gllm-inference-binary 0.4.56__cp311-cp311-manylinux_2_31_x86_64.whl → 0.4.58__cp311-cp311-manylinux_2_31_x86_64.whl

gllm_inference/lm_invoker/litellm_lm_invoker.pyi

@@ -3,6 +3,7 @@ from gllm_core.event import EventEmitter as EventEmitter
 from gllm_core.utils.retry import RetryConfig as RetryConfig
 from gllm_inference.constants import ALL_EXTENSIONS as ALL_EXTENSIONS
 from gllm_inference.lm_invoker.openai_compatible_lm_invoker import OpenAICompatibleLMInvoker as OpenAICompatibleLMInvoker
+from gllm_inference.lm_invoker.openai_lm_invoker import ReasoningEffort as ReasoningEffort
 from gllm_inference.schema import AttachmentType as AttachmentType, ModelId as ModelId, ModelProvider as ModelProvider, MultimodalOutput as MultimodalOutput, ResponseSchema as ResponseSchema
 from langchain_core.tools import Tool as Tool
 from typing import Any
@@ -172,6 +173,46 @@ class LiteLLMLMInvoker(OpenAICompatibleLMInvoker):
         lm_invoker = LiteLLMLMInvoker(..., retry_config=retry_config)
         ```
 
+     Reasoning:
+        Some language models support advanced reasoning capabilities. When using such reasoning-capable models,
+        you can configure how much reasoning the model should perform before generating a final response by setting
+        reasoning-related parameters.
+
+        The reasoning effort of reasoning models can be set via the `reasoning_effort` parameter. This parameter
+        will guide the models on how many reasoning tokens it should generate before creating a response to the prompt.
+        The reasoning effort is only supported by some language models.
+        Available options include:
+        1. "low": Favors speed and economical token usage.
+        2. "medium": Favors a balance between speed and reasoning accuracy.
+        3. "high": Favors more complete reasoning at the cost of more tokens generated and slower responses.
+        This may differ between models. When not set, the reasoning effort will be equivalent to None by default.
+
+        When using reasoning models, some providers might output the reasoning summary. These will be stored in the
+        `reasoning` attribute in the output.
+
+        Output example:
+        ```python
+        LMOutput(
+            response="Golden retriever is a good dog breed.",
+            reasoning=[Reasoning(id="", reasoning="Let me think about it...")],
+        )
+        ```
+
+        When streaming is enabled along with reasoning and the provider supports reasoning output, the reasoning token
+        will be streamed with the `EventType.DATA` event type.
+
+        Streaming output example:
+        ```python
+        {"type": "data", "value": \'{"data_type": "thinking_start", "data_value": ""}\', ...}
+        {"type": "data", "value": \'{"data_type": "thinking", "data_value": "Let me think "}\', ...}
+        {"type": "data", "value": \'{"data_type": "thinking", "data_value": "about it..."}\', ...}
+        {"type": "data", "value": \'{"data_type": "thinking_end", "data_value": ""}\', ...}
+        {"type": "response", "value": "Golden retriever ", ...}
+        {"type": "response", "value": "is a good dog breed.", ...}
+
+        Setting reasoning-related parameters for non-reasoning models will raise an error.
+
+
     Output types:
         The output of the `LiteLLMLMInvoker` is of type `MultimodalOutput`, which is a type alias that can represent:
         1. `str`: The text response if no additional output is needed.
@@ -193,7 +234,7 @@ class LiteLLMLMInvoker(OpenAICompatibleLMInvoker):
                 Defaults to an empty list.
     '''
     completion: Incomplete
-    def __init__(self, model_id: str, default_hyperparameters: dict[str, Any] | None = None, tools: list[Tool] | None = None, response_schema: ResponseSchema | None = None, output_analytics: bool = False, retry_config: RetryConfig | None = None) -> None:
+    def __init__(self, model_id: str, default_hyperparameters: dict[str, Any] | None = None, tools: list[Tool] | None = None, response_schema: ResponseSchema | None = None, output_analytics: bool = False, retry_config: RetryConfig | None = None, reasoning_effort: ReasoningEffort | None = None) -> None:
         """Initializes a new instance of the LiteLLMLMInvoker class.
 
         Args:
@@ -207,4 +248,6 @@ class LiteLLMLMInvoker(OpenAICompatibleLMInvoker):
             output_analytics (bool, optional): Whether to output the invocation analytics. Defaults to False.
             retry_config (RetryConfig | None, optional): The retry configuration for the language model.
                 Defaults to None, in which case a default config with no retry and 30.0 seconds timeout is used.
+            reasoning_effort (ReasoningEffort | None, optional): The reasoning effort for reasoning models.
+                Defaults to None.
         """

gllm_inference/lm_invoker/openai_compatible_lm_invoker.pyi

@@ -3,8 +3,9 @@ from gllm_core.event import EventEmitter as EventEmitter
 from gllm_core.utils.retry import RetryConfig as RetryConfig
 from gllm_inference.constants import ALL_EXTENSIONS as ALL_EXTENSIONS
 from gllm_inference.lm_invoker.lm_invoker import BaseLMInvoker as BaseLMInvoker
-from gllm_inference.lm_invoker.schema.openai_compatible import InputType as InputType, Key as Key
-from gllm_inference.schema import Attachment as Attachment, AttachmentType as AttachmentType, LMOutput as LMOutput, ModelId as ModelId, ModelProvider as ModelProvider, MultimodalPrompt as MultimodalPrompt, PromptRole as PromptRole, ResponseSchema as ResponseSchema, TokenUsage as TokenUsage, ToolCall as ToolCall, ToolResult as ToolResult
+from gllm_inference.lm_invoker.schema.openai_compatible import InputType as InputType, Key as Key, ReasoningEffort as ReasoningEffort
+from gllm_inference.schema import Attachment as Attachment, AttachmentType as AttachmentType, EmitDataType as EmitDataType, LMOutput as LMOutput, ModelId as ModelId, ModelProvider as ModelProvider, MultimodalPrompt as MultimodalPrompt, PromptRole as PromptRole, Reasoning as Reasoning, ResponseSchema as ResponseSchema, TokenUsage as TokenUsage, ToolCall as ToolCall, ToolResult as ToolResult
+from gllm_inference.utils.utils import validate_string_enum as validate_string_enum
 from langchain_core.tools import Tool as Tool
 from typing import Any
 
@@ -176,6 +177,45 @@ class OpenAICompatibleLMInvoker(BaseLMInvoker):
         lm_invoker = OpenAICompatibleLMInvoker(..., retry_config=retry_config)
         ```
 
+    Reasoning:
+        Some language models support advanced reasoning capabilities. When using such reasoning-capable models,
+        you can configure how much reasoning the model should perform before generating a final response by setting
+        reasoning-related parameters.
+
+        The reasoning effort of reasoning models can be set via the `reasoning_effort` parameter. This parameter
+        will guide the models on how many reasoning tokens it should generate before creating a response to the prompt.
+        The reasoning effort is only supported by some language models.
+        Available options include:
+        1. "low": Favors speed and economical token usage.
+        2. "medium": Favors a balance between speed and reasoning accuracy.
+        3. "high": Favors more complete reasoning at the cost of more tokens generated and slower responses.
+        This may differ between models. When not set, the reasoning effort will be equivalent to None by default.
+
+        When using reasoning models, some providers might output the reasoning summary. These will be stored in the
+        `reasoning` attribute in the output.
+
+        Output example:
+        ```python
+        LMOutput(
+            response="Golden retriever is a good dog breed.",
+            reasoning=[Reasoning(id="", reasoning="Let me think about it...")],
+        )
+        ```
+
+        When streaming is enabled along with reasoning and the provider supports reasoning output, the reasoning token
+        will be streamed with the `EventType.DATA` event type.
+
+        Streaming output example:
+        ```python
+        {"type": "data", "value": \'{"data_type": "thinking_start", "data_value": ""}\', ...}
+        {"type": "data", "value": \'{"data_type": "thinking", "data_value": "Let me think "}\', ...}
+        {"type": "data", "value": \'{"data_type": "thinking", "data_value": "about it..."}\', ...}
+        {"type": "data", "value": \'{"data_type": "thinking_end", "data_value": ""}\', ...}
+        {"type": "response", "value": "Golden retriever ", ...}
+        {"type": "response", "value": "is a good dog breed.", ...}
+
+        Setting reasoning-related parameters for non-reasoning models will raise an error.
+
     Output types:
         The output of the `OpenAICompatibleLMInvoker` is of type `MultimodalOutput`, which is a type alias that can
         represent:
@@ -198,7 +238,7 @@ class OpenAICompatibleLMInvoker(BaseLMInvoker):
                 Defaults to an empty list.
     '''
     client: Incomplete
-    def __init__(self, model_name: str, base_url: str, api_key: str | None = None, model_kwargs: dict[str, Any] | None = None, default_hyperparameters: dict[str, Any] | None = None, tools: list[Tool] | None = None, response_schema: ResponseSchema | None = None, output_analytics: bool = False, retry_config: RetryConfig | None = None, bind_tools_params: dict[str, Any] | None = None, with_structured_output_params: dict[str, Any] | None = None) -> None:
+    def __init__(self, model_name: str, base_url: str, api_key: str | None = None, model_kwargs: dict[str, Any] | None = None, default_hyperparameters: dict[str, Any] | None = None, tools: list[Tool] | None = None, response_schema: ResponseSchema | None = None, output_analytics: bool = False, retry_config: RetryConfig | None = None, reasoning_effort: ReasoningEffort | None = None, bind_tools_params: dict[str, Any] | None = None, with_structured_output_params: dict[str, Any] | None = None) -> None:
         """Initializes a new instance of the OpenAICompatibleLMInvoker class.
 
         Args:
@@ -217,6 +257,7 @@ class OpenAICompatibleLMInvoker(BaseLMInvoker):
             output_analytics (bool, optional): Whether to output the invocation analytics. Defaults to False.
             retry_config (RetryConfig | None, optional): The retry configuration for the language model.
                 Defaults to None, in which case a default config with no retry and 30.0 seconds timeout is used.
+            reasoning_effort (str | None, optional): The reasoning effort for the language model. Defaults to None.
             bind_tools_params (dict[str, Any] | None, optional): Deprecated parameter to add tool calling capability.
                 If provided, must at least include the `tools` key that is equivalent to the `tools` parameter.
                 Retained for backward compatibility. Defaults to None.

gllm_inference/lm_invoker/schema/openai_compatible.pyi

@@ -1,9 +1,13 @@
+from enum import StrEnum
+
 class Key:
-    """Defines valid keys in OpenAI."""
+    """Defines valid keys in OpenAI compatible models."""
     ARGUMENTS: str
     CONTENT: str
+    CHOICES: str
     DATA: str
     DEFS: str
+    EFFORT: str
     FINISH_REASON: str
     FORMAT: str
     FUNCTION: str
@@ -11,6 +15,7 @@ class Key:
     IMAGE_URL: str
     INPUT_AUDIO: str
     JSON_SCHEMA: str
+    MESSAGE: str
     NAME: str
     RESPONSE_FORMAT: str
     ROLE: str
@@ -24,12 +29,24 @@ class Key:
     TYPE: str
     URL: str
     USAGE: str
+    REASONING: str
+    REASONING_CONTENT: str
+    REASONING_EFFORT: str
+    SUMMARY: str
 
 class InputType:
-    """Defines valid input types in OpenAI."""
+    """Defines valid input types in OpenAI compatible models."""
     FUNCTION: str
     IMAGE_URL: str
     INPUT_AUDIO: str
     JSON_SCHEMA: str
     TEXT: str
     TOOL: str
+    REASONING: str
+    SUMMARY_TEXT: str
+
+class ReasoningEffort(StrEnum):
+    """Defines the reasoning effort for reasoning models."""
+    HIGH = 'high'
+    MEDIUM = 'medium'
+    LOW = 'low'

{gllm_inference_binary-0.4.56.dist-info → gllm_inference_binary-0.4.58.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: gllm-inference-binary
-Version: 0.4.56
+Version: 0.4.58
 Summary: A library containing components related to model inferences in Gen AI applications.
 Author: Henry Wicaksono
 Author-email: henry.wicaksono@gdplabs.id

{gllm_inference_binary-0.4.56.dist-info → gllm_inference_binary-0.4.58.dist-info}/RECORD

@@ -41,9 +41,9 @@ gllm_inference/lm_invoker/google_generativeai_lm_invoker.pyi,sha256=YIGzX4EFuVKh
 gllm_inference/lm_invoker/google_lm_invoker.pyi,sha256=yNIocjnM8FpnrRIM5mDTnTYqEXKX6AaJfRFBA9LT4wI,17535
 gllm_inference/lm_invoker/google_vertexai_lm_invoker.pyi,sha256=BC3uWeYxxQyzoDOlDACL6zqYi16BtoO581TJYMrbqCk,4271
 gllm_inference/lm_invoker/langchain_lm_invoker.pyi,sha256=tkFHv7zxz-SfA0zrtkil_WH4iLRgj5xwkeN4vJdnTyo,14315
-gllm_inference/lm_invoker/litellm_lm_invoker.pyi,sha256=g2yVUlTo21H9ieWYREUyDjO82eWSUAuidnE8VJXVvRs,10898
+gllm_inference/lm_invoker/litellm_lm_invoker.pyi,sha256=Qhcl8d4VwhD_9AOme3k9zris38wBjBkGlQBzqPd8MAE,13395
 gllm_inference/lm_invoker/lm_invoker.pyi,sha256=aMgai33nvs3LZDRnUSIqch1enkmA-2twx5wMPSxjuUY,7970
-gllm_inference/lm_invoker/openai_compatible_lm_invoker.pyi,sha256=BpomU9onejDtZMObQk2BJZhinZLbA0aF5D5w8FXSHXM,13229
+gllm_inference/lm_invoker/openai_compatible_lm_invoker.pyi,sha256=fYYFwK25Q3uQYBH_JdDdFVyRGZO2X6z9o4bhiAmmpXo,15781
 gllm_inference/lm_invoker/openai_lm_invoker.pyi,sha256=zvJyOM2pEJOEtPtlqSL9-3fKE9GtzPm6Wqw0Upysshg,20158
 gllm_inference/lm_invoker/schema/__init__.pyi,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 gllm_inference/lm_invoker/schema/anthropic.pyi,sha256=Y7NAjB7H0Wmiwv3YGZuG83K1rOCjSDB6hoZ6LjndWss,971
@@ -52,7 +52,7 @@ gllm_inference/lm_invoker/schema/datasaur.pyi,sha256=mEfrhYbwVmb1jA7UIwPBqtosqp_
 gllm_inference/lm_invoker/schema/google.pyi,sha256=ZwEAo30lif7v1EgpwmKng6rzCPxPyUypyKBYkvLjJJE,443
 gllm_inference/lm_invoker/schema/langchain.pyi,sha256=-0JIiMFofXoHDoMtpaFUOysvrPGJBvjDFcNeomnWTSY,371
 gllm_inference/lm_invoker/schema/openai.pyi,sha256=YogOvOZqPuWkNyfcvyzaxi-Bu7UMfcoRzk4gWtkPG08,1899
-gllm_inference/lm_invoker/schema/openai_compatible.pyi,sha256=uZkQUNwOjrgErXeiDK5sAUHkfoozni1MoNb6vDEZYsw,630
+gllm_inference/lm_invoker/schema/openai_compatible.pyi,sha256=iNaiEjYe_uQnhLdkp0XMhw-D1BCZR2qQZAwgMAM49us,1022
 gllm_inference/lm_invoker/tgi_lm_invoker.pyi,sha256=-Popi8_C7m90yyV1-2IQfEzBVjMVG2TFzdsZ-GTzR10,2173
 gllm_inference/multimodal_em_invoker/__init__.pyi,sha256=mvLLTF8a4hdNUECvEQO58inzf6MHWhJ9yabuV8N1vwk,385
 gllm_inference/multimodal_em_invoker/google_vertexai_multimodal_em_invoker.pyi,sha256=kTe0ZbMdSzYBELadlJOJffRfWAc4XiJ-jEcuijNXEjw,3073
@@ -98,8 +98,8 @@ gllm_inference/utils/openai_multimodal_lm_helper.pyi,sha256=eF3MVWpQOyu_oYdHRWpR
 gllm_inference/utils/retry.pyi,sha256=PVDHBDWfWj9Frvl0yf5X6nqI6oromc35LvOs-jDk3So,76
 gllm_inference/utils/utils.pyi,sha256=px3RqfrgMTR_IvC2byKjSkNSrvE9_80nIe5UUw-d09s,6017
 gllm_inference.build/.gitignore,sha256=aEiIwOuxfzdCmLZe4oB1JsBmCUxwG8x-u-HBCV9JT8E,1
-gllm_inference.cpython-311-x86_64-linux-gnu.so,sha256=4wR8MC1mV5dMiIerRytByKF-WL4RfnUJt67c1cjAuGU,4671328
+gllm_inference.cpython-311-x86_64-linux-gnu.so,sha256=XUKN54uBJsC5mZFbvYsCZ7ER1PBvs2jIU5Yow4n0cR8,4708192
 gllm_inference.pyi,sha256=RXY-iU4LtaxXcr06kRZpT9i9hRw_C65c51U-PAHH6go,4985
-gllm_inference_binary-0.4.56.dist-info/METADATA,sha256=3YCUKrCoTN-LYD8t9XOXXYX_YTTHLjS4PnUHupxSsOQ,4917
-gllm_inference_binary-0.4.56.dist-info/WHEEL,sha256=IFe_ZNdNTT_i6vUiBlaFC_vwJqKup8CcDJ489_L8YrY,110
-gllm_inference_binary-0.4.56.dist-info/RECORD,,
+gllm_inference_binary-0.4.58.dist-info/METADATA,sha256=EFxwUwdHZl9BM3T-8sMQInbMyY0IXwSEjoBJ1E_7PcE,4917
+gllm_inference_binary-0.4.58.dist-info/WHEEL,sha256=IFe_ZNdNTT_i6vUiBlaFC_vwJqKup8CcDJ489_L8YrY,110
+gllm_inference_binary-0.4.58.dist-info/RECORD,,