full support of o1 models + Update LLM dependencies in pyproject.toml to latest versions

[mrT23]

User description

for o1 models we should turn off stream always

also update litellm dependencies (and added the past fix)

---

PR Type

enhancement, bug_fix, dependencies

---

Description

• Adjusted the call_model method in AICaller.py to set stream to False for "o1-preview" and "o1-mini" models, and removed max_tokens from completion_params.
• Removed unnecessary model condition in UnitTestGenerator.py that set the model to "gpt-4o" for specific models.
• Updated test assertion in test_AICaller.py to reflect changes in exception messages.
• Updated litellm, openai, and tiktoken dependencies in pyproject.toml to the latest versions.

---

PRDescriptionHeader.CHANGES_WALKTHROUGH

	Relevant files
Enhancement	AICaller.py Adjust streaming and parameters for o1 models                        cover_agent/AICaller.py Set stream to False for "o1-preview" and "o1-mini" models. Removed max_tokens from completion_params for these models. +1/-0      UnitTestGenerator.py Remove unnecessary model condition in UnitTestGenerator    cover_agent/UnitTestGenerator.py - Removed condition setting model to "gpt-4o" for specific models. +0/-2
Tests	test_AICaller.py Update test assertion for exception message                            tests/test_AICaller.py - Updated test assertion for exception message. +2/-1
Dependencies	pyproject.toml Update LLM dependencies to latest versions                              pyproject.toml Updated litellm dependency to a specific Git repository. Updated openai and tiktoken dependencies to newer versions. +3/-3

---

💡 PR-Agent usage: Comment /help "your question" on any pull request to receive relevant information

[codiumai-pr-agent-pro[bot]]

PR Reviewer Guide 🔍

Here are some key observations to aid the review process:

⏱️ Estimated effort to review: 2 🔵🔵⚪⚪⚪

🧪 PR contains tests

🔒 No security concerns identified

⚡ Recommended focus areas for review Potential Inconsistency The `stream` parameter is set to False twice for o1 models. Verify if this redundancy is necessary or if it can be simplified. Possible Oversight The removal of the model-specific condition for setting the model to "gpt-4o" might affect the functionality. Ensure this change doesn't impact the expected behavior. Test Modification The assertion for the error message has been changed and a comment suggests it might vary. Verify if this change accurately reflects the expected behavior across different versions.

[codiumai-pr-agent-pro[bot]]

PR Code Suggestions ✨

Explore these optional code suggestions:

Category	Suggestion	Score
Best practice	Use a more flexible assertion for error messages in tests ___ **Consider using a more generic assertion for the error message to make the test more robust against minor changes in error reporting.** [tests/test_AICaller.py [54]](https://github.com/Codium-ai/cover-agent/pull/198/files#diff-e84aeed71db4a60462435bc58814bc055ba7779343c529c6531763006007a961R54-R54) ```diff -assert str(exc_info.value) == "'NoneType' object is not subscriptable" # this error message might change for different versions of litellm +assert "NoneType" in str(exc_info.value) and "subscriptable" in str(exc_info.value) ``` - [ ] **Apply this suggestion** Suggestion importance[1-10]: 7 Why: The suggestion to use a more generic assertion makes the test more robust against changes in error messages, enhancing the maintainability and reliability of the test suite.	7
Enhancement	Consolidate redundant comments to improve code readability ___ **Consider consolidating the two statements about streaming for o1 models into a single comment to avoid redundancy.** [cover_agent/AICaller.py [61-66]](https://github.com/Codium-ai/cover-agent/pull/198/files#diff-d94557e610fe5a4d747c06e8ac8d6a0e243e5c50b4df8f624fa1acc7d51f57b0R61-R66) ```diff if self.model in ["o1-preview", "o1-mini"]: - stream = False # o1 doesn't support streaming + # o1 models don't support streaming + stream = False completion_params["temperature"] = 1 - completion_params["stream"] = False # o1 doesn't support streaming + completion_params["stream"] = False completion_params["max_completion_tokens"] = max_tokens completion_params.pop("max_tokens", None) # Remove 'max_tokens' if present ``` - [ ] **Apply this suggestion** Suggestion importance[1-10]: 5 Why: The suggestion to consolidate comments improves code readability by reducing redundancy. However, it does not address any functional issues or bugs, so its impact is moderate.	5

💡 Need additional feedback ? start a PR chat

[EmbeddedDevops1]

⌛ Running regression testing: https://github.com/Codium-ai/cover-agent/actions/runs/11608162470