feat: add support for OpenAI 1.78 distribution

- Update litellm dependency to allow >=1.68.0,<1.72.0 for OpenAI 1.78 compatibility
- Add test to verify OpenAI 1.78 compatibility with multi-image input support
- Resolves #2910

Co-Authored-By: Joe Moura <joao@crewai.com>

tests/test_openai_178_compatibility.py

@@ -0,0 +1,53 @@
+import os
+
+import pytest
+
+from crewai import LLM, Agent, Crew, Task, TaskOutput
+
+
+@pytest.mark.skip(reason="Only run manually with valid API keys")
+def test_openai_178_compatibility_with_multimodal():
+    """
+    Test that CrewAI works with OpenAI 1.78.0 and multi-image input support.
+    This test verifies the fix for issue #2910.
+    """
+    OPENAI_API_KEY = os.getenv("OPENAI_API_KEY")
+    if not OPENAI_API_KEY:
+        pytest.skip("OPENAI_API_KEY environment variable not set")
+
+    llm = LLM(
+        model="openai/gpt-4o",  # model with vision capabilities
+        api_key=OPENAI_API_KEY,
+        temperature=0.7
+    )
+
+    visual_agent = Agent(
+        role="Multi-Image Analyst",
+        goal="Analyze multiple images and provide comprehensive reports",
+        backstory="Expert in visual analysis capable of processing multiple images simultaneously",
+        llm=llm,
+        verbose=True,
+        allow_delegation=False,
+        multimodal=True
+    )
+
+    analysis_task = Task(
+        description="""
+        Analyze these product images:
+        1. https://www.us.maguireshoes.com/cdn/shop/files/FW24-Edito-Lucena-Distressed-01_1920x.jpg?v=1736371244
+        2. https://example.com/sample-image.jpg
+        
+        Provide a comparative analysis focusing on design elements and quality indicators.
+        """,
+        expected_output="A comparative analysis of the provided images",
+        agent=visual_agent
+    )
+
+    crew = Crew(agents=[visual_agent], tasks=[analysis_task])
+    result = crew.kickoff()
+
+    assert result is not None
+    assert len(result.tasks_output) == 1
+    task_output = result.tasks_output[0]
+    assert isinstance(task_output, TaskOutput)
+    assert len(task_output.raw) > 0