elastic-opentelemetry-instrumentation-openai: add openai integration tests

This adds a bunch of integration tests hitting openai. This requires
setting a variable `OPENAI_API_KEY` in github secrets. Tests themselves
are a copy of the recorded ones asserting values from the response data.

Integration tests are not run by default and can be called with `pytest
--integration-tests`

Author: xrmx

.github/workflows/ci.yml

@@ -72,3 +72,9 @@ jobs:
       working-directory: ${{ env.working_dir }}
     - run: pytest
       working-directory: ${{ env.working_dir }}
+    - if: ${{ env[matrix.python-version] == '3.12' && !env[matrix.openai-version] }}
+      # Only run on latest python and openai client version because we are calling openai
+      run: pytest --integration-tests
+      working-directory: ${{ env.working_dir }}
+      env:
+        OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}

instrumentation/elastic-opentelemetry-instrumentation-openai/README.md

@@ -71,6 +71,12 @@ pip install -r dev-requirements.txt
 pytest
 ```
 
+To run integration tests doing real requests:
+
+```
+OPENAI_API_KEY=unused pytest --integration-tests
+```
+
 ## Refreshing HTTP payloads
 
 We use [VCR.py](https://vcrpy.readthedocs.io/en/latest/) to automatically record HTTP responses from

instrumentation/elastic-opentelemetry-instrumentation-openai/tests/conftest.py

@@ -498,3 +498,26 @@ def assert_token_usage_metric(
             ),
         ],
     )
+
+
+def pytest_addoption(parser):
+    parser.addoption(
+        "--integration-tests",
+        action="store_true",
+        default=False,
+        help="run integrations tests doing real requests",
+    )
+
+
+def pytest_configure(config):
+    config.addinivalue_line("markers", "integration: mark integration tests")
+
+
+def pytest_collection_modifyitems(config, items):
+    run_integration_tests = bool(config.getoption("integration_tests"))
+    reason = "running integrations tests only" if run_integration_tests else "skipping integration tests"
+    skip_mark = pytest.mark.skip(reason=reason)
+    for item in items:
+        test_is_integration = "integration" in item.keywords
+        if run_integration_tests != test_is_integration:
+            item.add_marker(skip_mark)