promptbuilder 0.4.29__tar.gz → 0.4.31__tar.gz

{promptbuilder-0.4.29/promptbuilder.egg-info → promptbuilder-0.4.31}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: promptbuilder
-Version: 0.4.29
+Version: 0.4.31
 Summary: Library for building prompts for LLMs
 Home-page: https://github.com/kapulkin/promptbuilder
 Author: Kapulkin Stanislav
@@ -18,6 +18,9 @@ Requires-Dist: google-genai>=1.4.0
 Requires-Dist: anthropic
 Requires-Dist: openai
 Requires-Dist: aioboto3
+Requires-Dist: litellm
+Requires-Dist: httpx
+Requires-Dist: aiohttp
 Dynamic: author
 Dynamic: author-email
 Dynamic: classifier

{promptbuilder-0.4.29 → promptbuilder-0.4.31}/promptbuilder/llm_client/aisuite_client.py

@@ -71,6 +71,7 @@ class AiSuiteLLMClient(BaseLLMClient):
         thinking_config: ThinkingConfig = ThinkingConfig(),
         system_message: str | None = None,
         max_tokens: int | None = None,
+        timeout: float | None = None,
         tools: list[Tool] | None = None,
         tool_config: ToolConfig = ToolConfig(),
     ) -> Response:
@@ -230,6 +231,7 @@ class AiSuiteLLMClientAsync(BaseLLMClientAsync):
         thinking_config: ThinkingConfig = ThinkingConfig(),
         system_message: str | None = None,
         max_tokens: int | None = None,
+        timeout: float | None = None,
         tools: list[Tool] | None = None,
         tool_config: ToolConfig = ToolConfig(),
     ) -> Response:

{promptbuilder-0.4.29 → promptbuilder-0.4.31}/promptbuilder/llm_client/anthropic_client.py

@@ -200,6 +200,7 @@ class AnthropicLLMClient(BaseLLMClient):
         thinking_config: ThinkingConfig | None = None,
         system_message: str | None = None,
         max_tokens: int | None = None,
+        timeout: float | None = None,
         tools: list[Tool] | None = None,
         tool_config: ToolConfig = ToolConfig(),
     ) -> Response:
@@ -215,6 +216,9 @@ class AnthropicLLMClient(BaseLLMClient):
             "max_tokens": max_tokens,
             "messages": anthropic_messages,
         }
+
+        if timeout is not None:
+            anthropic_kwargs["timeout"] = timeout
         
         if thinking_config is None:
             thinking_config = self.default_thinking_config
@@ -453,6 +457,7 @@ class AnthropicLLMClientAsync(BaseLLMClientAsync):
         thinking_config: ThinkingConfig | None = None,
         system_message: str | None = None,
         max_tokens: int | None = None,
+        timeout: float | None = None,
         tools: list[Tool] | None = None,
         tool_config: ToolConfig = ToolConfig(),
     ) -> Response:
@@ -474,6 +479,9 @@ class AnthropicLLMClientAsync(BaseLLMClientAsync):
             "max_tokens": max_tokens,
             "messages": anthropic_messages,
         }
+
+        if timeout is not None:
+            anthropic_kwargs["timeout"] = timeout
         
         if thinking_config is None:
             thinking_config = self.default_thinking_config

{promptbuilder-0.4.29 → promptbuilder-0.4.31}/promptbuilder/llm_client/base_client.py

@@ -82,6 +82,7 @@ class BaseLLMClient(ABC, utils.InheritDecoratorsMixin):
         thinking_config: ThinkingConfig | None = None,
         system_message: str | None = None,
         max_tokens: int | None = None,
+        timeout: float | None = None,
         tools: list[Tool] | None = None,
         tool_config: ToolConfig = ToolConfig(),
         autocomplete: bool = False
@@ -98,6 +99,7 @@ class BaseLLMClient(ABC, utils.InheritDecoratorsMixin):
             thinking_config=thinking_config,
             system_message=system_message,
             max_tokens=max_tokens,
+            timeout=timeout,
             tools=tools,
             tool_config=tool_config,
         )
@@ -105,25 +107,28 @@ class BaseLLMClient(ABC, utils.InheritDecoratorsMixin):
         total_count = BaseLLMClient._response_out_tokens(response)
 
         finish_reason = response.candidates[0].finish_reason.value if response.candidates and response.candidates[0].finish_reason else None
-        while autocomplete and response.candidates and finish_reason == FinishReason.MAX_TOKENS.value:
-            BaseLLMClient._append_generated_part(messages, response)
-
-            response = self._create(
-                messages=messages,
-                result_type=result_type,
-                thinking_config=thinking_config,
-                system_message=system_message,
-                max_tokens=max_tokens,
-                tools=tools,
-                tool_config=tool_config,
-            )
-            finish_reason = response.candidates[0].finish_reason.value if response.candidates and response.candidates[0].finish_reason else None
-            total_count += BaseLLMClient._response_out_tokens(response)
-            if max_tokens is not None and total_count >= max_tokens:
-                break
-        if response.candidates:
-            BaseLLMClient._append_generated_part(messages, response)
-            response.candidates[0].content = messages[-1] if len(messages) > 0 else None
+        if autocomplete:
+            while autocomplete and response.candidates and finish_reason == FinishReason.MAX_TOKENS.value:
+                BaseLLMClient._append_generated_part(messages, response)
+
+                response = self._create(
+                    messages=messages,
+                    result_type=result_type,
+                    thinking_config=thinking_config,
+                    system_message=system_message,
+                    max_tokens=max_tokens,
+                    timeout=timeout,
+                    tools=tools,
+                    tool_config=tool_config,
+                )
+                finish_reason = response.candidates[0].finish_reason.value if response.candidates and response.candidates[0].finish_reason else None
+                total_count += BaseLLMClient._response_out_tokens(response)
+                if max_tokens is not None and total_count >= max_tokens:
+                    break
+            if response.candidates and response.candidates[0].content:
+                appended_message = BaseLLMClient._append_generated_part(messages, response)
+                if appended_message is not None:
+                    response.candidates[0].content = appended_message
         return response
 
     @logfire_decorators.create
@@ -138,6 +143,7 @@ class BaseLLMClient(ABC, utils.InheritDecoratorsMixin):
         thinking_config: ThinkingConfig | None = None,
         system_message: str | None = None,
         max_tokens: int | None = None,
+        timeout: float | None = None,
         tools: list[Tool] | None = None,
         tool_config: ToolConfig = ToolConfig(),
     ) -> Response:
@@ -152,6 +158,7 @@ class BaseLLMClient(ABC, utils.InheritDecoratorsMixin):
         thinking_config: ThinkingConfig | None = None,
         system_message: str | None = None,
         max_tokens: int | None = None,
+        timeout: float | None = None,
         tools: None = None,
         tool_choice_mode: Literal["NONE"] = "NONE",
         autocomplete: bool = False,
@@ -165,6 +172,7 @@ class BaseLLMClient(ABC, utils.InheritDecoratorsMixin):
         thinking_config: ThinkingConfig | None = None,
         system_message: str | None = None,
         max_tokens: int | None = None,
+        timeout: float | None = None,
         tools: None = None,
         tool_choice_mode: Literal["NONE"] = "NONE",
         autocomplete: bool = False,
@@ -178,6 +186,7 @@ class BaseLLMClient(ABC, utils.InheritDecoratorsMixin):
         thinking_config: ThinkingConfig | None = None,
         system_message: str | None = None,
         max_tokens: int | None = None,
+        timeout: float | None = None,
         tools: None = None,
         tool_choice_mode: Literal["NONE"] = "NONE",
         autocomplete: bool = False,
@@ -191,6 +200,7 @@ class BaseLLMClient(ABC, utils.InheritDecoratorsMixin):
         thinking_config: ThinkingConfig | None = None,
         system_message: str | None = None,
         max_tokens: int | None = None,
+        timeout: float | None = None,
         tools: list[Tool],
         tool_choice_mode: Literal["ANY"],
         autocomplete: bool = False,
@@ -204,6 +214,7 @@ class BaseLLMClient(ABC, utils.InheritDecoratorsMixin):
         thinking_config: ThinkingConfig | None = None,
         system_message: str | None = None,
         max_tokens: int | None = None,
+        timeout: float | None = None,
         tools: list[Tool] | None = None,
         tool_choice_mode: Literal["ANY", "NONE"] = "NONE",
         autocomplete: bool = False,
@@ -215,6 +226,7 @@ class BaseLLMClient(ABC, utils.InheritDecoratorsMixin):
                 thinking_config=thinking_config,
                 system_message=system_message,
                 max_tokens=max_tokens,
+                timeout=timeout,
                 tools=tools,
                 tool_config=ToolConfig(function_calling_config=FunctionCallingConfig(mode=tool_choice_mode)),
             )
@@ -231,6 +243,7 @@ class BaseLLMClient(ABC, utils.InheritDecoratorsMixin):
             thinking_config=thinking_config,
             system_message=system_message,
             max_tokens=max_tokens,
+            timeout=timeout,
             tools=tools,
             tool_config=ToolConfig(function_calling_config=FunctionCallingConfig(mode=tool_choice_mode)),
             autocomplete=autocomplete,
@@ -245,7 +258,7 @@ class BaseLLMClient(ABC, utils.InheritDecoratorsMixin):
     
 
     @staticmethod
-    def _append_generated_part(messages: list[Content], response: Response):
+    def _append_generated_part(messages: list[Content], response: Response) -> Content | None:
         assert(response.candidates and response.candidates[0].content), "Response must contain at least one candidate with content."
 
         text_parts = [
@@ -262,7 +275,7 @@ class BaseLLMClient(ABC, utils.InheritDecoratorsMixin):
                 response_text = "".join(part.text for part in thought_parts)
                 is_thought = True
             else:
-                raise ValueError("No text or thought found in the response parts.")
+                return None
         
         if len(messages) > 0 and messages[-1].role == "model":
             message_to_append = messages[-1]
@@ -274,6 +287,7 @@ class BaseLLMClient(ABC, utils.InheritDecoratorsMixin):
                 message_to_append.parts.append(Part(text=response_text, thought=is_thought))
         else:
             messages.append(Content(parts=[Part(text=response_text, thought=is_thought)], role="model"))
+        return messages[-1]
 
     @staticmethod
     def _response_out_tokens(response: Response):
@@ -419,6 +433,7 @@ class BaseLLMClientAsync(ABC, utils.InheritDecoratorsMixin):
         thinking_config: ThinkingConfig | None = None,
         system_message: str | None = None,
         max_tokens: int | None = None,
+        timeout: float | None = None,
         tools: list[Tool] | None = None,
         tool_config: ToolConfig = ToolConfig(),
         autocomplete: bool = False,
@@ -435,6 +450,7 @@ class BaseLLMClientAsync(ABC, utils.InheritDecoratorsMixin):
             thinking_config=thinking_config,
             system_message=system_message,
             max_tokens=max_tokens,
+            timeout=timeout,
             tools=tools,
             tool_config=tool_config,
         )
@@ -442,25 +458,28 @@ class BaseLLMClientAsync(ABC, utils.InheritDecoratorsMixin):
         total_count = BaseLLMClient._response_out_tokens(response)
 
         finish_reason = response.candidates[0].finish_reason.value if response.candidates and response.candidates[0].finish_reason else None
-        while autocomplete and response.candidates and finish_reason == FinishReason.MAX_TOKENS.value:
-            BaseLLMClient._append_generated_part(messages, response)
-
-            response = await self._create(
-                messages=messages,
-                result_type=result_type,
-                thinking_config=thinking_config,
-                system_message=system_message,
-                max_tokens=max_tokens,
-                tools=tools,
-                tool_config=tool_config,
-            )
-            finish_reason = response.candidates[0].finish_reason.value if response.candidates and response.candidates[0].finish_reason else None
-            total_count += BaseLLMClient._response_out_tokens(response)
-            if max_tokens is not None and total_count >= max_tokens:
-                break
-        if response.candidates:
-            BaseLLMClient._append_generated_part(messages, response)
-            response.candidates[0].content = messages[-1] if len(messages) > 0 else None
+        if autocomplete:
+            while autocomplete and response.candidates and finish_reason == FinishReason.MAX_TOKENS.value:
+                BaseLLMClient._append_generated_part(messages, response)
+
+                response = await self._create(
+                    messages=messages,
+                    result_type=result_type,
+                    thinking_config=thinking_config,
+                    system_message=system_message,
+                    max_tokens=max_tokens,
+                    timeout=timeout,
+                    tools=tools,
+                    tool_config=tool_config,
+                )
+                finish_reason = response.candidates[0].finish_reason.value if response.candidates and response.candidates[0].finish_reason else None
+                total_count += BaseLLMClient._response_out_tokens(response)
+                if max_tokens is not None and total_count >= max_tokens:
+                    break
+            if response.candidates and response.candidates[0].content:
+                appended_message = BaseLLMClient._append_generated_part(messages, response)
+                if appended_message is not None:
+                    response.candidates[0].content = appended_message
         return response
 
     @logfire_decorators.create_async
@@ -475,6 +494,7 @@ class BaseLLMClientAsync(ABC, utils.InheritDecoratorsMixin):
         thinking_config: ThinkingConfig | None = None,
         system_message: str | None = None,
         max_tokens: int | None = None,
+        timeout: float | None = None,
         tools: list[Tool] | None = None,
         tool_config: ToolConfig = ToolConfig(),
     ) -> Response:
@@ -489,6 +509,7 @@ class BaseLLMClientAsync(ABC, utils.InheritDecoratorsMixin):
         thinking_config: ThinkingConfig | None = None,
         system_message: str | None = None,
         max_tokens: int | None = None,
+        timeout: float | None = None,
         tools: None = None,
         tool_choice_mode: Literal["NONE"] = "NONE",
         autocomplete: bool = False,
@@ -502,6 +523,7 @@ class BaseLLMClientAsync(ABC, utils.InheritDecoratorsMixin):
         thinking_config: ThinkingConfig | None = None,
         system_message: str | None = None,
         max_tokens: int | None = None,
+        timeout: float | None = None,
         tools: None = None,
         tool_choice_mode: Literal["NONE"] = "NONE",
         autocomplete: bool = False,
@@ -515,6 +537,7 @@ class BaseLLMClientAsync(ABC, utils.InheritDecoratorsMixin):
         thinking_config: ThinkingConfig | None = None,
         system_message: str | None = None,
         max_tokens: int | None = None,
+        timeout: float | None = None,
         tools: None = None,
         tool_choice_mode: Literal["NONE"] = "NONE",
         autocomplete: bool = False,
@@ -528,6 +551,7 @@ class BaseLLMClientAsync(ABC, utils.InheritDecoratorsMixin):
         thinking_config: ThinkingConfig | None = None,
         system_message: str | None = None,
         max_tokens: int | None = None,
+        timeout: float | None = None,
         tools: list[Tool],
         tool_choice_mode: Literal["ANY"],
         autocomplete: bool = False,
@@ -541,6 +565,7 @@ class BaseLLMClientAsync(ABC, utils.InheritDecoratorsMixin):
         thinking_config: ThinkingConfig | None = None,
         system_message: str | None = None,
         max_tokens: int | None = None,
+        timeout: float | None = None,
         tools: list[Tool] | None = None,
         tool_choice_mode: Literal["ANY", "NONE"] = "NONE",
         autocomplete: bool = False,
@@ -552,6 +577,7 @@ class BaseLLMClientAsync(ABC, utils.InheritDecoratorsMixin):
                 thinking_config=thinking_config,
                 system_message=system_message,
                 max_tokens=max_tokens,
+                timeout=timeout,
                 tools=tools,
                 tool_config=ToolConfig(function_calling_config=FunctionCallingConfig(mode=tool_choice_mode)),
             )
@@ -568,6 +594,7 @@ class BaseLLMClientAsync(ABC, utils.InheritDecoratorsMixin):
             thinking_config=thinking_config,
             system_message=system_message,
             max_tokens=max_tokens,
+            timeout=timeout,
             tools=tools,
             tool_config=ToolConfig(function_calling_config=FunctionCallingConfig(mode=tool_choice_mode)),
             autocomplete=autocomplete

{promptbuilder-0.4.29 → promptbuilder-0.4.31}/promptbuilder/llm_client/bedrock_client.py

@@ -111,6 +111,7 @@ class BedrockLLMClient(BaseLLMClient):
         thinking_config: ThinkingConfig | None = None,
         system_message: str | None = None,
         max_tokens: int | None = None,
+        timeout: float | None = None,
         tools: list[Tool] | None = None,
         tool_config: ToolConfig = ToolConfig(),
     ) -> Response:
@@ -120,7 +121,10 @@ class BedrockLLMClient(BaseLLMClient):
         
         if system_message is not None:
             bedrock_kwargs["system"] = [{"text": system_message}]
-        
+
+        if timeout is not None:
+            bedrock_kwargs["timeout"] = timeout
+
         if max_tokens is None:
             max_tokens = self.default_max_tokens
         if max_tokens is not None:
@@ -407,6 +411,7 @@ class BedrockLLMClientAsync(BaseLLMClientAsync):
         thinking_config: ThinkingConfig | None = None,
         system_message: str | None = None,
         max_tokens: int | None = None,
+        timeout: float | None = None,
         tools: list[Tool] | None = None,
         tool_config: ToolConfig = ToolConfig(),
     ) -> Response:
@@ -417,6 +422,9 @@ class BedrockLLMClientAsync(BaseLLMClientAsync):
         if system_message is not None:
             bedrock_kwargs["system"] = [{"text": system_message}]
         
+        if timeout is not None:
+            bedrock_kwargs["timeout"] = timeout
+
         if max_tokens is None:
             max_tokens = self.default_max_tokens
         if max_tokens is not None:

{promptbuilder-0.4.29 → promptbuilder-0.4.31}/promptbuilder/llm_client/google_client.py

@@ -57,7 +57,7 @@ class GoogleLLMClient(BaseLLMClient):
             raise ValueError("To create a google llm client you need to either set the environment variable GOOGLE_API_KEY or pass the api_key in string format")
         super().__init__(GoogleLLMClient.PROVIDER, model, decorator_configs=decorator_configs, default_thinking_config=default_thinking_config, default_max_tokens=default_max_tokens)
         self._api_key = api_key
-        self.client = Client(api_key=api_key)
+        self.client = Client(api_key=api_key, **kwargs)
     
     @property
     def api_key(self) -> str:
@@ -95,6 +95,7 @@ class GoogleLLMClient(BaseLLMClient):
         thinking_config: ThinkingConfig | None = None,
         system_message: str | None = None,
         max_tokens: int | None = None,
+        timeout: float | None = None,
         tools: list[Tool] | None = None,
         tool_config: ToolConfig = ToolConfig(),
     ) -> Response:
@@ -107,6 +108,9 @@ class GoogleLLMClient(BaseLLMClient):
             tools=tools,
             tool_config=tool_config,
         )
+        if timeout is not None:
+            # Google processes timeout via HttpOptions on the request/config
+            config.http_options = types.HttpOptions(timeout=int(timeout * 1_000))
         
         if thinking_config is None:
             thinking_config = self.default_thinking_config
@@ -233,8 +237,8 @@ class GoogleLLMClientAsync(BaseLLMClientAsync):
             raise ValueError("To create a google llm client you need to either set the environment variable GOOGLE_API_KEY or pass the api_key in string format")
         super().__init__(GoogleLLMClientAsync.PROVIDER, model, decorator_configs=decorator_configs, default_thinking_config=default_thinking_config, default_max_tokens=default_max_tokens)
         self._api_key = api_key
-        self.client = Client(api_key=api_key)
-    
+        self.client = Client(api_key=api_key, **kwargs)
+
     @property
     def api_key(self) -> str:
         return self._api_key
@@ -248,6 +252,7 @@ class GoogleLLMClientAsync(BaseLLMClientAsync):
         thinking_config: ThinkingConfig | None = None,
         system_message: str | None = None,
         max_tokens: int | None = None,
+        timeout: float | None = None,
         tools: list[Tool] | None = None,
         tool_config: ToolConfig = ToolConfig(),
     ) -> Response:
@@ -260,11 +265,13 @@ class GoogleLLMClientAsync(BaseLLMClientAsync):
             tools=tools,
             tool_config=tool_config,
         )
-        
+        if timeout is not None:
+            config.http_options = types.HttpOptions(timeout=int(timeout * 1_000))
+
         if thinking_config is None:
             thinking_config = self.default_thinking_config
         config.thinking_config = thinking_config
-        
+
         if result_type is None or result_type == "json":
             return await self.client.aio.models.generate_content(
                 model=self.model,

{promptbuilder-0.4.29 → promptbuilder-0.4.31}/promptbuilder/llm_client/litellm_client.py

@@ -139,6 +139,7 @@ class LiteLLMLLMClient(BaseLLMClient):
         thinking_config: ThinkingConfig | None = None,
         system_message: str | None = None,
         max_tokens: int | None = None,
+        timeout: float | None = None,
         tools: list[Tool] | None = None,
         tool_config: ToolConfig = ToolConfig(),
     ) -> Response:
@@ -167,6 +168,9 @@ class LiteLLMLLMClient(BaseLLMClient):
 
         if max_tokens is not None:
             kwargs["max_tokens"] = max_tokens
+        if timeout is not None:
+            # LiteLLM supports request_timeout in seconds
+            kwargs["request_timeout"] = timeout
 
         if tools is not None:
             lite_tools = []
@@ -351,6 +355,7 @@ class LiteLLMLLMClientAsync(BaseLLMClientAsync):
         thinking_config: ThinkingConfig | None = None,
         system_message: str | None = None,
         max_tokens: int | None = None,
+        timeout: float | None = None,
         tools: list[Tool] | None = None,
         tool_config: ToolConfig = ToolConfig(),
     ) -> Response:
@@ -377,6 +382,8 @@ class LiteLLMLLMClientAsync(BaseLLMClientAsync):
 
         if max_tokens is not None:
             kwargs["max_tokens"] = max_tokens
+        if timeout is not None:
+            kwargs["request_timeout"] = timeout
 
         if tools is not None:
             lite_tools = []

{promptbuilder-0.4.29 → promptbuilder-0.4.31}/promptbuilder/llm_client/main.py

@@ -10,7 +10,7 @@ from promptbuilder.llm_client.anthropic_client import AnthropicLLMClient, Anthro
 from promptbuilder.llm_client.openai_client import OpenaiLLMClient, OpenaiLLMClientAsync
 from promptbuilder.llm_client.bedrock_client import BedrockLLMClient, BedrockLLMClientAsync
 from promptbuilder.llm_client.aisuite_client import AiSuiteLLMClient, AiSuiteLLMClientAsync
-from promptbuilder.llm_client.litellm_client import LiteLLMLLMClient, LiteLLMLLMClientAsync
+from promptbuilder.llm_client.litellm_client import LiteLLMClient, LiteLLMClientAsync
 
 
 
@@ -24,14 +24,17 @@ def get_client(
     decorator_configs: DecoratorConfigs | None = None,
     default_thinking_config: ThinkingConfig | None = None,
     default_max_tokens: int | None = None,
+    **kwargs,
 ) -> BaseLLMClient:
     global _memory
     
-    kwargs = {
+    explicit_kwargs = {
         "decorator_configs": decorator_configs,
         "default_thinking_config": default_thinking_config,
         "default_max_tokens": default_max_tokens,
     }
+    # Merge explicit kwargs with additional kwargs, with explicit taking precedence
+    merged_kwargs = {**kwargs, **explicit_kwargs}
     provider_to_client_class: dict[str, type[BaseLLMClient]] = {
         "google": GoogleLLMClient,
         "anthropic": AnthropicLLMClient,
@@ -41,9 +44,9 @@ def get_client(
     provider, model = full_model_name.split(":", 1)
     if provider in provider_to_client_class:
         client_class = provider_to_client_class[provider]
-        client = client_class(model, api_key, **kwargs)
+        client = client_class(model, api_key, **merged_kwargs)
     else:
-        client = LiteLLMLLMClient(full_model_name, api_key, **kwargs)
+        client = LiteLLMClient(full_model_name, api_key, **merged_kwargs)
     
     if (full_model_name, client.api_key) in _memory:
         client = _memory[(full_model_name, client.api_key)]
@@ -65,14 +68,17 @@ def get_async_client(
     decorator_configs: DecoratorConfigs | None = None,
     default_thinking_config: ThinkingConfig | None = None,
     default_max_tokens: int | None = None,
+    **kwargs,
 ) -> BaseLLMClientAsync:
     global _memory_async
     
-    kwargs = {
+    explicit_kwargs = {
         "decorator_configs": decorator_configs,
         "default_thinking_config": default_thinking_config,
         "default_max_tokens": default_max_tokens,
     }
+    # Merge explicit kwargs with additional kwargs, with explicit taking precedence
+    merged_kwargs = {**kwargs, **explicit_kwargs}
     provider_to_client_class: dict[str, type[BaseLLMClientAsync]] = {
         "google": GoogleLLMClientAsync,
         "anthropic": AnthropicLLMClientAsync,
@@ -82,9 +88,9 @@ def get_async_client(
     provider, model = full_model_name.split(":", 1)
     if provider in provider_to_client_class:
         client_class = provider_to_client_class[provider]
-        client = client_class(model, api_key, **kwargs)
+        client = client_class(model, api_key, **merged_kwargs)
     else:
-        client = LiteLLMLLMClientAsync(full_model_name, api_key, **kwargs)
+        client = LiteLLMClientAsync(full_model_name, api_key, **merged_kwargs)
         
     if (full_model_name, client.api_key) in _memory_async:
         client = _memory_async[(full_model_name, client.api_key)]

{promptbuilder-0.4.29 → promptbuilder-0.4.31}/promptbuilder/llm_client/openai_client.py

@@ -150,6 +150,7 @@ class OpenaiLLMClient(BaseLLMClient):
         thinking_config: ThinkingConfig | None = None,
         system_message: str | None = None,
         max_tokens: int | None = None,
+        timeout: float | None = None,
         tools: list[Tool] | None = None,
         tool_config: ToolConfig = ToolConfig(),
     ) -> Response:
@@ -205,6 +206,9 @@ class OpenaiLLMClient(BaseLLMClient):
                 openai_kwargs["tool_choice"] = "required"
         
         if result_type is None or result_type == "json":
+            # Forward timeout to OpenAI per-request if provided
+            if timeout is not None:
+                openai_kwargs["timeout"] = timeout
             response = self.client.responses.create(**openai_kwargs)
             
             parts: list[Part] = []
@@ -227,6 +231,8 @@ class OpenaiLLMClient(BaseLLMClient):
                 ),
             )
         elif isinstance(result_type, type(BaseModel)):
+            if timeout is not None:
+                openai_kwargs["timeout"] = timeout
             response = self.client.responses.parse(**openai_kwargs, text_format=result_type)
             
             parts: list[Part] = []
@@ -385,6 +391,7 @@ class OpenaiLLMClientAsync(BaseLLMClientAsync):
         thinking_config: ThinkingConfig | None = None,
         system_message: str | None = None,
         max_tokens: int | None = None,
+        timeout: float | None = None,
         tools: list[Tool] | None = None,
         tool_config: ToolConfig = ToolConfig(),
     ) -> Response:
@@ -447,8 +454,9 @@ class OpenaiLLMClientAsync(BaseLLMClientAsync):
                 openai_kwargs["tool_choice"] = "required"
         
         if result_type is None or result_type == "json":
+            if timeout is not None:
+                openai_kwargs["timeout"] = timeout
             response = await self.client.responses.create(**openai_kwargs)
-            
             parts: list[Part] = []
             for output_item in response.output:
                 if output_item.type == "message":
@@ -469,6 +477,8 @@ class OpenaiLLMClientAsync(BaseLLMClientAsync):
                 ),
             )
         elif isinstance(result_type, type(BaseModel)):
+            if timeout is not None:
+                openai_kwargs["timeout"] = timeout
             response = await self.client.responses.parse(**openai_kwargs, text_format=result_type)
             
             parts: list[Part] = []