test(fireworks): mock remaining live smoke tests

test_completion_fireworks_ai and test_completion_cost_fireworks_ai
made real Fireworks calls and broke whenever Fireworks rotated its
serverless catalog (no externally-verifiable model list exists).
They also asserted nothing — just printed.

Mock the HTTP post and assert real behavior instead: the request is
built with the right model/messages and the OpenAI-compatible
response parses back; the cost path yields a non-zero cost against
the local cost map. No network, no model dependency, stronger than
the old smoke checks.

tests/local_testing/test_completion.py

@@ -1047,22 +1047,50 @@ def test_completion_openai_params(model):
 
 
 def test_completion_fireworks_ai():
-    try:
-        litellm.set_verbose = True
-        messages = [
-            {"role": "system", "content": "You're a good bot"},
+    """
+    Mocked so it does not depend on Fireworks' rotating serverless catalog
+    (no externally-verifiable model list exists). Asserts the request is
+    built correctly and the OpenAI-compatible response is parsed back.
+    """
+    litellm.set_verbose = True
+    messages = [
+        {"role": "system", "content": "You're a good bot"},
+        {"role": "user", "content": "Hey"},
+    ]
+
+    mock_response = MagicMock()
+    mock_response.status_code = 200
+    mock_response.headers = {"content-type": "application/json"}
+    mock_response.json.return_value = {
+        "id": "chatcmpl-test",
+        "object": "chat.completion",
+        "created": 1234567890,
+        "model": "accounts/fireworks/models/deepseek-v3p1",
+        "choices": [
             {
-                "role": "user",
-                "content": "Hey",
-            },
-        ]
+                "index": 0,
+                "message": {"role": "assistant", "content": "Hello there!"},
+                "finish_reason": "stop",
+            }
+        ],
+        "usage": {"prompt_tokens": 10, "completion_tokens": 2, "total_tokens": 12},
+    }
+    mock_response.text = json.dumps(mock_response.json.return_value)
+
+    client = HTTPHandler()
+    with patch.object(client, "post", return_value=mock_response) as mock_post:
         response = completion(
             model="fireworks_ai/accounts/fireworks/models/deepseek-v3p1",
             messages=messages,
+            client=client,
         )
-        print(response)
-    except Exception as e:
-        pytest.fail(f"Error occurred: {e}")
+
+    mock_post.assert_called_once()
+    request_body = json.loads(mock_post.call_args.kwargs["data"])
+    assert "deepseek-v3p1" in request_body["model"]
+    assert request_body["messages"] == messages
+    assert response.choices[0].message.content == "Hello there!"
+    assert response.usage.total_tokens == 12
 
 
 @pytest.mark.parametrize(

tests/local_testing/test_completion_cost.py

@@ -1186,14 +1186,43 @@ def test_get_model_params_fireworks_ai(model, base_model):
     ],
 )
 def test_completion_cost_fireworks_ai(model):
+    """
+    Mocked so it does not depend on Fireworks' rotating serverless catalog.
+    Validates the Fireworks cost path: a parsed response with usage yields a
+    non-zero cost against the local cost map.
+    """
     os.environ["LITELLM_LOCAL_MODEL_COST_MAP"] = "True"
     litellm.model_cost = litellm.get_model_cost_map(url="")
 
-    messages = [{"role": "user", "content": "Hey, how's it going?"}]
-    resp = litellm.completion(model=model, messages=messages)  # works fine
+    mock_response_data = {
+        "id": "chatcmpl-test",
+        "object": "chat.completion",
+        "created": 1234567890,
+        "model": model.split("fireworks_ai/")[-1],
+        "choices": [
+            {
+                "index": 0,
+                "message": {"role": "assistant", "content": "Going great, thanks!"},
+                "finish_reason": "stop",
+            }
+        ],
+        "usage": {"prompt_tokens": 8, "completion_tokens": 5, "total_tokens": 13},
+    }
+
+    mock_response = MagicMock()
+    mock_response.status_code = 200
+    mock_response.headers = {"content-type": "application/json"}
+    mock_response.json.return_value = mock_response_data
+    mock_response.text = json.dumps(mock_response_data)
+
+    sync_handler = HTTPHandler()
+    messages = [{"role": "user", "content": "Hey, how's it going?"}]
+
+    with patch.object(HTTPHandler, "post", return_value=mock_response):
+        resp = litellm.completion(model=model, messages=messages, client=sync_handler)
 
-    print(resp)
     cost = completion_cost(completion_response=resp)
+    assert cost > 0
 
 
 def test_cost_azure_openai_prompt_caching():