llama-index-llms-openai 0.2.2__py3-none-any.whl → 0.2.4__py3-none-any.whl

llama_index/llms/openai/base.py

@@ -2,6 +2,7 @@ import functools
 from typing import (
     TYPE_CHECKING,
     Any,
+    Generator,
     Awaitable,
     Callable,
     Dict,
@@ -49,8 +50,9 @@ from llama_index.core.llms.callbacks import (
 )
 from llama_index.core.llms.function_calling import FunctionCallingLLM
 from llama_index.core.llms.llm import ToolSelection
-from llama_index.core.types import BaseOutputParser, PydanticProgramMode
+from llama_index.core.types import BaseOutputParser, PydanticProgramMode, Model
 from llama_index.llms.openai.utils import (
+    O1_MODELS,
     OpenAIToolCall,
     create_retry_decorator,
     from_openai_completion_logprobs,
@@ -62,6 +64,9 @@ from llama_index.llms.openai.utils import (
     resolve_openai_credentials,
     to_openai_message_dicts,
 )
+from llama_index.core.bridge.pydantic import (
+    BaseModel,
+)
 
 from openai import AsyncOpenAI, AzureOpenAI
 from openai import OpenAI as SyncOpenAI
@@ -72,6 +77,10 @@ from openai.types.chat.chat_completion_chunk import (
 )
 from llama_index.core.llms.utils import parse_partial_json
 
+import llama_index.core.instrumentation as instrument
+
+dispatcher = instrument.get_dispatcher(__name__)
+
 if TYPE_CHECKING:
     from llama_index.core.tools.types import BaseTool
 
@@ -323,6 +332,10 @@ class OpenAI(FunctionCallingLLM):
                 model=self._get_model_name()
             ),
             model_name=self.model,
+            # TODO: Temp for O1 beta
+            system_role=MessageRole.USER
+            if self.model in O1_MODELS
+            else MessageRole.SYSTEM,
         )
 
     @llm_chat_callback()
@@ -402,7 +415,7 @@ class OpenAI(FunctionCallingLLM):
     @llm_retry_decorator
     def _chat(self, messages: Sequence[ChatMessage], **kwargs: Any) -> ChatResponse:
         client = self._get_client()
-        message_dicts = to_openai_message_dicts(messages)
+        message_dicts = to_openai_message_dicts(messages, model=self.model)
 
         if self.reuse_client:
             response = client.chat.completions.create(
@@ -484,7 +497,7 @@ class OpenAI(FunctionCallingLLM):
         self, messages: Sequence[ChatMessage], **kwargs: Any
     ) -> ChatResponseGen:
         client = self._get_client()
-        message_dicts = to_openai_message_dicts(messages)
+        message_dicts = to_openai_message_dicts(messages, model=self.model)
 
         def gen() -> ChatResponseGen:
             content = ""
@@ -690,7 +703,7 @@ class OpenAI(FunctionCallingLLM):
         self, messages: Sequence[ChatMessage], **kwargs: Any
     ) -> ChatResponse:
         aclient = self._get_aclient()
-        message_dicts = to_openai_message_dicts(messages)
+        message_dicts = to_openai_message_dicts(messages, model=self.model)
 
         if self.reuse_client:
             response = await aclient.chat.completions.create(
@@ -723,7 +736,7 @@ class OpenAI(FunctionCallingLLM):
         self, messages: Sequence[ChatMessage], **kwargs: Any
     ) -> ChatResponseAsyncGen:
         aclient = self._get_aclient()
-        message_dicts = to_openai_message_dicts(messages)
+        message_dicts = to_openai_message_dicts(messages, model=self.model)
 
         async def gen() -> ChatResponseAsyncGen:
             content = ""
@@ -940,3 +953,55 @@ class OpenAI(FunctionCallingLLM):
             )
 
         return tool_selections
+
+    @dispatcher.span
+    def structured_predict(
+        self, *args: Any, llm_kwargs: Optional[Dict[str, Any]] = None, **kwargs: Any
+    ) -> BaseModel:
+        """Structured predict."""
+        llm_kwargs = llm_kwargs or {}
+        llm_kwargs["tool_choice"] = (
+            "required" if "tool_choice" not in llm_kwargs else llm_kwargs["tool_choice"]
+        )
+        # by default structured prediction uses function calling to extract structured outputs
+        # here we force tool_choice to be required
+        return super().structured_predict(*args, llm_kwargs=llm_kwargs, **kwargs)
+
+    @dispatcher.span
+    async def astructured_predict(
+        self, *args: Any, llm_kwargs: Optional[Dict[str, Any]] = None, **kwargs: Any
+    ) -> BaseModel:
+        """Structured predict."""
+        llm_kwargs = llm_kwargs or {}
+        llm_kwargs["tool_choice"] = (
+            "required" if "tool_choice" not in llm_kwargs else llm_kwargs["tool_choice"]
+        )
+        # by default structured prediction uses function calling to extract structured outputs
+        # here we force tool_choice to be required
+        return await super().astructured_predict(*args, llm_kwargs=llm_kwargs, **kwargs)
+
+    @dispatcher.span
+    def stream_structured_predict(
+        self, *args: Any, llm_kwargs: Optional[Dict[str, Any]] = None, **kwargs: Any
+    ) -> Generator[Union[Model, List[Model]], None, None]:
+        """Stream structured predict."""
+        llm_kwargs = llm_kwargs or {}
+        llm_kwargs["tool_choice"] = (
+            "required" if "tool_choice" not in llm_kwargs else llm_kwargs["tool_choice"]
+        )
+        # by default structured prediction uses function calling to extract structured outputs
+        # here we force tool_choice to be required
+        return super().stream_structured_predict(*args, llm_kwargs=llm_kwargs, **kwargs)
+
+    @dispatcher.span
+    def stream_structured_predict(
+        self, *args: Any, llm_kwargs: Optional[Dict[str, Any]] = None, **kwargs: Any
+    ) -> Generator[Union[Model, List[Model]], None, None]:
+        """Stream structured predict."""
+        llm_kwargs = llm_kwargs or {}
+        llm_kwargs["tool_choice"] = (
+            "required" if "tool_choice" not in llm_kwargs else llm_kwargs["tool_choice"]
+        )
+        # by default structured prediction uses function calling to extract structured outputs
+        # here we force tool_choice to be required
+        return super().stream_structured_predict(*args, llm_kwargs=llm_kwargs, **kwargs)

llama_index/llms/openai/utils.py

@@ -29,6 +29,10 @@ DEFAULT_OPENAI_API_TYPE = "open_ai"
 DEFAULT_OPENAI_API_BASE = "https://api.openai.com/v1"
 DEFAULT_OPENAI_API_VERSION = ""
 
+O1_MODELS: Dict[str, int] = {
+    "o1-preview": 128000,
+    "o1-mini": 128000,
+}
 
 GPT4_MODELS: Dict[str, int] = {
     # stable model names:
@@ -109,6 +113,7 @@ GPT3_MODELS: Dict[str, int] = {
 }
 
 ALL_AVAILABLE_MODELS = {
+    **O1_MODELS,
     **GPT4_MODELS,
     **TURBO_MODELS,
     **GPT3_5_MODELS,
@@ -117,6 +122,7 @@ ALL_AVAILABLE_MODELS = {
 }
 
 CHAT_MODELS = {
+    **O1_MODELS,
     **GPT4_MODELS,
     **TURBO_MODELS,
     **AZURE_TURBO_MODELS,
@@ -220,11 +226,15 @@ def is_chat_model(model: str) -> bool:
 def is_function_calling_model(model: str) -> bool:
     is_chat_model_ = is_chat_model(model)
     is_old = "0314" in model or "0301" in model
-    return is_chat_model_ and not is_old
+
+    # TODO: This is temporary for openai's beta
+    is_o1_beta = "o1" in model
+
+    return is_chat_model_ and not is_old and not is_o1_beta
 
 
 def to_openai_message_dict(
-    message: ChatMessage, drop_none: bool = False
+    message: ChatMessage, drop_none: bool = False, model: Optional[str] = None
 ) -> ChatCompletionMessageParam:
     """Convert generic message to OpenAI message dict."""
     message_dict = {
@@ -232,6 +242,11 @@ def to_openai_message_dict(
         "content": message.content,
     }
 
+    # TODO: O1 models do not support system prompts
+    if model is not None and model in O1_MODELS:
+        if message_dict["role"] == "system":
+            message_dict["role"] = "user"
+
     # NOTE: openai messages have additional arguments:
     # - function messages have `name`
     # - assistant messages have optional `function_call`
@@ -247,11 +262,14 @@ def to_openai_message_dict(
 
 
 def to_openai_message_dicts(
-    messages: Sequence[ChatMessage], drop_none: bool = False
+    messages: Sequence[ChatMessage],
+    drop_none: bool = False,
+    model: Optional[str] = None,
 ) -> List[ChatCompletionMessageParam]:
     """Convert generic messages to OpenAI message dicts."""
     return [
-        to_openai_message_dict(message, drop_none=drop_none) for message in messages
+        to_openai_message_dict(message, drop_none=drop_none, model=model)
+        for message in messages
     ]
 
 

{llama_index_llms_openai-0.2.2.dist-info → llama_index_llms_openai-0.2.4.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: llama-index-llms-openai
-Version: 0.2.2
+Version: 0.2.4
 Summary: llama-index llms openai integration
 License: MIT
 Author: llama-index
@@ -11,8 +11,8 @@ Classifier: Programming Language :: Python :: 3.9
 Classifier: Programming Language :: Python :: 3.10
 Classifier: Programming Language :: Python :: 3.11
 Classifier: Programming Language :: Python :: 3.12
-Requires-Dist: llama-index-agent-openai (>=0.3.0,<0.4.0)
-Requires-Dist: llama-index-core (>=0.11.0,<0.12.0)
+Requires-Dist: llama-index-agent-openai (>=0.3.1,<0.4.0)
+Requires-Dist: llama-index-core (>=0.11.7,<0.12.0)
 Requires-Dist: openai (>=1.40.0,<2.0.0)
 Description-Content-Type: text/markdown
 

llama_index_llms_openai-0.2.4.dist-info/RECORD

@@ -0,0 +1,6 @@
+llama_index/llms/openai/__init__.py,sha256=vm3cIBSGkBFlE77GyfyN0EhpJcnJZN95QMhPN53EkbE,148
+llama_index/llms/openai/base.py,sha256=5wlEeigvfvdrhlm-0EEXvwuJm2t1o6ZJ9Dl2ob8oNLk,36998
+llama_index/llms/openai/utils.py,sha256=MM8xilN70w4jFXsf-aoE4HdYWD3zptLEYjpeBZ02lEc,13571
+llama_index_llms_openai-0.2.4.dist-info/METADATA,sha256=_OiQkheD5dE-b24cKH-uehLz8hnEUwSTjXQgJYEcFBA,705
+llama_index_llms_openai-0.2.4.dist-info/WHEEL,sha256=sP946D7jFCHeNz5Iq4fL4Lu-PrWrFsgfLXbbkciIZwg,88
+llama_index_llms_openai-0.2.4.dist-info/RECORD,,

llama_index_llms_openai-0.2.2.dist-info/RECORD

@@ -1,6 +0,0 @@
-llama_index/llms/openai/__init__.py,sha256=vm3cIBSGkBFlE77GyfyN0EhpJcnJZN95QMhPN53EkbE,148
-llama_index/llms/openai/base.py,sha256=C_nxCljq-5mV8jcfE3cR7BiRE6UfJ7CxO239lCvRfcI,34064
-llama_index/llms/openai/utils.py,sha256=VuDXkLR_BGVqoZc9IJqiJlVloZwG9Z7s1nGPAhlbvWE,13079
-llama_index_llms_openai-0.2.2.dist-info/METADATA,sha256=dHi8MVPHPI2icQKVEsePMYwYCCyRbl_OXAkuq_1I2JQ,705
-llama_index_llms_openai-0.2.2.dist-info/WHEEL,sha256=sP946D7jFCHeNz5Iq4fL4Lu-PrWrFsgfLXbbkciIZwg,88
-llama_index_llms_openai-0.2.2.dist-info/RECORD,,