change gpt-4o and install in .yml

ci/test/.test.yml

@@ -9,6 +9,7 @@
   <<: *use_base_container
   script:
     - pip3 install -e .
+    - pip install inline-snapshot pytest-asyncio graphviz pytest-mock
     - pytest -s $TEST_PATH
   tags:
     - p40

tests/core/test_openai_chatcompletions.py

@@ -31,7 +31,8 @@ from cai.sdk.agents import (
     generation_span,
 )
 from cai.sdk.agents.models.fake_id import FAKE_RESPONSES_ID
-
+import os
+cai_model = os.getenv('CAI_MODEL', "qwen2.5:14b")
 
 @pytest.mark.allow_call_model_methods
 @pytest.mark.asyncio
@@ -57,7 +58,7 @@ async def test_get_response_with_text_message(monkeypatch) -> None:
         return chat
 
     monkeypatch.setattr(OpenAIChatCompletionsModel, "_fetch_response", patched_fetch_response)
-    model = OpenAIProvider(use_responses=False).get_model("gpt-4")
+    model = OpenAIProvider(use_responses=False).get_model(cai_model)
     resp: ModelResponse = await model.get_response(
         system_instructions=None,
         input="",
@@ -105,7 +106,7 @@ async def test_get_response_with_refusal(monkeypatch) -> None:
         return chat
 
     monkeypatch.setattr(OpenAIChatCompletionsModel, "_fetch_response", patched_fetch_response)
-    model = OpenAIProvider(use_responses=False).get_model("gpt-4")
+    model = OpenAIProvider(use_responses=False).get_model(cai_model)
     resp: ModelResponse = await model.get_response(
         system_instructions=None,
         input="",
@@ -154,7 +155,7 @@ async def test_get_response_with_tool_call(monkeypatch) -> None:
         return chat
 
     monkeypatch.setattr(OpenAIChatCompletionsModel, "_fetch_response", patched_fetch_response)
-    model = OpenAIProvider(use_responses=False).get_model("gpt-4")
+    model = OpenAIProvider(use_responses=False).get_model(cai_model)
     resp: ModelResponse = await model.get_response(
         system_instructions=None,
         input="",
@@ -208,7 +209,7 @@ async def test_fetch_response_non_stream(monkeypatch) -> None:
     )
     completions = DummyCompletions()
     dummy_client = DummyClient(completions)
-    model = OpenAIChatCompletionsModel(model="gpt-4", openai_client=dummy_client)  # type: ignore
+    model = OpenAIChatCompletionsModel(model=cai_model, openai_client=dummy_client)  # type: ignore
     # Execute the private fetch with a system instruction and simple string input.
     with generation_span(disabled=True) as span:
         result = await model._fetch_response(
@@ -227,7 +228,7 @@ async def test_fetch_response_non_stream(monkeypatch) -> None:
     kwargs = completions.kwargs
     assert kwargs["stream"] is False
     assert kwargs["store"] is True
-    assert kwargs["model"] == "gpt-4"
+    assert kwargs["model"] == cai_model
     assert kwargs["messages"][0]["role"] == "system"
     assert kwargs["messages"][0]["content"] == "sys"
     assert kwargs["messages"][1]["role"] == "user"
@@ -265,7 +266,7 @@ async def test_fetch_response_stream(monkeypatch) -> None:
 
     completions = DummyCompletions()
     dummy_client = DummyClient(completions)
-    model = OpenAIChatCompletionsModel(model="gpt-4", openai_client=dummy_client)  # type: ignore
+    model = OpenAIChatCompletionsModel(model=cai_model, openai_client=dummy_client)  # type: ignore
     with generation_span(disabled=True) as span:
         response, stream = await model._fetch_response(
             system_instructions=None,
@@ -285,7 +286,7 @@ async def test_fetch_response_stream(monkeypatch) -> None:
     # Response is a proper openai Response
     assert isinstance(response, Response)
     assert response.id == FAKE_RESPONSES_ID
-    assert response.model == "gpt-4"
+    assert response.model == cai_model
     assert response.object == "response"
     assert response.output == []
     # We returned the async iterator produced by our dummy.

tests/core/test_openai_chatcompletions_stream.py

@@ -22,6 +22,8 @@ from cai.sdk.agents.models.interface import ModelTracing
 from cai.sdk.agents.models.openai_chatcompletions import OpenAIChatCompletionsModel
 from cai.sdk.agents.models.openai_provider import OpenAIProvider
 
+import os
+cai_model = os.getenv('CAI_MODEL', "qwen2.5:14b")
 
 @pytest.mark.allow_call_model_methods
 @pytest.mark.asyncio
@@ -69,7 +71,7 @@ async def test_stream_response_yields_events_for_text_content(monkeypatch) -> No
         return resp, fake_stream()
 
     monkeypatch.setattr(OpenAIChatCompletionsModel, "_fetch_response", patched_fetch_response)
-    model = OpenAIProvider(use_responses=False).get_model("gpt-4")
+    model = OpenAIProvider(use_responses=False).get_model(cai_model)
     output_events = []
     async for event in model.stream_response(
         system_instructions=None,
@@ -158,7 +160,7 @@ async def test_stream_response_yields_events_for_refusal_content(monkeypatch) ->
         return resp, fake_stream()
 
     monkeypatch.setattr(OpenAIChatCompletionsModel, "_fetch_response", patched_fetch_response)
-    model = OpenAIProvider(use_responses=False).get_model("gpt-4")
+    model = OpenAIProvider(use_responses=False).get_model(cai_model)
     output_events = []
     async for event in model.stream_response(
         system_instructions=None,
@@ -245,7 +247,7 @@ async def test_stream_response_yields_events_for_tool_call(monkeypatch) -> None:
         return resp, fake_stream()
 
     monkeypatch.setattr(OpenAIChatCompletionsModel, "_fetch_response", patched_fetch_response)
-    model = OpenAIProvider(use_responses=False).get_model("gpt-4")
+    model = OpenAIProvider(use_responses=False).get_model(cai_model)
     output_events = []
     async for event in model.stream_response(
         system_instructions=None,

tests/others/test_config.py

@@ -9,22 +9,25 @@ from cai.sdk.agents.models.openai_provider import OpenAIProvider
 from cai.sdk.agents.models.openai_responses import OpenAIResponsesModel
 
 
+import os
+cai_model = os.getenv('CAI_MODEL', "qwen2.5:14b")
+
 def test_cc_no_default_key_errors(monkeypatch):
     monkeypatch.delenv("OPENAI_API_KEY", raising=False)
     with pytest.raises(openai.OpenAIError):
-        OpenAIProvider(use_responses=False).get_model("gpt-4")
+        OpenAIProvider(use_responses=False).get_model(cai_model)
 
 
 def test_cc_set_default_openai_key():
     set_default_openai_key("test_key")
-    chat_model = OpenAIProvider(use_responses=False).get_model("gpt-4")
+    chat_model = OpenAIProvider(use_responses=False).get_model(cai_model)
     assert chat_model._client.api_key == "test_key"  # type: ignore
 
 
 def test_cc_set_default_openai_client():
     client = openai.AsyncOpenAI(api_key="test_key")
     set_default_openai_client(client)
-    chat_model = OpenAIProvider(use_responses=False).get_model("gpt-4")
+    chat_model = OpenAIProvider(use_responses=False).get_model(cai_model)
     assert chat_model._client.api_key == "test_key"  # type: ignore
 
 
@@ -32,33 +35,33 @@ def test_resp_no_default_key_errors(monkeypatch):
     monkeypatch.delenv("OPENAI_API_KEY", raising=False)
     assert os.getenv("OPENAI_API_KEY") is None
     with pytest.raises(openai.OpenAIError):
-        OpenAIProvider(use_responses=True).get_model("gpt-4")
+        OpenAIProvider(use_responses=True).get_model(cai_model)
 
 
 def test_resp_set_default_openai_key():
     set_default_openai_key("test_key")
-    resp_model = OpenAIProvider(use_responses=True).get_model("gpt-4")
+    resp_model = OpenAIProvider(use_responses=True).get_model(cai_model)
     assert resp_model._client.api_key == "test_key"  # type: ignore
 
 
 def test_resp_set_default_openai_client():
     client = openai.AsyncOpenAI(api_key="test_key")
     set_default_openai_client(client)
-    resp_model = OpenAIProvider(use_responses=True).get_model("gpt-4")
+    resp_model = OpenAIProvider(use_responses=True).get_model(cai_model)
     assert resp_model._client.api_key == "test_key"  # type: ignore
 
 
 def test_set_default_openai_api():
-    assert isinstance(OpenAIProvider().get_model("gpt-4"), OpenAIResponsesModel), (
+    assert isinstance(OpenAIProvider().get_model(cai_model), OpenAIResponsesModel), (
         "Default should be responses"
     )
 
     set_default_openai_api("chat_completions")
-    assert isinstance(OpenAIProvider().get_model("gpt-4"), OpenAIChatCompletionsModel), (
+    assert isinstance(OpenAIProvider().get_model(cai_model), OpenAIChatCompletionsModel), (
         "Should be chat completions model"
     )
 
     set_default_openai_api("responses")
-    assert isinstance(OpenAIProvider().get_model("gpt-4"), OpenAIResponsesModel), (
+    assert isinstance(OpenAIProvider().get_model(cai_model), OpenAIResponsesModel), (
         "Should be responses model"
     )