LLM-Bridge 1.15.2__tar.gz → 1.15.3__tar.gz

{llm_bridge-1.15.2 → llm_bridge-1.15.3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: LLM-Bridge
-Version: 1.15.2
+Version: 1.15.3
 Summary: A Bridge for LLMs
 Author-email: windsnow1025 <windsnow1025@gmail.com>
 License-Expression: MIT
@@ -50,7 +50,7 @@ The features listed represent the maximum capabilities of each API type supporte
 |----------|--------------------------------|---------------------------------------------------------|-------------------|
 | OpenAI   | Text, Image, PDF               | Thinking, Web Search, Code Execution                    | Text, Image       |
 | Gemini   | Text, Image, Video, Audio, PDF | Thinking, Web Search, Code Execution, Structured Output | Text, Image, File |
-| Claude   | Text, Image, PDF               | Thinking, Web Search, Code Execution                    | Text              |
+| Claude   | Text, Image, PDF               | Thinking, Web Search, Code Execution                    | Text, File        |
 | Grok     | Text, Image                    |                                                         | Text              |
 
 #### Planned Features

{llm_bridge-1.15.2 → llm_bridge-1.15.3}/README.md

@@ -26,7 +26,7 @@ The features listed represent the maximum capabilities of each API type supporte
 |----------|--------------------------------|---------------------------------------------------------|-------------------|
 | OpenAI   | Text, Image, PDF               | Thinking, Web Search, Code Execution                    | Text, Image       |
 | Gemini   | Text, Image, Video, Audio, PDF | Thinking, Web Search, Code Execution, Structured Output | Text, Image, File |
-| Claude   | Text, Image, PDF               | Thinking, Web Search, Code Execution                    | Text              |
+| Claude   | Text, Image, PDF               | Thinking, Web Search, Code Execution                    | Text, File        |
 | Grok     | Text, Image                    |                                                         | Text              |
 
 #### Planned Features

{llm_bridge-1.15.2 → llm_bridge-1.15.3}/llm_bridge/client/implementations/claude/claude_response_handler.py

@@ -73,6 +73,39 @@ async def process_content_block(
     )
 
 
+async def build_chat_response_with_tokens(
+        text: str,
+        thought: str,
+        code: str,
+        code_output: str,
+        files: list[File],
+        input_tokens: int,
+        client: AsyncAnthropic,
+        model: str,
+) -> ChatResponse:
+    chat_response = ChatResponse(
+        text=text,
+        thought=thought,
+        code=code,
+        code_output=code_output,
+        files=files,
+    )
+    output_tokens = await count_claude_output_tokens(
+        client=client,
+        model=model,
+        chat_response=chat_response,
+    )
+    return ChatResponse(
+        text=text,
+        thought=thought,
+        code=code,
+        code_output=code_output,
+        files=files,
+        input_tokens=input_tokens,
+        output_tokens=output_tokens,
+    )
+
+
 async def process_claude_non_stream_response(
         message: BetaMessage,
         input_tokens: int,
@@ -93,26 +126,15 @@ async def process_claude_non_stream_response(
         code_output += content_block_chat_response.code_output
         files.extend(content_block_chat_response.files)
 
-    chat_response = ChatResponse(
+    return await build_chat_response_with_tokens(
         text=text,
         thought=thought,
         code=code,
         code_output=code_output,
         files=files,
-    )
-    output_tokens = await count_claude_output_tokens(
+        input_tokens=input_tokens,
         client=client,
         model=model,
-        chat_response=chat_response,
-    )
-    return ChatResponse(
-        text=text,
-        thought=thought,
-        code=code,
-        code_output=code_output,
-        files=files,
-        input_tokens=input_tokens,
-        output_tokens=output_tokens,
     )
 
 
@@ -152,24 +174,13 @@ async def process_claude_stream_response(
         code_output += content_block_chat_response.code_output
         files.extend(content_block_chat_response.files)
 
-    chat_response = ChatResponse(
+    return await build_chat_response_with_tokens(
         text=text,
         thought=thought,
         code=code,
         code_output=code_output,
         files=files,
-    )
-    output_tokens = await count_claude_output_tokens(
+        input_tokens=input_tokens,
         client=client,
         model=model,
-        chat_response=chat_response,
-    )
-    return ChatResponse(
-        text=text,
-        thought=thought,
-        code=code,
-        code_output=code_output,
-        files=files,
-        input_tokens=input_tokens,
-        output_tokens=output_tokens,
     )

{llm_bridge-1.15.2 → llm_bridge-1.15.3}/llm_bridge/client/implementations/gemini/gemini_response_handler.py

@@ -1,4 +1,3 @@
-import base64
 import mimetypes
 from typing import Optional
 

{llm_bridge-1.15.2 → llm_bridge-1.15.3}/llm_bridge/client/implementations/openai/non_stream_openai_responses_client.py

@@ -5,57 +5,16 @@ import httpx
 import openai
 from fastapi import HTTPException
 from openai import APIStatusError
-from openai.types.responses import Response, ResponseOutputItem, ResponseOutputMessage, \
-    ResponseOutputText, ResponseReasoningItem
+from openai.types.responses import Response
 
-from llm_bridge.client.implementations.openai.openai_token_couter import count_openai_responses_input_tokens, \
-    count_openai_output_tokens
+from llm_bridge.client.implementations.openai.openai_responses_response_handler import \
+    process_openai_responses_non_stream_response
+from llm_bridge.client.implementations.openai.openai_token_couter import count_openai_responses_input_tokens
 from llm_bridge.client.model_client.openai_client import OpenAIClient
-from llm_bridge.type.chat_response import ChatResponse, File
+from llm_bridge.type.chat_response import ChatResponse
 from llm_bridge.type.serializer import serialize
 
 
-def process_openai_responses_non_stream_response(
-        response: Response,
-        input_tokens: int,
-) -> ChatResponse:
-
-    output_list: list[ResponseOutputItem] = response.output
-
-    text: str = ""
-    thought: str = ""
-    files: list[File] = []
-
-    for output in output_list:
-        if output.type == "message":
-            output_message: ResponseOutputMessage = output
-            for content in output_message.content:
-                if content.type == "output_text":
-                    output_text: ResponseOutputText = content
-                    text += output_text.text
-        elif output.type == "reasoning":
-            reasoning_item: ResponseReasoningItem = output
-            for summary_delta in reasoning_item.summary:
-                thought += summary_delta.text
-        if output.type == "image_generation_call":
-            file = File(
-                name="generated_image.png",
-                data=output.result,
-                type="image/png",
-            )
-            files.append(file)
-
-    chat_response = ChatResponse(text=text, files=files)
-    output_tokens = count_openai_output_tokens(chat_response)
-    return ChatResponse(
-        text=text,
-        thought=thought,
-        files=files,
-        input_tokens=input_tokens,
-        output_tokens=output_tokens,
-    )
-
-
 class NonStreamOpenAIResponsesClient(OpenAIClient):
     async def generate_non_stream_response(self) -> ChatResponse:
         try:
@@ -72,10 +31,11 @@ class NonStreamOpenAIResponsesClient(OpenAIClient):
                 temperature=self.temperature,
                 stream=False,
                 tools=self.tools,
+                include=self.include,
                 # text_format=self.structured_output_base_model, # Async OpenAPI Responses Client does not support structured output
             )
 
-            return process_openai_responses_non_stream_response(
+            return await process_openai_responses_non_stream_response(
                 response=response,
                 input_tokens=input_tokens,
             )

llm_bridge-1.15.3/llm_bridge/client/implementations/openai/openai_responses_response_handler.py

@@ -0,0 +1,129 @@
+from openai.types.responses import Response, ResponseOutputItem, ResponseOutputMessage, \
+    ResponseOutputText, ResponseReasoningItem
+from openai.types.responses import ResponseStreamEvent, ResponseReasoningSummaryTextDeltaEvent, ResponseTextDeltaEvent, \
+    ResponseCodeInterpreterCallCodeDeltaEvent, ResponseImageGenCallPartialImageEvent, ResponseOutputItemDoneEvent, \
+    ResponseCodeInterpreterToolCall
+from openai.types.responses.response_code_interpreter_tool_call import Output, OutputLogs, OutputImage
+from openai.types.responses.response_output_item import ImageGenerationCall
+
+from llm_bridge.client.implementations.openai.openai_token_couter import count_openai_output_tokens
+from llm_bridge.logic.chat_generate.media_processor import get_base64_content_from_url
+from llm_bridge.type.chat_response import ChatResponse, File
+
+
+async def process_code_interpreter_outputs(interpreter_outputs: list[Output]) -> tuple[str, list[File]]:
+    code_output: str = ""
+    files: list[File] = []
+
+    for interpreter_output in interpreter_outputs:
+        if interpreter_output.type == "logs":
+            output_logs: OutputLogs = interpreter_output
+            code_output += output_logs.logs
+        if interpreter_output.type == "image":
+            output_image: OutputImage = interpreter_output
+            data, _ = await get_base64_content_from_url(output_image.url)
+            file = File(
+                name="code_interpreter_call_output.png",
+                data=data,
+                type="image/png",
+            )
+            files.append(file)
+
+    return code_output, files
+
+
+async def process_openai_responses_non_stream_response(
+        response: Response,
+        input_tokens: int,
+) -> ChatResponse:
+
+    output_list: list[ResponseOutputItem] = response.output
+
+    text: str = ""
+    thought: str = ""
+    code: str = ""
+    code_output: str = ""
+    files: list[File] = []
+
+    for output in output_list:
+        if output.type == "message":
+            output_message: ResponseOutputMessage = output
+            for content in output_message.content:
+                if content.type == "output_text":
+                    output_text: ResponseOutputText = content
+                    text += output_text.text
+        elif output.type == "reasoning":
+            reasoning_item: ResponseReasoningItem = output
+            for summary_delta in reasoning_item.summary:
+                thought += summary_delta.text
+        elif output.type == "code_interpreter_call":
+            code_interpreter_tool_call: ResponseCodeInterpreterToolCall = output
+            if interpreter_code := code_interpreter_tool_call.code:
+                code += interpreter_code
+            if interpreter_outputs := code_interpreter_tool_call.outputs:
+                interpreter_code_output, interpreter_files = await process_code_interpreter_outputs(interpreter_outputs)
+                code_output += interpreter_code_output
+                files.extend(interpreter_files)
+        elif output.type == "image_generation_call":
+            image_generation_call: ImageGenerationCall = output
+            file = File(
+                name="image_generation_call_output.png",
+                data=image_generation_call.result,
+                type="image/png",
+            )
+            files.append(file)
+
+    chat_response = ChatResponse(text=text, files=files)
+    output_tokens = count_openai_output_tokens(chat_response)
+    return ChatResponse(
+        text=text,
+        thought=thought,
+        code=code,
+        code_output=code_output,
+        files=files,
+        input_tokens=input_tokens,
+        output_tokens=output_tokens,
+    )
+
+
+async def process_openai_responses_stream_response(event: ResponseStreamEvent) -> ChatResponse:
+    text: str = ""
+    thought: str = ""
+    code: str = ""
+    code_output: str = ""
+    files: list[File] = []
+
+    if event.type == "response.output_text.delta":
+        text_delta_event: ResponseTextDeltaEvent = event
+        text = text_delta_event.delta
+    elif event.type == "response.reasoning_summary_text.delta":
+        reasoning_summary_text_delta_event: ResponseReasoningSummaryTextDeltaEvent = event
+        thought = reasoning_summary_text_delta_event.delta
+    elif event.type == "response.code_interpreter_call_code.delta":
+        code_interpreter_call_code_delta_event: ResponseCodeInterpreterCallCodeDeltaEvent = event
+        code = code_interpreter_call_code_delta_event.delta
+    elif event.type == "response.output_item.done":
+        output_item_done_event: ResponseOutputItemDoneEvent = event
+        if output_item_done_event.item.type == "code_interpreter_call":
+            code_interpreter_tool_call: ResponseCodeInterpreterToolCall = output_item_done_event.item
+            if interpreter_outputs := code_interpreter_tool_call.outputs:
+                interpreter_code_output, interpreter_files = await process_code_interpreter_outputs(interpreter_outputs)
+                code_output += interpreter_code_output
+                files.extend(interpreter_files)
+    elif event.type == "response.image_generation_call.partial_image":
+        image_gen_call_partial_image_event: ResponseImageGenCallPartialImageEvent = event
+        file = File(
+            name="generated_image.png",
+            data=image_gen_call_partial_image_event.partial_image_b64,
+            type="image/png",
+        )
+        files.append(file)
+
+    chat_response = ChatResponse(
+        text=text,
+        thought=thought,
+        code=code,
+        code_output=code_output,
+        files=files,
+    )
+    return chat_response

{llm_bridge-1.15.2 → llm_bridge-1.15.3}/llm_bridge/client/implementations/openai/openai_token_couter.py

@@ -35,6 +35,10 @@ def count_openai_responses_input_tokens(messages: list[OpenAIResponsesMessage])
 
 def count_openai_output_tokens(chat_response: ChatResponse) -> int:
     text = chat_response.text
+    if chat_response.thought:
+        text += chat_response.thought
+    if chat_response.code:
+        text += chat_response.code
     file_count = len(chat_response.files) if chat_response.files else 0
 
     return num_tokens_from_text(text) + file_count * 1000

{llm_bridge-1.15.2 → llm_bridge-1.15.3}/llm_bridge/client/implementations/openai/steam_openai_responses_client.py

@@ -6,53 +6,30 @@ import httpx
 import openai
 from fastapi import HTTPException
 from openai import APIStatusError, AsyncStream
-from openai.types.responses import ResponseStreamEvent, ResponseReasoningSummaryTextDeltaEvent, ResponseTextDeltaEvent
+from openai.types.responses import ResponseStreamEvent
 
+from llm_bridge.client.implementations.openai.openai_responses_response_handler import \
+    process_openai_responses_stream_response
 from llm_bridge.client.implementations.openai.openai_token_couter import count_openai_responses_input_tokens, \
     count_openai_output_tokens
 from llm_bridge.client.model_client.openai_client import OpenAIClient
-from llm_bridge.type.chat_response import ChatResponse, File
+from llm_bridge.type.chat_response import ChatResponse
 from llm_bridge.type.serializer import serialize
 
 
-def process_delta(event: ResponseStreamEvent) -> ChatResponse:
-    text: str = ""
-    thought: str = ""
-    files: list[File] = []
-
-    if event.type == "response.output_text.delta":
-        text_delta_event: ResponseTextDeltaEvent = event
-        text = text_delta_event.delta
-    elif event.type == "response.reasoning_summary_text.delta":
-        reasoning_summary_text_delta_event: ResponseReasoningSummaryTextDeltaEvent = event
-        thought = reasoning_summary_text_delta_event.delta
-    if event.type == "response.image_generation_call.partial_image":
-        file = File(
-            name="generated_image.png",
-            data=event.partial_image_b64,
-            type="image/png",
-        )
-        files.append(file)
-
-    chat_response = ChatResponse(
-        text=text,
-        thought=thought,
-        files=files,
-    )
-    return chat_response
-
-
 async def generate_chunk(
         stream: AsyncStream[ResponseStreamEvent],
         input_tokens: int,
 ) -> AsyncGenerator[ChatResponse, None]:
     try:
         async for event in stream:
-            chat_response = process_delta(event)
+            chat_response = await process_openai_responses_stream_response(event)
             output_tokens = count_openai_output_tokens(chat_response)
             yield ChatResponse(
                 text=chat_response.text,
                 thought=chat_response.thought,
+                code=chat_response.code,
+                code_output=chat_response.code_output,
                 files=chat_response.files,
                 input_tokens=input_tokens,
                 output_tokens=output_tokens,
@@ -78,6 +55,7 @@ class StreamOpenAIResponsesClient(OpenAIClient):
                 temperature=self.temperature,
                 stream=True,
                 tools=self.tools,
+                include=self.include,
                 # text_format=self.structured_output_base_model, # Async OpenAPI Responses Client does not support structured output
             )
 

{llm_bridge-1.15.2 → llm_bridge-1.15.3}/llm_bridge/client/model_client/openai_client.py

@@ -3,7 +3,7 @@ from pydantic import BaseModel
 
 import openai.lib.azure
 from openai.types import Reasoning
-from openai.types.responses import ToolParam
+from openai.types.responses import ToolParam, ResponseIncludable
 
 from llm_bridge.client.chat_client import ChatClient
 from llm_bridge.type.chat_response import ChatResponse
@@ -21,6 +21,7 @@ class OpenAIClient(ChatClient):
             client: openai.AsyncOpenAI | openai.lib.azure.AsyncAzureOpenAI,
             tools: Iterable[ToolParam],
             reasoning: Reasoning,
+            include: list[ResponseIncludable],
             structured_output_base_model: Type[BaseModel] | None = None,
     ):
         self.model = model
@@ -30,6 +31,7 @@ class OpenAIClient(ChatClient):
         self.client = client
         self.tools = tools
         self.reasoning = reasoning
+        self.include = include
         self.structured_output_base_model = structured_output_base_model
 
     async def generate_non_stream_response(self) -> ChatResponse:

{llm_bridge-1.15.2 → llm_bridge-1.15.3}/llm_bridge/logic/chat_generate/model_client_factory/openai_client_factory.py

@@ -3,10 +3,11 @@ from typing import Any
 
 import openai
 from fastapi import HTTPException
+from openai import Omit
 from openai.types import Reasoning
-from openai.types.responses import WebSearchToolParam
+from openai.types.responses import WebSearchToolParam, ResponseIncludable
 from openai.types.responses.tool_param import CodeInterpreter, CodeInterpreterContainerCodeInterpreterToolAuto, \
-    ImageGeneration
+    ImageGeneration, ToolParam
 
 from llm_bridge.client.implementations.openai.non_stream_openai_client import NonStreamOpenAIClient
 from llm_bridge.client.implementations.openai.non_stream_openai_responses_client import NonStreamOpenAIResponsesClient
@@ -29,6 +30,8 @@ async def create_openai_client(
         code_execution: bool,
         structured_output_schema: dict[str, Any] | None,
 ):
+    omit = Omit()
+
     if api_type == "OpenAI":
         client = openai.AsyncOpenAI(
             api_key=api_keys["OPENAI_API_KEY"],
@@ -62,8 +65,9 @@ async def create_openai_client(
     else:
         openai_messages = await convert_messages_to_openai(messages)
 
-    tools = []
-    reasoning = None
+    tools: list[ToolParam] = []
+    reasoning: Reasoning | Omit = omit
+    include: list[ResponseIncludable] = ["code_interpreter_call.outputs"]
 
     if model not in ["gpt-5-pro", "gpt-5.2-pro"] and "codex" not in model:
         if code_execution:
@@ -109,6 +113,7 @@ async def create_openai_client(
                 client=client,
                 tools=tools,
                 reasoning=reasoning,
+                include=include,
                 structured_output_base_model=structured_output_base_model,
             )
         else:
@@ -120,6 +125,7 @@ async def create_openai_client(
                 client=client,
                 tools=tools,
                 reasoning=reasoning,
+                include=include,
                 structured_output_base_model=structured_output_base_model,
             )
     else:

{llm_bridge-1.15.2 → llm_bridge-1.15.3}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "LLM-Bridge"
-version = "1.15.2"
+version = "1.15.3"
 dependencies = [
     "fastapi",
     "httpx",

{llm_bridge-1.15.2 → llm_bridge-1.15.3}/usage/main.py

@@ -70,7 +70,7 @@ structured_output_schema = {
     "price"
   ]
 }
-# structured_output_schema = None
+structured_output_schema = None
 
 messages = [
     Message(
@@ -101,8 +101,8 @@ messages = [
             # Content(type=ContentType.Text, data="What's the weather in NYC today?"),
 
             # Image Understanding
-            Content(type=ContentType.File, data="https://www.gstatic.com/webp/gallery3/1.png"),
-            Content(type=ContentType.Text, data="What is in this image?"),
+            # Content(type=ContentType.File, data="https://www.gstatic.com/webp/gallery3/1.png"),
+            # Content(type=ContentType.Text, data="What is in this image?"),
 
             # Image Generation
             # Content(type=ContentType.Text, data="Please generate an image of a cat."),
@@ -114,7 +114,7 @@ messages = [
             # Content(type=ContentType.Text, data="What is the sum of the first 50 prime numbers? Generate and run code for the calculation, and make sure you get all 50."),
 
             # File Output
-            # Content(type=ContentType.Text, data="Create a matplotlib visualization and save it as output.png"),
+            Content(type=ContentType.Text, data="Create a matplotlib visualization and save it as output.png"),
 
             # Structured Output
             # Content(type=ContentType.Text, data="Please generate a product."),
@@ -133,12 +133,12 @@ messages = [
     # ),
 ]
 # See /llm_bridge/resources/model_prices.json for available models
-# model = "gpt-5.2"
+model = "gpt-5.2"
 # model = "gpt-5.1"
 # model = "gpt-5-pro"
 # model = "gpt-5"
 # model = "gpt-4.1"
-model = "gpt-5-codex"
+# model = "gpt-5-codex"
 # model = "gemini-3-pro-preview"
 # model = "gemini-3-pro-image-preview"
 # model = "gemini-3-flash-preview"

{llm_bridge-1.15.2 → llm_bridge-1.15.3}/uv.lock

@@ -357,7 +357,7 @@ wheels = [
 
 [[package]]
 name = "llm-bridge"
-version = "1.15.2"
+version = "1.15.3"
 source = { editable = "." }
 dependencies = [
     { name = "anthropic" },