LLM-Bridge 1.12.7__tar.gz → 1.13.1__tar.gz

{llm_bridge-1.12.7 → llm_bridge-1.13.1/LLM_Bridge.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: LLM-Bridge
-Version: 1.12.7
+Version: 1.13.1
 Summary: A Bridge for LLMs
 Author-email: windsnow1025 <windsnow1025@gmail.com>
 License-Expression: MIT
@@ -13,10 +13,10 @@ License-File: LICENSE
 Requires-Dist: fastapi
 Requires-Dist: httpx
 Requires-Dist: tenacity
-Requires-Dist: openai==1.106.1
+Requires-Dist: openai==2.9.0
 Requires-Dist: tiktoken==0.11.0
 Requires-Dist: google-genai==1.46.0
-Requires-Dist: anthropic==0.66.0
+Requires-Dist: anthropic==0.75.0
 Requires-Dist: PyMuPDF
 Requires-Dist: docxlatex>=1.1.1
 Requires-Dist: openpyxl

{llm_bridge-1.12.7 → llm_bridge-1.13.1}/LLM_Bridge.egg-info/SOURCES.txt

@@ -43,6 +43,7 @@ llm_bridge/logic/chat_generate/model_client_factory/__init__.py
 llm_bridge/logic/chat_generate/model_client_factory/claude_client_factory.py
 llm_bridge/logic/chat_generate/model_client_factory/gemini_client_factory.py
 llm_bridge/logic/chat_generate/model_client_factory/openai_client_factory.py
+llm_bridge/logic/chat_generate/model_client_factory/schema_converter.py
 llm_bridge/logic/chat_generate/model_message_converter/__init__.py
 llm_bridge/logic/chat_generate/model_message_converter/claude_message_converter.py
 llm_bridge/logic/chat_generate/model_message_converter/gemini_message_converter.py

{llm_bridge-1.12.7 → llm_bridge-1.13.1}/LLM_Bridge.egg-info/requires.txt

@@ -1,10 +1,10 @@
 fastapi
 httpx
 tenacity
-openai==1.106.1
+openai==2.9.0
 tiktoken==0.11.0
 google-genai==1.46.0
-anthropic==0.66.0
+anthropic==0.75.0
 PyMuPDF
 docxlatex>=1.1.1
 openpyxl

{llm_bridge-1.12.7/LLM_Bridge.egg-info → llm_bridge-1.13.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: LLM-Bridge
-Version: 1.12.7
+Version: 1.13.1
 Summary: A Bridge for LLMs
 Author-email: windsnow1025 <windsnow1025@gmail.com>
 License-Expression: MIT
@@ -13,10 +13,10 @@ License-File: LICENSE
 Requires-Dist: fastapi
 Requires-Dist: httpx
 Requires-Dist: tenacity
-Requires-Dist: openai==1.106.1
+Requires-Dist: openai==2.9.0
 Requires-Dist: tiktoken==0.11.0
 Requires-Dist: google-genai==1.46.0
-Requires-Dist: anthropic==0.66.0
+Requires-Dist: anthropic==0.75.0
 Requires-Dist: PyMuPDF
 Requires-Dist: docxlatex>=1.1.1
 Requires-Dist: openpyxl

{llm_bridge-1.12.7 → llm_bridge-1.13.1}/llm_bridge/client/implementations/claude/non_stream_claude_client.py

@@ -58,27 +58,17 @@ class NonStreamClaudeClient(ClaudeClient):
         try:
             logging.info(f"messages: {self.messages}")
 
-            if self.thinking:
-                message = await self.client.beta.messages.create(
-                    model=self.model,
-                    max_tokens=self.max_tokens,
-                    temperature=self.temperature,
-                    system=self.system,
-                    messages=serialize(self.messages),
-                    betas=self.betas,
-                    tools=self.tools,
-                    thinking=self.thinking,
-                )
-            else:
-                message = await self.client.beta.messages.create(
-                    model=self.model,
-                    max_tokens=self.max_tokens,
-                    temperature=self.temperature,
-                    system=self.system,
-                    messages=serialize(self.messages),
-                    betas=self.betas,
-                    tools=self.tools,
-                )
+            message = await self.client.beta.messages.create(
+                model=self.model,
+                max_tokens=self.max_tokens,
+                temperature=self.temperature,
+                system=self.system,
+                messages=serialize(self.messages),
+                betas=self.betas,
+                tools=self.tools,
+                thinking=self.thinking,
+                output_format=self.output_format,
+            )
 
             return await process_claude_non_stream_response(
                 message=message,

llm_bridge-1.13.1/llm_bridge/client/implementations/claude/stream_claude_client.py

@@ -0,0 +1,54 @@
+import logging
+import re
+from typing import AsyncGenerator
+
+import httpx
+from fastapi import HTTPException
+
+from llm_bridge.client.implementations.claude.claude_response_handler import process_claude_stream_response
+from llm_bridge.client.model_client.claude_client import ClaudeClient
+from llm_bridge.type.chat_response import ChatResponse
+from llm_bridge.type.serializer import serialize
+
+class StreamClaudeClient(ClaudeClient):
+    async def generate_stream_response(self) -> AsyncGenerator[ChatResponse, None]:
+        try:
+            logging.info(f"messages: {self.messages}")
+
+            try:
+                async with self.client.beta.messages.stream(
+                    model=self.model,
+                    max_tokens=self.max_tokens,
+                    temperature=self.temperature,
+                    system=self.system,
+                    messages=serialize(self.messages),
+                    betas=self.betas,
+                    tools=self.tools,
+                    thinking=self.thinking,
+                    output_format=self.output_format,
+                ) as stream:
+                    async for event in stream:
+                        yield await process_claude_stream_response(
+                            event=event,
+                            input_tokens=self.input_tokens,
+                            client=self.client,
+                            model=self.model,
+                        )
+
+            except Exception as e:
+                logging.exception(e)
+                yield ChatResponse(error=repr(e))
+
+        except httpx.HTTPStatusError as e:
+            status_code = e.response.status_code
+            text = e.response.text
+            raise HTTPException(status_code=status_code, detail=text)
+        except Exception as e:
+            logging.exception(e)
+            match = re.search(r'\d{3}', str(e))
+            if match:
+                error_code = int(match.group(0))
+            else:
+                error_code = 500
+
+            raise HTTPException(status_code=error_code, detail=str(e))

{llm_bridge-1.12.7 → llm_bridge-1.13.1}/llm_bridge/client/implementations/openai/non_stream_openai_responses_client.py

@@ -86,6 +86,7 @@ class NonStreamOpenAIResponsesClient(OpenAIClient):
                 temperature=self.temperature,
                 stream=False,
                 tools=self.tools,
+                # text_format=self.structured_output_base_model, # Async OpenAPI Responses Client does not support structured output
             )
 
             return process_openai_responses_non_stream_response(

{llm_bridge-1.12.7 → llm_bridge-1.13.1}/llm_bridge/client/implementations/openai/steam_openai_responses_client.py

@@ -86,6 +86,7 @@ class StreamOpenAIResponsesClient(OpenAIClient):
                 temperature=self.temperature,
                 stream=True,
                 tools=self.tools,
+                # text_format=self.structured_output_base_model, # Async OpenAPI Responses Client does not support structured output
             )
 
         except httpx.HTTPStatusError as e:

{llm_bridge-1.12.7 → llm_bridge-1.13.1}/llm_bridge/client/model_client/claude_client.py

@@ -22,6 +22,7 @@ class ClaudeClient(ChatClient):
             input_tokens: int,
             tools: list[BetaToolUnionParam],
             thinking: ThinkingConfigEnabledParam,
+            output_format: dict[str, Any] | None,
     ):
         self.model = model
         self.messages = messages
@@ -33,8 +34,10 @@ class ClaudeClient(ChatClient):
         self.input_tokens = input_tokens
         self.tools = tools
         self.thinking = thinking
+        self.output_format = output_format
 
     async def generate_non_stream_response(self) -> ChatResponse:
         raise NotImplementedError
+
     async def generate_stream_response(self) -> AsyncGenerator[ChatResponse, None]:
         raise NotImplementedError

{llm_bridge-1.12.7 → llm_bridge-1.13.1}/llm_bridge/client/model_client/openai_client.py

@@ -1,4 +1,5 @@
-from typing import AsyncGenerator, Iterable
+from typing import AsyncGenerator, Iterable, Type, Any
+from pydantic import BaseModel
 
 import openai.lib.azure
 from openai.types import Reasoning
@@ -20,6 +21,7 @@ class OpenAIClient(ChatClient):
             client: openai.AsyncOpenAI | openai.lib.azure.AsyncAzureOpenAI,
             tools: Iterable[ToolParam],
             reasoning: Reasoning,
+            structured_output_base_model: Type[BaseModel] | None = None,
     ):
         self.model = model
         self.messages = messages
@@ -28,8 +30,10 @@ class OpenAIClient(ChatClient):
         self.client = client
         self.tools = tools
         self.reasoning = reasoning
+        self.structured_output_base_model = structured_output_base_model
 
     async def generate_non_stream_response(self) -> ChatResponse:
         raise NotImplementedError
+
     async def generate_stream_response(self) -> AsyncGenerator[ChatResponse, None]:
         raise NotImplementedError

{llm_bridge-1.12.7 → llm_bridge-1.13.1}/llm_bridge/logic/chat_generate/chat_client_factory.py

@@ -1,3 +1,5 @@
+from typing import Any
+
 from fastapi import HTTPException
 
 from llm_bridge.client.chat_client import ChatClient
@@ -16,7 +18,7 @@ async def create_chat_client(
         stream: bool,
         thought: bool,
         code_execution: bool,
-        structured_output_schema: dict | None,
+        structured_output_schema: dict[str, Any] | None,
 ) -> ChatClient:
     if api_type == 'OpenAI':
         return await create_openai_client(
@@ -28,6 +30,7 @@ async def create_chat_client(
             stream=stream,
             thought=thought,
             code_execution=code_execution,
+            structured_output_schema=structured_output_schema,
         )
     elif api_type == 'OpenAI-Azure':
         return await create_openai_client(
@@ -42,6 +45,7 @@ async def create_chat_client(
             stream=stream,
             thought=thought,
             code_execution=code_execution,
+            structured_output_schema=structured_output_schema,
         )
     elif api_type == 'OpenAI-GitHub':
         return await create_openai_client(
@@ -53,6 +57,7 @@ async def create_chat_client(
             stream=stream,
             thought=thought,
             code_execution=code_execution,
+            structured_output_schema=structured_output_schema,
         )
     elif api_type == 'Grok':
         return await create_openai_client(
@@ -64,6 +69,7 @@ async def create_chat_client(
             stream=stream,
             thought=thought,
             code_execution=code_execution,
+            structured_output_schema=structured_output_schema,
         )
     elif api_type == 'Gemini-Free':
         return await create_gemini_client(
@@ -110,6 +116,7 @@ async def create_chat_client(
             stream=stream,
             thought=thought,
             code_execution=code_execution,
+            structured_output_schema=structured_output_schema,
         )
     else:
         raise HTTPException(status_code=400, detail="Invalid API type")

{llm_bridge-1.12.7 → llm_bridge-1.13.1}/llm_bridge/logic/chat_generate/model_client_factory/claude_client_factory.py

@@ -1,4 +1,7 @@
+from typing import Any
+
 import anthropic
+from anthropic import Omit, transform_schema
 from anthropic.types import ThinkingConfigEnabledParam, AnthropicBetaParam
 from anthropic.types.beta import BetaWebSearchTool20250305Param, BetaToolUnionParam, BetaCodeExecutionTool20250825Param
 
@@ -18,7 +21,10 @@ async def create_claude_client(
         stream: bool,
         thought: bool,
         code_execution: bool,
+        structured_output_schema: dict[str, Any] | None
 ):
+    omit = Omit()
+
     client = anthropic.AsyncAnthropic(
         api_key=api_key,
     )
@@ -44,7 +50,7 @@ async def create_claude_client(
         max_output,
         context_window - input_tokens,
     )
-    thinking = None
+    thinking = omit
     if thought:
         thinking = ThinkingConfigEnabledParam(
             type="enabled",
@@ -55,6 +61,7 @@ async def create_claude_client(
         "context-1m-2025-08-07",
         "output-128k-2025-02-19",
         "code-execution-2025-08-25",
+        "structured-outputs-2025-11-13"
     ]
     tools: list[BetaToolUnionParam] = []
     tools.append(
@@ -71,6 +78,13 @@ async def create_claude_client(
             )
         )
 
+    output_format = omit
+    # if structured_output_schema: # Claude output format raises: TypeError: unhashable type: 'dict'
+    #     output_format = {
+    #         "type": "json_schema",
+    #         "schema": transform_schema(structured_output_schema),
+    #     }
+
     if stream:
         return StreamClaudeClient(
             model=model,
@@ -83,6 +97,7 @@ async def create_claude_client(
             input_tokens=input_tokens,
             tools=tools,
             thinking=thinking,
+            output_format=output_format,
         )
     else:
         return NonStreamClaudeClient(
@@ -96,4 +111,5 @@ async def create_claude_client(
             input_tokens=input_tokens,
             tools=tools,
             thinking=thinking,
+            output_format=output_format,
         )

{llm_bridge-1.12.7 → llm_bridge-1.13.1}/llm_bridge/logic/chat_generate/model_client_factory/gemini_client_factory.py

@@ -1,6 +1,8 @@
+from typing import Any
+
 from google import genai
 from google.genai import types
-from google.genai.types import Modality, HttpOptions, MediaResolution
+from google.genai.types import Modality, MediaResolution
 
 from llm_bridge.client.implementations.gemini.non_stream_gemini_client import NonStreamGeminiClient
 from llm_bridge.client.implementations.gemini.stream_gemini_client import StreamGeminiClient
@@ -18,7 +20,7 @@ async def create_gemini_client(
         stream: bool,
         thought: bool,
         code_execution: bool,
-        structured_output_schema: dict | None,
+        structured_output_schema: dict[str, Any] | None,
 ):
     client = genai.Client(
         vertexai=vertexai,
@@ -59,7 +61,6 @@ async def create_gemini_client(
     config = types.GenerateContentConfig(
         system_instruction=system_instruction,
         temperature=temperature,
-        media_resolution=MediaResolution.MEDIA_RESOLUTION_HIGH,
         safety_settings=[
             types.SafetySetting(
                 category=types.HarmCategory.HARM_CATEGORY_HATE_SPEECH,
@@ -87,7 +88,10 @@ async def create_gemini_client(
         response_modalities=response_modalities,
     )
 
-    if structured_output_schema is not None:
+    if vertexai:
+        config.media_resolution=MediaResolution.MEDIA_RESOLUTION_HIGH
+
+    if structured_output_schema:
         config.response_mime_type = "application/json"
         config.response_json_schema = structured_output_schema
 

{llm_bridge-1.12.7 → llm_bridge-1.13.1}/llm_bridge/logic/chat_generate/model_client_factory/openai_client_factory.py

@@ -1,4 +1,5 @@
 import re
+from typing import Any
 
 import openai
 from fastapi import HTTPException
@@ -13,6 +14,7 @@ from llm_bridge.client.implementations.openai.steam_openai_responses_client impo
 from llm_bridge.client.implementations.openai.stream_openai_client import StreamOpenAIClient
 from llm_bridge.logic.chat_generate.chat_message_converter import convert_messages_to_openai_responses, \
     convert_messages_to_openai
+from llm_bridge.logic.chat_generate.model_client_factory.schema_converter import json_schema_to_pydantic_model
 from llm_bridge.type.message import Message
 
 
@@ -25,6 +27,7 @@ async def create_openai_client(
         stream: bool,
         thought: bool,
         code_execution: bool,
+        structured_output_schema: dict[str, Any] | None,
 ):
     if api_type == "OpenAI":
         client = openai.AsyncOpenAI(
@@ -91,6 +94,11 @@ async def create_openai_client(
             )
         )
 
+    structured_output_base_model = None
+    if structured_output_schema:
+        structured_output_base_model = json_schema_to_pydantic_model(structured_output_schema)
+
+
     if use_responses_api:
         if stream:
             return StreamOpenAIResponsesClient(
@@ -101,6 +109,7 @@ async def create_openai_client(
                 client=client,
                 tools=tools,
                 reasoning=reasoning,
+                structured_output_base_model=structured_output_base_model,
             )
         else:
             return NonStreamOpenAIResponsesClient(
@@ -111,6 +120,7 @@ async def create_openai_client(
                 client=client,
                 tools=tools,
                 reasoning=reasoning,
+                structured_output_base_model=structured_output_base_model,
             )
     else:
         if stream:

llm_bridge-1.13.1/llm_bridge/logic/chat_generate/model_client_factory/schema_converter.py

@@ -0,0 +1,7 @@
+from typing import Any
+
+from pydantic import BaseModel
+
+
+def json_schema_to_pydantic_model(json_schema: dict[str, Any]) -> BaseModel:
+    pass

{llm_bridge-1.12.7 → llm_bridge-1.13.1}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "LLM-Bridge"
-version = "1.12.7"
+version = "1.13.1"
 authors = [
     {name = "windsnow1025", email = "windsnow1025@gmail.com"}
 ]
@@ -21,10 +21,10 @@ dependencies = [
     "fastapi",
     "httpx",
     "tenacity",
-    "openai==1.106.1",
+    "openai==2.9.0",
     "tiktoken==0.11.0",
     "google-genai==1.46.0",
-    "anthropic==0.66.0",
+    "anthropic==0.75.0",
     "PyMuPDF",
     "docxlatex>=1.1.1",
     "openpyxl",

llm_bridge-1.12.7/llm_bridge/client/implementations/claude/stream_claude_client.py

@@ -1,71 +0,0 @@
-import logging
-import re
-from typing import AsyncGenerator
-
-import httpx
-from fastapi import HTTPException
-
-from llm_bridge.client.implementations.claude.claude_response_handler import process_claude_stream_response
-from llm_bridge.client.model_client.claude_client import ClaudeClient
-from llm_bridge.type.chat_response import ChatResponse
-from llm_bridge.type.serializer import serialize
-
-class StreamClaudeClient(ClaudeClient):
-    async def generate_stream_response(self) -> AsyncGenerator[ChatResponse, None]:
-        try:
-            logging.info(f"messages: {self.messages}")
-
-            try:
-                if self.thinking:
-                    async with self.client.beta.messages.stream(
-                        model=self.model,
-                        max_tokens=self.max_tokens,
-                        temperature=self.temperature,
-                        system=self.system,
-                        messages=serialize(self.messages),
-                        betas=self.betas,
-                        tools=self.tools,
-                        thinking=self.thinking,
-                    ) as stream:
-                        async for event in stream:
-                            yield await process_claude_stream_response(
-                                event=event,
-                                input_tokens=self.input_tokens,
-                                client=self.client,
-                                model=self.model,
-                            )
-                else:
-                    async with self.client.beta.messages.stream(
-                            model=self.model,
-                            max_tokens=self.max_tokens,
-                            temperature=self.temperature,
-                            system=self.system,
-                            messages=serialize(self.messages),
-                            betas=self.betas,
-                            tools=self.tools,
-                    ) as stream:
-                        async for event in stream:
-                            yield await process_claude_stream_response(
-                                event=event,
-                                input_tokens=self.input_tokens,
-                                client=self.client,
-                                model=self.model,
-                            )
-
-            except Exception as e:
-                logging.exception(e)
-                yield ChatResponse(error=repr(e))
-
-        except httpx.HTTPStatusError as e:
-            status_code = e.response.status_code
-            text = e.response.text
-            raise HTTPException(status_code=status_code, detail=text)
-        except Exception as e:
-            logging.exception(e)
-            match = re.search(r'\d{3}', str(e))
-            if match:
-                error_code = int(match.group(0))
-            else:
-                error_code = 500
-
-            raise HTTPException(status_code=error_code, detail=str(e))