From 989ef138fc4fc90375640b62d689b0b46be280ea Mon Sep 17 00:00:00 2001
From: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Date: Fri, 18 Apr 2025 08:44:20 +0000
Subject: [PATCH] Update litellm dependency to v1.66.3 to fix #2640

Co-Authored-By: Joe Moura <joao@crewai.com>
---
 pyproject.toml               |  2 +-
 tests/litellm_update_test.py | 47 ++++++++++++++++++++++++++++++++++++
 2 files changed, 48 insertions(+), 1 deletion(-)
 create mode 100644 tests/litellm_update_test.py

diff --git a/pyproject.toml b/pyproject.toml
index 7528a2ecc..65efbcd68 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -11,7 +11,7 @@ dependencies = [
     # Core Dependencies
     "pydantic>=2.4.2",
     "openai>=1.13.3",
-    "litellm==1.60.2",
+    "litellm==1.66.3",
     "instructor>=1.3.3",
     # Text Processing
     "pdfplumber>=0.11.4",
diff --git a/tests/litellm_update_test.py b/tests/litellm_update_test.py
new file mode 100644
index 000000000..902f0ffc2
--- /dev/null
+++ b/tests/litellm_update_test.py
@@ -0,0 +1,47 @@
+import pytest
+from unittest.mock import MagicMock, patch
+
+from crewai.llm import LLM
+
+
+def test_llm_call_with_litellm_1_66_3():
+    """Test that the LLM class works with litellm v1.66.3+"""
+    llm = LLM(
+        model="gpt-3.5-turbo",
+        temperature=0.7,
+        max_tokens=50,
+        stop=["STOP"],
+        presence_penalty=0.1,
+        frequency_penalty=0.1,
+    )
+    messages = [{"role": "user", "content": "Say 'Hello, World!' and then say STOP"}]
+
+    with patch("litellm.completion") as mocked_completion:
+        mock_message = MagicMock()
+        mock_message.content = "Hello, World! I won't say the stop word."
+        mock_choice = MagicMock()
+        mock_choice.message = mock_message
+        mock_response = MagicMock()
+        mock_response.choices = [mock_choice]
+        mock_response.usage = {
+            "prompt_tokens": 10,
+            "completion_tokens": 10,
+            "total_tokens": 20,
+        }
+
+        mocked_completion.return_value = mock_response
+
+        response = llm.call(messages)
+
+        mocked_completion.assert_called_once()
+
+        assert "Hello, World!" in response
+        assert "STOP" not in response
+
+        _, kwargs = mocked_completion.call_args
+        assert kwargs["model"] == "gpt-3.5-turbo"
+        assert kwargs["temperature"] == 0.7
+        assert kwargs["max_tokens"] == 50
+        assert kwargs["stop"] == ["STOP"]
+        assert kwargs["presence_penalty"] == 0.1
+        assert kwargs["frequency_penalty"] == 0.1