Migrate openai framework to use any_llm (#828)

* Tested with copied over LitellmModel class

* Replace litellm with anyllm acompletion - fix type errors

* No need of provider field as it is not being used

* Fix openai integration test failing for error: Only strict function tools can be auto-parsed

* api_base is obtained from model_args

* Add strict_mode=True to wrap_tool test as well

* Fix mock_any_llm_response usage in unit tests

* Fix litellm / any_llm usage in unit test: test_tool_error_llm_mocked

* Drop debug script

* Properly handle omit and NotGiven cases for streaming

* Update comment for clarity

* Use class-based API - define once and reuse

* Bump openai-agents sdk version

* Fix openai unit test by correct reference to LLM_IMPORT_PATHS

* mock_anyllm_model instead of mock_litellm_model in openai unit tests

* Rename mock_anyllm_model -> mock_model

* add fixture to mock API keys in unit tests that use AnyLLM.create internally

Author: HareeshBahuleyan

pyproject.toml

@@ -42,7 +42,7 @@ llama_index = [
 ]
 
 openai = [
-  "openai-agents>=0.2.6, <1",
+  "openai-agents>=0.3.0, <1",
 ]
 
 smolagents = [