0,0 +1,73 @@
+import unittest
+
+import pytest
+
+import langsmith.utils as ls_utils
+
+
+class LangSmithProjectNameTest(unittest.TestCase):
+    class GetTracerProjectTestCase:
+        def __init__(
+            self, test_name, envvars, expected_project_name, return_default_value=None
+        ):
+            self.test_name = test_name
+            self.envvars = envvars
+            self.expected_project_name = expected_project_name
+            self.return_default_value = return_default_value
+
+    def test_correct_get_tracer_project(self):
+        cases = [
+            self.GetTracerProjectTestCase(
+                test_name="default to 'default' when no project provided",
+                envvars={},
+                expected_project_name="default",
+            ),
+            self.GetTracerProjectTestCase(
+                test_name="default to 'default' when "
+                + "return_default_value=True and no project provided",
+                envvars={},
+                expected_project_name="default",
+            ),
+            self.GetTracerProjectTestCase(
+                test_name="do not default if return_default_value=False "
+                + "when no project provided",
+                envvars={},
+                expected_project_name=None,
+                return_default_value=False,
+            ),
+            self.GetTracerProjectTestCase(
+                test_name="use session_name for legacy tracers",
+                envvars={"LANGCHAIN_SESSION": "old_timey_session"},
+                expected_project_name="old_timey_session",
+            ),
+            self.GetTracerProjectTestCase(
+                test_name="use LANGCHAIN_PROJECT over SESSION_NAME",
+                envvars={
+                    "LANGCHAIN_SESSION": "old_timey_session",
+                    "LANGCHAIN_PROJECT": "modern_session",
+                },
+                expected_project_name="modern_session",
+            ),
+            self.GetTracerProjectTestCase(
+                test_name="hosted projects get precedence over all other defaults",
+                envvars={
+                    "HOSTED_LANGSERVE_PROJECT_NAME": "hosted_project",
+                    "LANGCHAIN_SESSION": "old_timey_session",
+                    "LANGCHAIN_PROJECT": "modern_session",
+                },
+                expected_project_name="hosted_project",
+            ),
+        ]
+
+        for case in cases:
+            with self.subTest(msg=case.test_name):
+                with pytest.MonkeyPatch.context() as mp:
+                    for k, v in case.envvars.items():
+                        mp.setenv(k, v)
+
+                    project = (
+                        ls_utils.get_tracer_project()
+                        if case.return_default_value is None
+                        else ls_utils.get_tracer_project(case.return_default_value)
+                    )
+                    self.assertEqual(project, case.expected_project_name)