LLM-Bridge 1.15.0a0__tar.gz → 1.15.2__tar.gz

{llm_bridge-1.15.0a0 → llm_bridge-1.15.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: LLM-Bridge
-Version: 1.15.0a0
+Version: 1.15.2
 Summary: A Bridge for LLMs
 Author-email: windsnow1025 <windsnow1025@gmail.com>
 License-Expression: MIT
@@ -48,7 +48,7 @@ The features listed represent the maximum capabilities of each API type supporte
 
 | API Type | Input Format                   | Capabilities                                            | Output Format     |
 |----------|--------------------------------|---------------------------------------------------------|-------------------|
-| OpenAI   | Text, Image, PDF               | Thinking, Web Search, Code Execution                    | Text              |
+| 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              |
 | Grok     | Text, Image                    |                                                         | Text              |

{llm_bridge-1.15.0a0 → llm_bridge-1.15.2}/README.md

@@ -24,7 +24,7 @@ The features listed represent the maximum capabilities of each API type supporte
 
 | API Type | Input Format                   | Capabilities                                            | Output Format     |
 |----------|--------------------------------|---------------------------------------------------------|-------------------|
-| OpenAI   | Text, Image, PDF               | Thinking, Web Search, Code Execution                    | Text              |
+| 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              |
 | Grok     | Text, Image                    |                                                         | Text              |

{llm_bridge-1.15.0a0 → llm_bridge-1.15.2}/llm_bridge/client/implementations/claude/claude_response_handler.py

@@ -1,20 +1,36 @@
 from anthropic import BetaMessageStreamEvent, AsyncAnthropic
+from anthropic._response import AsyncBinaryAPIResponse
 from anthropic.types.beta import BetaRawContentBlockDelta, BetaThinkingDelta, BetaTextDelta, BetaInputJSONDelta, \
     BetaBashCodeExecutionToolResultBlock, \
     BetaTextEditorCodeExecutionToolResultBlock, BetaTextEditorCodeExecutionViewResultBlock, \
     BetaTextEditorCodeExecutionStrReplaceResultBlock, \
-    BetaServerToolUseBlock, BetaBashCodeExecutionResultBlock, BetaTextBlock, BetaThinkingBlock
+    BetaServerToolUseBlock, BetaBashCodeExecutionResultBlock, BetaTextBlock, BetaThinkingBlock, \
+    BetaBashCodeExecutionOutputBlock, BetaMessage, FileMetadata
 from anthropic.types.beta.beta_raw_content_block_start_event import ContentBlock
 
 from llm_bridge.client.implementations.claude.claude_token_counter import count_claude_output_tokens
-from llm_bridge.type.chat_response import ChatResponse
-
+from llm_bridge.logic.chat_generate.media_processor import bytes_to_base64
+from llm_bridge.type.chat_response import ChatResponse, File
+
+
+async def download_claude_file(client: AsyncAnthropic, file_id: str) -> File:
+    file_metadata: FileMetadata = await client.beta.files.retrieve_metadata(file_id)
+    file_content: AsyncBinaryAPIResponse = await client.beta.files.download(file_id)
+    data = await file_content.read()
+    return File(
+        name=file_metadata.filename,
+        data=bytes_to_base64(data),
+        type=file_metadata.mime_type,
+    )
 
-def process_content_block(content_block: ContentBlock) -> ChatResponse:
-    text = ""
-    thought = ""
-    code = ""
-    code_output = ""
+async def process_content_block(
+        content_block: ContentBlock, client: AsyncAnthropic
+) -> ChatResponse:
+    text: str = ""
+    thought: str = ""
+    code: str = ""
+    code_output: str = ""
+    files: list[File] = []
 
     if content_block.type == "text":
         text_block: BetaTextBlock = content_block
@@ -31,23 +47,72 @@ def process_content_block(content_block: ContentBlock) -> ChatResponse:
     elif content_block.type == "bash_code_execution_tool_result":
         bash_code_execution_tool_result_block: BetaBashCodeExecutionToolResultBlock = content_block
         if bash_code_execution_tool_result_block.content.type == "bash_code_execution_result":
-            content: BetaBashCodeExecutionResultBlock = content_block.content
-            code_output += content.stdout
+            result: BetaBashCodeExecutionResultBlock = content_block.content
+            code_output += result.stdout
+            outputs: list[BetaBashCodeExecutionOutputBlock] = result.content
+            file_ids = [output.file_id for output in outputs]
+            for file_id in file_ids:
+                file = await download_claude_file(client, file_id)
+                files.append(file)
 
     elif content_block.type == "text_editor_code_execution_tool_result":
         text_editor_code_execution_tool_result: BetaTextEditorCodeExecutionToolResultBlock = content_block
         if text_editor_code_execution_tool_result.content.type == "text_editor_code_execution_view_result":
-            content: BetaTextEditorCodeExecutionViewResultBlock = content_block.content
-            code_output += content.content
+            result: BetaTextEditorCodeExecutionViewResultBlock = content_block.content
+            code_output += result.content
         elif text_editor_code_execution_tool_result.content.type == "text_editor_code_execution_str_replace_result":
-            content: BetaTextEditorCodeExecutionStrReplaceResultBlock = content_block.content
-            code_output += content.lines
+            result: BetaTextEditorCodeExecutionStrReplaceResultBlock = content_block.content
+            code_output += result.lines
+
+    return ChatResponse(
+        text=text,
+        thought=thought,
+        code=code,
+        code_output=code_output,
+        files=files,
+    )
+
+
+async def process_claude_non_stream_response(
+        message: BetaMessage,
+        input_tokens: int,
+        client: AsyncAnthropic,
+        model: str,
+) -> ChatResponse:
+    text = ""
+    thought = ""
+    code = ""
+    code_output = ""
+    files: list[File] = []
 
+    for content_block in message.content:
+        content_block_chat_response = await process_content_block(content_block, client)
+        text += content_block_chat_response.text
+        thought += content_block_chat_response.thought
+        code += content_block_chat_response.code
+        code_output += content_block_chat_response.code_output
+        files.extend(content_block_chat_response.files)
+
+    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,
     )
 
 
@@ -61,6 +126,7 @@ async def process_claude_stream_response(
     thought = ""
     code = ""
     code_output = ""
+    files: list[File] = []
 
     if event.type == "content_block_delta":
         event_delta: BetaRawContentBlockDelta = event.delta
@@ -79,17 +145,19 @@ async def process_claude_stream_response(
 
     if event.type == "content_block_start":
         content_block: ContentBlock = event.content_block
-        content_block_chat_response = process_content_block(content_block)
+        content_block_chat_response = await process_content_block(content_block, client)
         text += content_block_chat_response.text
         thought += content_block_chat_response.thought
         code += content_block_chat_response.code
         code_output += content_block_chat_response.code_output
+        files.extend(content_block_chat_response.files)
 
     chat_response = ChatResponse(
         text=text,
         thought=thought,
         code=code,
         code_output=code_output,
+        files=files,
     )
     output_tokens = await count_claude_output_tokens(
         client=client,
@@ -101,6 +169,7 @@ async def process_claude_stream_response(
         thought=thought,
         code=code,
         code_output=code_output,
+        files=files,
         input_tokens=input_tokens,
         output_tokens=output_tokens,
     )

{llm_bridge-1.15.0a0 → llm_bridge-1.15.2}/llm_bridge/client/implementations/claude/non_stream_claude_client.py

@@ -2,57 +2,14 @@ import logging
 import re
 
 import httpx
-from anthropic import AsyncAnthropic
-from anthropic.types.beta import BetaMessage, BetaBashCodeExecutionToolResultBlock, BetaTextBlock, BetaThinkingBlock, \
-    BetaServerToolUseBlock
 from fastapi import HTTPException
 
-from llm_bridge.client.implementations.claude.claude_response_handler import process_content_block
-from llm_bridge.client.implementations.claude.claude_token_counter import count_claude_output_tokens
+from llm_bridge.client.implementations.claude.claude_response_handler import process_claude_non_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
 
 
-async def process_claude_non_stream_response(
-        message: BetaMessage,
-        input_tokens: int,
-        client: AsyncAnthropic,
-        model: str,
-) -> ChatResponse:
-    text = ""
-    thought = ""
-    code = ""
-    code_output = ""
-
-    for content_block in message.content:
-        content_block_chat_response = process_content_block(content_block)
-        text += content_block_chat_response.text
-        thought += content_block_chat_response.thought
-        code += content_block_chat_response.code
-        code_output += content_block_chat_response.code_output
-
-    chat_response = ChatResponse(
-        text=text,
-        thought=thought,
-        code=code,
-        code_output=code_output,
-    )
-    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,
-        input_tokens=input_tokens,
-        output_tokens=output_tokens,
-    )
-
-
 class NonStreamClaudeClient(ClaudeClient):
     async def generate_non_stream_response(self) -> ChatResponse:
         try:

{llm_bridge-1.15.0a0 → llm_bridge-1.15.2}/llm_bridge/client/implementations/gemini/gemini_response_handler.py

@@ -7,6 +7,7 @@ from google.genai.types import Part
 
 from llm_bridge.client.implementations.gemini.gemini_token_counter import count_gemini_tokens
 from llm_bridge.client.implementations.printing_status import PrintingStatus
+from llm_bridge.logic.chat_generate.media_processor import bytes_to_base64
 from llm_bridge.type.chat_response import ChatResponse, File
 
 
@@ -57,7 +58,7 @@ class GeminiResponseHandler:
                 extension = mimetypes.guess_extension(mime_type) or ""
                 file = File(
                     name=f"generated_file{extension}",
-                    data=base64.b64encode(part.inline_data.data).decode('utf-8'),
+                    data=bytes_to_base64(part.inline_data.data),
                     type=mime_type,
                 )
                 files.append(file)

llm_bridge-1.15.2/llm_bridge/logic/chat_generate/media_processor.py

@@ -0,0 +1,23 @@
+import base64
+
+from llm_bridge.logic.file_fetch import fetch_file_data
+
+
+def bytes_to_base64(data: bytes) -> str:
+    return base64.b64encode(data).decode('utf-8')
+
+
+async def get_bytes_content_from_url(req_url: str) -> tuple[bytes, str]:
+    file_data, media_type = await fetch_file_data(req_url)
+    return file_data, media_type
+
+
+async def get_base64_content_from_url(req_url: str) -> tuple[str, str]:
+    media_data, media_type = await get_bytes_content_from_url(req_url)
+    base64_media = bytes_to_base64(media_data)
+    return base64_media, media_type
+
+
+async def get_openai_image_content_from_url(req_img_url: str) -> str:
+    base64_image, media_type = await get_base64_content_from_url(req_img_url)
+    return f"data:{media_type};base64,{base64_image}"

{llm_bridge-1.15.0a0 → llm_bridge-1.15.2}/llm_bridge/logic/chat_generate/model_client_factory/claude_client_factory.py

@@ -3,7 +3,8 @@ 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
+from anthropic.types.beta import BetaWebSearchTool20250305Param, BetaToolUnionParam, BetaCodeExecutionTool20250825Param, \
+    BetaToolBash20250124Param
 
 from llm_bridge.client.implementations.claude.claude_token_counter import count_claude_input_tokens
 from llm_bridge.client.implementations.claude.non_stream_claude_client import NonStreamClaudeClient
@@ -61,7 +62,8 @@ 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"
+        "files-api-2025-04-14",
+        "structured-outputs-2025-11-13",
     ]
     tools: list[BetaToolUnionParam] = []
     tools.append(
@@ -71,6 +73,12 @@ async def create_claude_client(
         )
     )
     if code_execution:
+        tools.append(
+            BetaToolBash20250124Param(
+                type="bash_20250124",
+                name="bash",
+            )
+        )
         tools.append(
             BetaCodeExecutionTool20250825Param(
                 type="code_execution_20250825",

{llm_bridge-1.15.0a0 → llm_bridge-1.15.2}/llm_bridge/logic/chat_generate/model_client_factory/openai_client_factory.py

@@ -65,7 +65,7 @@ async def create_openai_client(
     tools = []
     reasoning = None
 
-    if model not in ["gpt-5-pro", "gpt-5.2-pro"]:
+    if model not in ["gpt-5-pro", "gpt-5.2-pro"] and "codex" not in model:
         if code_execution:
             tools.append(
                 CodeInterpreter(
@@ -87,6 +87,7 @@ async def create_openai_client(
                 effort="high",
                 summary="auto",
             )
+    if re.match(r"gpt-5.*", model) and "codex" not in model:
         tools.append(
             ImageGeneration(
                 type="image_generation",

{llm_bridge-1.15.0a0 → llm_bridge-1.15.2}/llm_bridge/logic/chat_generate/model_message_converter/claude_message_converter.py

@@ -25,7 +25,7 @@ async def convert_message_to_claude(message: Message) -> ClaudeMessage:
             file_url = content_item.data
             file_type, sub_type = await get_file_type(file_url)
             if file_type == "image":
-                base64_image, media_type = await media_processor.get_encoded_content_from_url(file_url)
+                base64_image, media_type = await media_processor.get_base64_content_from_url(file_url)
                 image_content = ImageBlockParam(
                     type="image",
                     source=Base64ImageSourceParam(
@@ -36,7 +36,7 @@ async def convert_message_to_claude(message: Message) -> ClaudeMessage:
                 )
                 claude_content.append(image_content)
             elif sub_type == "pdf":
-                file_data, media_type = await media_processor.get_encoded_content_from_url(file_url)
+                file_data, media_type = await media_processor.get_base64_content_from_url(file_url)
                 pdf_content = DocumentBlockParam(
                     type="document",
                     source=Base64PDFSourceParam(

{llm_bridge-1.15.0a0 → llm_bridge-1.15.2}/llm_bridge/logic/chat_generate/model_message_converter/gemini_message_converter.py

@@ -23,7 +23,7 @@ async def convert_message_to_gemini(message: Message) -> GeminiMessage:
             file_url = content_item.data
             file_type, sub_type = await get_file_type(file_url)
             if sub_type == "pdf" or file_type in ("image", "video", "audio"):
-                file_data, media_type = await media_processor.get_raw_content_from_url(file_url)
+                file_data, media_type = await media_processor.get_bytes_content_from_url(file_url)
                 if media_type == 'video/webm':
                     media_type = 'audio/webm'
                 parts.append(types.Part.from_bytes(data=file_data, mime_type=media_type))

{llm_bridge-1.15.0a0 → llm_bridge-1.15.2}/llm_bridge/logic/chat_generate/model_message_converter/openai_message_converter.py

@@ -28,7 +28,7 @@ async def convert_message_to_openai(message: Message) -> OpenAIMessage:
                 )
                 content.append(image_content)
             elif file_type == "audio":
-                encoded_string, _ = await media_processor.get_encoded_content_from_url(file_url)
+                encoded_string, _ = await media_processor.get_base64_content_from_url(file_url)
                 audio_content = ChatCompletionContentPartInputAudioParam(
                     type="input_audio",
                     input_audio=InputAudio(data=encoded_string, format=sub_type)

{llm_bridge-1.15.0a0 → llm_bridge-1.15.2}/llm_bridge/logic/chat_generate/model_message_converter/openai_responses_message_converter.py

@@ -30,7 +30,7 @@ async def convert_message_to_openai_responses(message: Message) -> OpenAIRespons
                 )
                 content.append(image_content)
             elif sub_type == "pdf":
-                file_data, _ = await media_processor.get_encoded_content_from_url(file_url)
+                file_data, _ = await media_processor.get_base64_content_from_url(file_url)
                 pdf_content = ResponseInputFileParam(
                     type="input_file",
                     filename=get_file_name(file_url),

{llm_bridge-1.15.0a0 → llm_bridge-1.15.2}/llm_bridge/resources/model_prices.json

@@ -1,51 +1,51 @@
 [
   {
-    "apiType": "Gemini-Vertex",
+    "apiType": "Gemini-Paid",
     "model": "gemini-3-pro-preview",
     "input": 4,
     "output": 18
   },
   {
-    "apiType": "Gemini-Vertex",
+    "apiType": "Gemini-Paid",
     "model": "gemini-3-flash-preview",
     "input": 1,
     "output": 3
   },
   {
-    "apiType": "Gemini-Vertex",
+    "apiType": "Gemini-Paid",
     "model": "gemini-3-pro-image-preview",
     "input": 2,
     "output": 120
   },
-  {
-    "apiType": "Gemini-Free",
-    "model": "gemini-3-flash-preview",
-    "input": 0,
-    "output": 0
-  },
   {
     "apiType": "Gemini-Paid",
-    "model": "gemini-3-flash-preview",
+    "model": "gemini-flash-latest",
     "input": 1,
-    "output": 3
+    "output": 2.5
   },
   {
-    "apiType": "Gemini-Paid",
+    "apiType": "Gemini-Vertex",
     "model": "gemini-3-pro-preview",
     "input": 4,
     "output": 18
   },
   {
-    "apiType": "Gemini-Paid",
+    "apiType": "Gemini-Vertex",
+    "model": "gemini-3-flash-preview",
+    "input": 1,
+    "output": 3
+  },
+  {
+    "apiType": "Gemini-Vertex",
     "model": "gemini-3-pro-image-preview",
     "input": 2,
     "output": 120
   },
   {
-    "apiType": "Gemini-Paid",
-    "model": "gemini-flash-latest",
-    "input": 1,
-    "output": 2.5
+    "apiType": "Gemini-Free",
+    "model": "gemini-3-flash-preview",
+    "input": 0,
+    "output": 0
   },
   {
     "apiType": "OpenAI",
@@ -59,12 +59,30 @@
     "input": 1.25,
     "output": 10
   },
+  {
+    "apiType": "OpenAI",
+    "model": "gpt-5.1-codex-max",
+    "input": 1.25,
+    "output": 10
+  },
+  {
+    "apiType": "OpenAI",
+    "model": "gpt-5.1-codex",
+    "input": 1.25,
+    "output": 10
+  },
   {
     "apiType": "OpenAI",
     "model": "gpt-5",
     "input": 1.25,
     "output": 10
   },
+  {
+    "apiType": "OpenAI",
+    "model": "gpt-5-codex",
+    "input": 1.25,
+    "output": 10
+  },
   {
     "apiType": "OpenAI",
     "model": "gpt-5-mini",

{llm_bridge-1.15.0a0 → llm_bridge-1.15.2}/llm_bridge/type/chat_response.py

@@ -5,7 +5,7 @@ from typing import Optional
 @dataclass
 class File:
     name: str
-    data: str
+    data: str # Base64
     type: str
 
 

{llm_bridge-1.15.0a0 → llm_bridge-1.15.2}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "LLM-Bridge"
-version = "1.15.0-alpha.0"
+version = "1.15.2"
 dependencies = [
     "fastapi",
     "httpx",

{llm_bridge-1.15.0a0 → llm_bridge-1.15.2}/usage/main.py

@@ -98,11 +98,11 @@ messages = [
             # Content(type=ContentType.Text, data="Explain the concept of Occam's Razor and provide a simple, everyday example."),
 
             # Web Search
-            Content(type=ContentType.Text, data="What's the weather in NYC today?"),
+            # 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."),
@@ -114,8 +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.File, data="https://www.windsnow1025.com/minio/windsnow/uploads/1/1758384216123-script.py"),
-            # Content(type=ContentType.Text, data="Please implement a minimum example of Neural Network in `script.py`"),
+            # 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."),
@@ -139,23 +138,24 @@ messages = [
 # model = "gpt-5-pro"
 # model = "gpt-5"
 # model = "gpt-4.1"
-model = "gemini-3-pro-preview"
+model = "gpt-5-codex"
+# model = "gemini-3-pro-preview"
 # model = "gemini-3-pro-image-preview"
 # model = "gemini-3-flash-preview"
 # model = "grok-4-1-fast-reasoning"
 # model = "claude-sonnet-4-5"
 # model = "claude-opus-4-5"
-api_type = "Gemini-Vertex"
+# api_type = "Gemini-Vertex"
 # api_type = "Gemini-Free"
 # api_type = "Gemini-Paid"
-# api_type = "OpenAI"
+api_type = "OpenAI"
 # api_type = "OpenAI-Azure"
 # api_type = "OpenAI-GitHub"
 # api_type = "Claude"
 # api_type = "Grok"
 temperature = 0
-# stream = True
-stream = False
+stream = True
+# stream = False
 thought = True
 # thought = False
 code_execution = True

{llm_bridge-1.15.0a0 → llm_bridge-1.15.2}/uv.lock

@@ -357,7 +357,7 @@ wheels = [
 
 [[package]]
 name = "llm-bridge"
-version = "1.14.1"
+version = "1.15.2"
 source = { editable = "." }
 dependencies = [
     { name = "anthropic" },

llm_bridge-1.15.0a0/llm_bridge/logic/chat_generate/media_processor.py

@@ -1,20 +0,0 @@
-import base64
-
-from llm_bridge.logic.file_fetch import fetch_file_data
-
-
-async def get_raw_content_from_url(req_url: str) -> tuple[bytes, str]:
-    file_data, media_type = await fetch_file_data(req_url)
-    return file_data, media_type
-
-
-# Base64 Encoded
-async def get_encoded_content_from_url(req_url: str) -> tuple[str, str]:
-    media_data, media_type = await get_raw_content_from_url(req_url)
-    base64_media = base64.b64encode(media_data).decode('utf-8')
-    return base64_media, media_type
-
-
-async def get_openai_image_content_from_url(req_img_url: str) -> str:
-    base64_image, media_type = await get_encoded_content_from_url(req_img_url)
-    return f"data:{media_type};base64,{base64_image}"