Promote staging to production (6caf31b)

.github/workflows/test.yml

@@ -0,0 +1,83 @@
+# Separate from the generated ci.yml. stlc keeps workflow files across builds, but
+# `stlc build --rewrite-scaffold` would drop this one; re-add it after such a run.
+name: Tests
+on:
+  # No job-level `if:` (unlike ci.yml): a job skipped on pull_request counts as passing
+  # for required checks, and the PRs to gate (promote, release-please) are same-repo.
+  pull_request:
+  push:
+    branches: [main]
+  workflow_dispatch:
+
+concurrency:
+  group: test-${{ github.ref }}
+  # Only PR runs: a cancelled run on main would fail the required check for a good commit.
+  cancel-in-progress: ${{ github.event_name == 'pull_request' }}
+
+permissions:
+  contents: read
+
+jobs:
+  test:
+    timeout-minutes: 10
+    name: full suite (py3.9)
+    runs-on: ${{ startsWith(github.repository, 'stainless-sdks/') && 'depot-ubuntu-24.04' || 'ubuntu-latest' }}
+    steps:
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+        with:
+          persist-credentials: false
+
+      - name: Install Rye
+        run: |
+          curl -sSf https://rye.astral.sh/get | bash
+          echo "$HOME/.rye/shims" >> $GITHUB_PATH
+        env:
+          RYE_VERSION: '0.44.0'
+          RYE_INSTALL_OPTION: '--yes'
+
+      - name: Bootstrap
+        run: ./scripts/bootstrap
+
+      - name: Run tests
+        run: ./scripts/test
+
+  test-integrations:
+    timeout-minutes: 10
+    name: integrations with optional SDKs (py3.12)
+    runs-on: ${{ startsWith(github.repository, 'stainless-sdks/') && 'depot-ubuntu-24.04' || 'ubuntu-latest' }}
+    steps:
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+        with:
+          persist-credentials: false
+
+      - uses: astral-sh/setup-uv@c18668ad3cf93ea998bef934396af7bb5c839dc7 # v10.2.0
+
+      # The `test` job skips integration tests whose optional package isn't
+      # installed; this one installs them. Python 3.12 because claude-agent-sdk
+      # needs 3.10+, and requirements-dev.lock only resolves on 3.9, so only the
+      # test tools are pinned from it.
+      - name: Install dependencies
+        run: |
+          uv venv --python 3.12
+          grep -E '^(pytest|pytest-asyncio|pytest-xdist|respx|dirty-equals|time-machine)==' requirements-dev.lock > test-tools.txt
+          # Resolve as of a day ago: a release still uploading its wheels forces a
+          # source build that can fail (SQLAlchemy 2.1.0 did).
+          # google-adk<2: 2.x wraps callback exceptions and fails test_user_exception_propagates_without_chaining.
+          uv pip install --exclude-newer "$(date -u -d '1 day ago' +%Y-%m-%dT%H:%M:%SZ)" \
+            -e '.[aiohttp]' -r test-tools.txt \
+            openai boto3 claude-agent-sdk 'google-adk<2' google-genai langchain-core wrapt \
+            azure-cognitiveservices-speech
+
+      # Generated API tests need the mock server and already run in `test`.
+      - name: Run integration tests
+        run: .venv/bin/pytest --ignore=tests/api_resources
+
+  # The one check to require in rulesets, so jobs can be added or renamed freely.
+  # The integrations job stays advisory until its optional dependencies are pinned.
+  # always(), not !cancelled(): a skipped required check would count as passing.
+  required:
+    if: always()
+    needs: [test]
+    runs-on: ubuntu-latest
+    steps:
+      - run: test "${{ needs.test.result }}" = success

tests/integrations/test_claude_agent_sdk.py

@@ -6,7 +6,8 @@
 # ruff: noqa: ARG001
 
 import asyncio
-from typing import Any, List
+import inspect
+from typing import Any, List, Generator
 from unittest.mock import patch
 
 import pytest
@@ -33,6 +34,27 @@ def _disable_publish(monkeypatch: pytest.MonkeyPatch) -> None:
     monkeypatch.setenv("OPENLAYER_DISABLE_PUBLISH", "true")
     monkeypatch.setenv("OPENLAYER_API_KEY", "fake")
     monkeypatch.setattr(ol_tracer, "_publish", False, raising=False)
+    # Publish inline so tests that capture the upload don't race the executor.
+    monkeypatch.setitem(ol_tracer._tracer_config, "background_publish_enabled", False)
+
+
+@pytest.fixture(autouse=True)
+def _undo_sdk_patches(monkeypatch: pytest.MonkeyPatch) -> Generator[None, None, None]:
+    """trace_claude_agent_sdk() patches the SDK globally; undo it after each test."""
+    try:
+        import claude_agent_sdk as cas
+    except ImportError:
+        yield
+        return
+    monkeypatch.setattr(cas, "query", cas.query)
+    for name in ("__init__", "query", "receive_response"):
+        monkeypatch.setattr(
+            cas.ClaudeSDKClient, name, inspect.getattr_static(cas.ClaudeSDKClient, name)
+        )
+    yield
+    # Not via monkeypatch: some tests delete the flag themselves.
+    if "_openlayer_patched" in vars(cas.ClaudeSDKClient):
+        del cas.ClaudeSDKClient._openlayer_patched
 
 
 def _capture_trace_publish():
@@ -65,6 +87,7 @@ def test_traced_query_emits_root_agent_step_with_cost_and_tokens():
     The root AGENT step gets cost/tokens/session_id populated from the final
     ``ResultMessage``.
     """
+    pytest.importorskip("claude_agent_sdk")
     from openlayer.lib.integrations.claude_agent_sdk import traced_query
 
     messages = [
@@ -178,6 +201,7 @@ async def run():
 
 def test_assistant_message_emits_chat_completion_step():
     """Each AssistantMessage becomes a CHAT_COMPLETION child of the root step."""
+    pytest.importorskip("claude_agent_sdk")
     from openlayer.lib.integrations.claude_agent_sdk import traced_query
 
     messages = [
@@ -242,6 +266,7 @@ def _extract_hook_callbacks(options, event: str):
 
 def test_tool_call_creates_tool_step_with_input_and_output():
     """A tool call yields a TOOL step with input/output/latency/tool_use_id."""
+    pytest.importorskip("claude_agent_sdk")
     import claude_agent_sdk as cas
 
     from openlayer.lib.integrations.claude_agent_sdk import traced_query
@@ -311,6 +336,7 @@ async def run():
 
 def test_mcp_tool_name_is_parsed_into_metadata():
     """A tool named ``mcp__playwright__browser_click`` records the parsed metadata."""
+    pytest.importorskip("claude_agent_sdk")
     import claude_agent_sdk as cas
 
     from openlayer.lib.integrations.claude_agent_sdk import traced_query
@@ -375,6 +401,7 @@ async def run():
 
 def test_subagent_messages_nest_under_agent_tool_step():
     """A message with ``parent_tool_use_id`` nests under the spawning Agent ToolStep."""
+    pytest.importorskip("claude_agent_sdk")
     import claude_agent_sdk as cas
 
     from openlayer.lib.integrations.claude_agent_sdk import traced_query
@@ -481,6 +508,7 @@ def test_subagent_internal_tool_calls_nest_under_agent_step():
     message says ``[tool call: Grep]`` but the actual Grep TOOL step was being
     created as a child of root instead of the ``Agent: ...`` AGENT step.
     """
+    pytest.importorskip("claude_agent_sdk")
     import claude_agent_sdk as cas
 
     from openlayer.lib.integrations.claude_agent_sdk import traced_query
@@ -612,6 +640,7 @@ async def run():
 
 def test_result_message_error_subtype_marks_root_step():
     """An error ResultMessage subtype is reflected on the root step's metadata."""
+    pytest.importorskip("claude_agent_sdk")
     from openlayer.lib.integrations.claude_agent_sdk import traced_query
 
     messages = [
@@ -649,6 +678,7 @@ async def run():
 
 def test_post_tool_use_failure_marks_tool_step_as_error():
     """PostToolUseFailure fires instead of PostToolUse — the tool step is marked errored."""
+    pytest.importorskip("claude_agent_sdk")
     import claude_agent_sdk as cas
 
     from openlayer.lib.integrations.claude_agent_sdk import traced_query
@@ -708,6 +738,7 @@ async def run():
 
 def test_user_hooks_compose_with_openlayer_hooks():
     """User-provided hooks run alongside ours; neither replaces the other."""
+    pytest.importorskip("claude_agent_sdk")
     import claude_agent_sdk as cas
 
     from openlayer.lib.integrations.claude_agent_sdk import traced_query
@@ -785,6 +816,7 @@ async def run():
 
 def test_mcp_env_is_stripped_from_agent_config_metadata():
     """``env`` and ``headers`` of MCP server configs must be redacted."""
+    pytest.importorskip("claude_agent_sdk")
     from openlayer.lib.integrations.claude_agent_sdk import traced_query
 
     messages = [
@@ -833,6 +865,7 @@ async def run():
 
 def test_trace_claude_agent_sdk_patches_module_query():
     """``trace_claude_agent_sdk()`` monkey-patches ``claude_agent_sdk.query``."""
+    pytest.importorskip("claude_agent_sdk")
     import claude_agent_sdk
 
     from openlayer.lib.integrations.claude_agent_sdk import trace_claude_agent_sdk
@@ -854,6 +887,7 @@ def test_trace_claude_agent_sdk_patches_module_query():
 
 def test_trace_claude_agent_sdk_config_persists():
     """Init kwargs are persisted into the module-level config."""
+    pytest.importorskip("claude_agent_sdk")
     import claude_agent_sdk
 
     from openlayer.lib.integrations import claude_agent_sdk as integration
@@ -882,6 +916,7 @@ def test_trace_claude_agent_sdk_config_persists():
 
 def test_trace_claude_agent_sdk_patches_claude_sdk_client():
     """``trace_claude_agent_sdk()`` also patches ``ClaudeSDKClient.query`` / ``.receive_response``."""
+    pytest.importorskip("claude_agent_sdk")
     import claude_agent_sdk
 
     from openlayer.lib.integrations.claude_agent_sdk import trace_claude_agent_sdk
@@ -912,6 +947,7 @@ def test_trace_claude_agent_sdk_patches_claude_sdk_client():
 
 def test_wrapped_stream_yields_identical_messages_in_identical_order():
     """The wrapper is a pure observer — output must equal the underlying stream."""
+    pytest.importorskip("claude_agent_sdk")
     from openlayer.lib.integrations.claude_agent_sdk import traced_query
 
     original_messages = [

tests/lib/integrations/test_langchain_callback.py

@@ -52,6 +52,9 @@ def _disable_publish(monkeypatch: pytest.MonkeyPatch) -> None:
     from openlayer.lib.tracing import tracer as _tracer
 
     monkeypatch.setattr(_tracer, "_publish", False, raising=False)
+    # Process traces inline: a background upload outlives this fixture, runs
+    # after _publish is restored, and builds a real client for later tests.
+    monkeypatch.setitem(_tracer._tracer_config, "background_publish_enabled", False)
 
 
 def _tool_call(name: str, args: dict, call_id: str) -> dict:

tests/test_async_openai_embedding_integration.py

@@ -7,6 +7,8 @@
 import asyncio
 from unittest.mock import Mock, AsyncMock, MagicMock, patch
 
+import pytest
+
 
 class TestOpenAIAsyncEmbedding:
     def _fake_response(self, embeddings, prompt_tokens=4, model="text-embedding-3-small"):
@@ -117,7 +119,7 @@ def test_handle_embedding_async_labels_azure_clients_correctly(self) -> None:
         assert kwargs["provider"] == "Azure"
 
     def test_trace_async_openai_patches_embeddings_create(self) -> None:
-        import openai  # pyright: ignore[reportMissingImports]
+        openai = pytest.importorskip("openai")
 
         from openlayer.lib.integrations.async_openai_tracer import trace_async_openai
 

tests/test_bedrock_integration.py

@@ -8,6 +8,10 @@
 import json
 from unittest.mock import MagicMock, patch
 
+import pytest
+
+pytest.importorskip("boto3")  # bedrock_tracer needs boto3 (and botocore, its dependency)
+
 
 class TestBedrockChatRegression:
     """Lock in existing chat-completion behaviour before refactoring."""

tests/test_litellm_integration.py

@@ -30,7 +30,7 @@ def test_trace_litellm_raises_import_error_without_dependency(self):
             assert "pip install litellm" in str(exc_info.value)  # type: ignore
 
     @patch('openlayer.lib.integrations.litellm_tracer.HAVE_LITELLM', True)
-    @patch('openlayer.lib.integrations.litellm_tracer.litellm')
+    @patch('openlayer.lib.integrations.litellm_tracer.litellm', create=True)
     def test_trace_litellm_patches_completion(self, mock_litellm: Mock) -> None:
         """Test that trace_litellm successfully patches litellm.completion."""
         from openlayer.lib.integrations.litellm_tracer import trace_litellm
@@ -261,7 +261,7 @@ def test_extract_litellm_metadata(self):
         assert metadata == expected_metadata
 
     @patch('openlayer.lib.integrations.litellm_tracer.HAVE_LITELLM', True)
-    @patch('openlayer.lib.integrations.litellm_tracer.litellm')
+    @patch('openlayer.lib.integrations.litellm_tracer.litellm', create=True)
     def test_detect_provider_with_litellm_method(self, mock_litellm: Mock) -> None:
         """Test provider detection using LiteLLM's get_llm_provider method."""
         from openlayer.lib.integrations.litellm_tracer import detect_provider_from_response
@@ -281,7 +281,7 @@ class TestLiteLLMEmbedding:
     """Embedding calls must be traced via add_embedding_step_to_trace."""
 
     @patch("openlayer.lib.integrations.litellm_tracer.HAVE_LITELLM", True)
-    @patch("openlayer.lib.integrations.litellm_tracer.litellm")
+    @patch("openlayer.lib.integrations.litellm_tracer.litellm", create=True)
     def test_trace_litellm_patches_embedding(self, mock_litellm: Mock) -> None:
         from openlayer.lib.integrations import litellm_tracer
 
@@ -428,7 +428,7 @@ def test_handle_embedding_failure_does_not_break_client(self) -> None:
         assert result is fake_response
 
     @patch("openlayer.lib.integrations.litellm_tracer.HAVE_LITELLM", True)
-    @patch("openlayer.lib.integrations.litellm_tracer.litellm")
+    @patch("openlayer.lib.integrations.litellm_tracer.litellm", create=True)
     def test_completion_path_unchanged_after_embedding_patch(
         self, mock_litellm: Mock
     ) -> None:

tests/test_openai_embedding_integration.py

@@ -6,6 +6,8 @@
 
 from unittest.mock import Mock, MagicMock, patch
 
+import pytest
+
 
 class TestOpenAISyncEmbedding:
     """Sync OpenAI client.embeddings.create must be traced."""
@@ -112,7 +114,7 @@ def test_handle_embedding_labels_azure_clients_correctly(self) -> None:
 
     def test_trace_openai_patches_embeddings_create(self) -> None:
         """After trace_openai, client.embeddings.create is replaced."""
-        import openai  # pyright: ignore[reportMissingImports]
+        openai = pytest.importorskip("openai")
 
         from openlayer.lib.integrations.openai_tracer import trace_openai