llama-index-llms-openai 0.3.30__tar.gz → 0.3.31__tar.gz

{llama_index_llms_openai-0.3.30 → llama_index_llms_openai-0.3.31}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: llama-index-llms-openai
-Version: 0.3.30
+Version: 0.3.31
 Summary: llama-index llms openai integration
 License: MIT
 Author: llama-index

{llama_index_llms_openai-0.3.30 → llama_index_llms_openai-0.3.31}/llama_index/llms/openai/responses.py

@@ -247,6 +247,10 @@ class OpenAIResponses(FunctionCallingLLM):
         default=None,
         description="The effort to use for reasoning models.",
     )
+    context_window: Optional[int] = Field(
+        default=None,
+        description="The context window override for the model.",
+    )
 
     _client: SyncOpenAI = PrivateAttr()
     _aclient: AsyncOpenAI = PrivateAttr()
@@ -281,6 +285,7 @@ class OpenAIResponses(FunctionCallingLLM):
         async_http_client: Optional[httpx.AsyncClient] = None,
         openai_client: Optional[SyncOpenAI] = None,
         async_openai_client: Optional[AsyncOpenAI] = None,
+        context_window: Optional[int] = None,
         **kwargs: Any,
     ) -> None:
         additional_kwargs = additional_kwargs or {}
@@ -316,6 +321,7 @@ class OpenAIResponses(FunctionCallingLLM):
             default_headers=default_headers,
             call_metadata=call_metadata,
             strict=strict,
+            context_window=context_window,
             **kwargs,
         )
 
@@ -339,7 +345,8 @@ class OpenAIResponses(FunctionCallingLLM):
     @property
     def metadata(self) -> LLMMetadata:
         return LLMMetadata(
-            context_window=openai_modelname_to_contextsize(self._get_model_name()),
+            context_window=self.context_window
+            or openai_modelname_to_contextsize(self._get_model_name()),
             num_output=self.max_output_tokens or -1,
             is_chat_model=True,
             is_function_calling_model=is_function_calling_model(

{llama_index_llms_openai-0.3.30 → llama_index_llms_openai-0.3.31}/llama_index/llms/openai/utils.py

@@ -257,6 +257,10 @@ def is_chat_model(model: str) -> bool:
 
 
 def is_function_calling_model(model: str) -> bool:
+    # default to True for models that are not in the ALL_AVAILABLE_MODELS dict
+    if model not in ALL_AVAILABLE_MODELS:
+        return True
+
     # checking whether the model is fine-tuned or not.
     # fine-tuned model names these days look like:
     # ft:gpt-3.5-turbo:acemeco:suffix:abc123
@@ -510,7 +514,7 @@ def to_openai_message_dicts(
             message_dicts = to_openai_responses_message_dict(
                 message,
                 drop_none=drop_none,
-                model=model,
+                model="o3-mini",  # hardcode to ensure developer messages are used
             )
             if isinstance(message_dicts, list):
                 final_message_dicts.extend(message_dicts)

{llama_index_llms_openai-0.3.30 → llama_index_llms_openai-0.3.31}/pyproject.toml

@@ -28,7 +28,7 @@ exclude = ["**/BUILD"]
 license = "MIT"
 name = "llama-index-llms-openai"
 readme = "README.md"
-version = "0.3.30"
+version = "0.3.31"
 
 [tool.poetry.dependencies]
 python = ">=3.9,<4.0"
@@ -42,6 +42,7 @@ mypy = "0.991"
 pre-commit = "3.2.0"
 pylint = "2.15.10"
 pytest = "7.2.1"
+pytest-asyncio = "0.21.0"
 pytest-mock = "3.11.1"
 ruff = "0.0.292"
 tree-sitter-languages = "^1.8.0"