LLM-Bridge 1.11.9__tar.gz → 1.12.0__tar.gz

llm_bridge-1.12.0/LLM_Bridge.egg-info/PKG-INFO

@@ -0,0 +1,88 @@
+Metadata-Version: 2.4
+Name: LLM-Bridge
+Version: 1.12.0
+Summary: A Bridge for LLMs
+Author-email: windsnow1025 <windsnow1025@gmail.com>
+License-Expression: MIT
+Keywords: llm,ai
+Classifier: Framework :: FastAPI
+Classifier: Programming Language :: Python :: 3
+Requires-Python: >=3.12
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: fastapi
+Requires-Dist: httpx
+Requires-Dist: tenacity
+Requires-Dist: openai==1.106.1
+Requires-Dist: tiktoken==0.11.0
+Requires-Dist: google-genai==1.46.0
+Requires-Dist: anthropic==0.66.0
+Requires-Dist: PyMuPDF
+Requires-Dist: docxlatex>=1.1.1
+Requires-Dist: openpyxl
+Requires-Dist: python-pptx
+Provides-Extra: test
+Requires-Dist: pytest; extra == "test"
+Requires-Dist: pytest-asyncio; extra == "test"
+Requires-Dist: python-dotenv; extra == "test"
+Requires-Dist: protobuf; extra == "test"
+Dynamic: license-file
+
+# LLM Bridge
+
+LLM Bridge is a unified Python interface for interacting with LLMs, including OpenAI (Native / Azure / GitHub), Gemini (AI Studio / Vertex), Claude, and Grok.
+
+GitHub: [https://github.com/windsnow1025/LLM-Bridge](https://github.com/windsnow1025/LLM-Bridge)
+
+PyPI: [https://pypi.org/project/LLM-Bridge/](https://pypi.org/project/LLM-Bridge/)
+
+## Workflow and Features
+
+1. **Message Preprocessor**: extracts text content from documents (Word, Excel, PPT, Code files, PDFs) which are not natively supported by the target model.
+2. **Chat Client Factory**: creates a client for the specific LLM API with model parameters
+    1. **Model Message Converter**: converts general messages to model messages
+        1. **Media Processor**: converts general media (Image, Audio, Video, PDF) to model compatible formats.
+3. **Chat Client**: generate stream or non-stream responses
+    - **Model Thoughts**: captures and formats the model's thinking process
+    - **Code Execution**: auto generate and execute Python code
+    - **Web Search + Citations**: extracts and formats citations from search results
+    - **Token Counter**: tracks and reports input and output token usage
+
+### Supported Features for API Types
+
+The features listed represent the maximum capabilities of each API type supported by LLM Bridge.
+
+| API Type | Input Format                   | Capabilities                                     | Output Format     |
+|----------|--------------------------------|--------------------------------------------------|-------------------|
+| OpenAI   | Text, Image, PDF               | Thinking, Web Search, Code Execution             | Text              |
+| Gemini   | Text, Image, Video, Audio, PDF | Thinking, Web Search + Citations, Code Execution | Text, Image, File |
+| Claude   | Text, Image, PDF               | Thinking, Web Search, Code Execution             | Text              |
+| Grok     | Text, Image                    |                                                  | Text              |
+
+#### Planned Features
+
+- Structured Output
+- More features for API Types
+- Native support for Grok
+
+## Installation
+
+```bash
+pip install --upgrade llm_bridge
+```
+
+## Test
+
+```bash
+pytest
+```
+
+## Quick Start
+
+### Setup
+
+1. Copy `./usage/.env.example` and rename it to `./usage/.env`, then fill in the environment variables.
+2. Install requirements: `pip install -r requirements.txt`
+3. In PyCharm, add a new Python configuration:
+   - script: `./usage/main.py`
+   - Paths to ".env" files: `./usage/.env`

llm_bridge-1.12.0/PKG-INFO

@@ -0,0 +1,88 @@
+Metadata-Version: 2.4
+Name: LLM-Bridge
+Version: 1.12.0
+Summary: A Bridge for LLMs
+Author-email: windsnow1025 <windsnow1025@gmail.com>
+License-Expression: MIT
+Keywords: llm,ai
+Classifier: Framework :: FastAPI
+Classifier: Programming Language :: Python :: 3
+Requires-Python: >=3.12
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: fastapi
+Requires-Dist: httpx
+Requires-Dist: tenacity
+Requires-Dist: openai==1.106.1
+Requires-Dist: tiktoken==0.11.0
+Requires-Dist: google-genai==1.46.0
+Requires-Dist: anthropic==0.66.0
+Requires-Dist: PyMuPDF
+Requires-Dist: docxlatex>=1.1.1
+Requires-Dist: openpyxl
+Requires-Dist: python-pptx
+Provides-Extra: test
+Requires-Dist: pytest; extra == "test"
+Requires-Dist: pytest-asyncio; extra == "test"
+Requires-Dist: python-dotenv; extra == "test"
+Requires-Dist: protobuf; extra == "test"
+Dynamic: license-file
+
+# LLM Bridge
+
+LLM Bridge is a unified Python interface for interacting with LLMs, including OpenAI (Native / Azure / GitHub), Gemini (AI Studio / Vertex), Claude, and Grok.
+
+GitHub: [https://github.com/windsnow1025/LLM-Bridge](https://github.com/windsnow1025/LLM-Bridge)
+
+PyPI: [https://pypi.org/project/LLM-Bridge/](https://pypi.org/project/LLM-Bridge/)
+
+## Workflow and Features
+
+1. **Message Preprocessor**: extracts text content from documents (Word, Excel, PPT, Code files, PDFs) which are not natively supported by the target model.
+2. **Chat Client Factory**: creates a client for the specific LLM API with model parameters
+    1. **Model Message Converter**: converts general messages to model messages
+        1. **Media Processor**: converts general media (Image, Audio, Video, PDF) to model compatible formats.
+3. **Chat Client**: generate stream or non-stream responses
+    - **Model Thoughts**: captures and formats the model's thinking process
+    - **Code Execution**: auto generate and execute Python code
+    - **Web Search + Citations**: extracts and formats citations from search results
+    - **Token Counter**: tracks and reports input and output token usage
+
+### Supported Features for API Types
+
+The features listed represent the maximum capabilities of each API type supported by LLM Bridge.
+
+| API Type | Input Format                   | Capabilities                                     | Output Format     |
+|----------|--------------------------------|--------------------------------------------------|-------------------|
+| OpenAI   | Text, Image, PDF               | Thinking, Web Search, Code Execution             | Text              |
+| Gemini   | Text, Image, Video, Audio, PDF | Thinking, Web Search + Citations, Code Execution | Text, Image, File |
+| Claude   | Text, Image, PDF               | Thinking, Web Search, Code Execution             | Text              |
+| Grok     | Text, Image                    |                                                  | Text              |
+
+#### Planned Features
+
+- Structured Output
+- More features for API Types
+- Native support for Grok
+
+## Installation
+
+```bash
+pip install --upgrade llm_bridge
+```
+
+## Test
+
+```bash
+pytest
+```
+
+## Quick Start
+
+### Setup
+
+1. Copy `./usage/.env.example` and rename it to `./usage/.env`, then fill in the environment variables.
+2. Install requirements: `pip install -r requirements.txt`
+3. In PyCharm, add a new Python configuration:
+   - script: `./usage/main.py`
+   - Paths to ".env" files: `./usage/.env`

llm_bridge-1.12.0/README.md

@@ -0,0 +1,58 @@
+# LLM Bridge
+
+LLM Bridge is a unified Python interface for interacting with LLMs, including OpenAI (Native / Azure / GitHub), Gemini (AI Studio / Vertex), Claude, and Grok.
+
+GitHub: [https://github.com/windsnow1025/LLM-Bridge](https://github.com/windsnow1025/LLM-Bridge)
+
+PyPI: [https://pypi.org/project/LLM-Bridge/](https://pypi.org/project/LLM-Bridge/)
+
+## Workflow and Features
+
+1. **Message Preprocessor**: extracts text content from documents (Word, Excel, PPT, Code files, PDFs) which are not natively supported by the target model.
+2. **Chat Client Factory**: creates a client for the specific LLM API with model parameters
+    1. **Model Message Converter**: converts general messages to model messages
+        1. **Media Processor**: converts general media (Image, Audio, Video, PDF) to model compatible formats.
+3. **Chat Client**: generate stream or non-stream responses
+    - **Model Thoughts**: captures and formats the model's thinking process
+    - **Code Execution**: auto generate and execute Python code
+    - **Web Search + Citations**: extracts and formats citations from search results
+    - **Token Counter**: tracks and reports input and output token usage
+
+### Supported Features for API Types
+
+The features listed represent the maximum capabilities of each API type supported by LLM Bridge.
+
+| API Type | Input Format                   | Capabilities                                     | Output Format     |
+|----------|--------------------------------|--------------------------------------------------|-------------------|
+| OpenAI   | Text, Image, PDF               | Thinking, Web Search, Code Execution             | Text              |
+| Gemini   | Text, Image, Video, Audio, PDF | Thinking, Web Search + Citations, Code Execution | Text, Image, File |
+| Claude   | Text, Image, PDF               | Thinking, Web Search, Code Execution             | Text              |
+| Grok     | Text, Image                    |                                                  | Text              |
+
+#### Planned Features
+
+- Structured Output
+- More features for API Types
+- Native support for Grok
+
+## Installation
+
+```bash
+pip install --upgrade llm_bridge
+```
+
+## Test
+
+```bash
+pytest
+```
+
+## Quick Start
+
+### Setup
+
+1. Copy `./usage/.env.example` and rename it to `./usage/.env`, then fill in the environment variables.
+2. Install requirements: `pip install -r requirements.txt`
+3. In PyCharm, add a new Python configuration:
+   - script: `./usage/main.py`
+   - Paths to ".env" files: `./usage/.env`

{llm_bridge-1.11.9 → llm_bridge-1.12.0}/llm_bridge/client/implementations/claude/non_stream_claude_client.py

@@ -58,16 +58,27 @@ class NonStreamClaudeClient(ClaudeClient):
         try:
             logging.info(f"messages: {self.messages}")
 
-            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),
-                thinking=self.thinking,
-                betas=self.betas,
-                tools=self.tools,
-            )
+            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,
+                )
 
             return await process_claude_non_stream_response(
                 message=message,

llm_bridge-1.12.0/llm_bridge/client/implementations/claude/stream_claude_client.py

@@ -0,0 +1,71 @@
+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))

{llm_bridge-1.11.9 → llm_bridge-1.12.0}/llm_bridge/logic/chat_generate/chat_client_factory.py

@@ -8,86 +8,104 @@ from llm_bridge.type.message import Message
 
 
 async def create_chat_client(
+        api_keys: dict,
         messages: list[Message],
         model: str,
         api_type: str,
         temperature: float,
         stream: bool,
-        api_keys: dict
+        thought: bool,
+        code_execution: bool,
 ) -> ChatClient:
     if api_type == 'OpenAI':
         return await create_openai_client(
+            api_keys={"OPENAI_API_KEY": api_keys["OPENAI_API_KEY"]},
             messages=messages,
             model=model,
             api_type=api_type,
             temperature=temperature,
             stream=stream,
-            api_keys={"OPENAI_API_KEY": api_keys["OPENAI_API_KEY"]}
+            thought=thought,
+            code_execution=code_execution,
         )
     elif api_type == 'OpenAI-Azure':
         return await create_openai_client(
+            api_keys={
+                "AZURE_API_KEY": api_keys["AZURE_API_KEY"],
+                "AZURE_API_BASE": api_keys["AZURE_API_BASE"]
+            },
             messages=messages,
             model=model,
             api_type=api_type,
             temperature=temperature,
             stream=stream,
-            api_keys={
-                "AZURE_API_KEY": api_keys["AZURE_API_KEY"],
-                "AZURE_API_BASE": api_keys["AZURE_API_BASE"]
-            }
+            thought=thought,
+            code_execution=code_execution,
         )
     elif api_type == 'OpenAI-GitHub':
         return await create_openai_client(
+            api_keys={"GITHUB_API_KEY": api_keys["GITHUB_API_KEY"]},
             messages=messages,
             model=model,
             api_type=api_type,
             temperature=temperature,
             stream=stream,
-            api_keys={"GITHUB_API_KEY": api_keys["GITHUB_API_KEY"]}
+            thought=thought,
+            code_execution=code_execution,
         )
     elif api_type == 'Grok':
         return await create_openai_client(
+            api_keys={"XAI_API_KEY": api_keys["XAI_API_KEY"]},
             messages=messages,
             model=model,
             api_type=api_type,
             temperature=temperature,
             stream=stream,
-            api_keys={"XAI_API_KEY": api_keys["XAI_API_KEY"]}
+            thought=thought,
+            code_execution=code_execution,
         )
     elif api_type == 'Gemini-Free':
         return await create_gemini_client(
+            api_key=api_keys["GEMINI_FREE_API_KEY"],
+            vertexai=False,
             messages=messages,
             model=model,
             temperature=temperature,
             stream=stream,
-            api_key=api_keys["GEMINI_FREE_API_KEY"],
-            vertexai=False,
+            thought=thought,
+            code_execution=code_execution,
         )
     elif api_type == 'Gemini-Paid':
         return await create_gemini_client(
+            api_key=api_keys["GEMINI_PAID_API_KEY"],
+            vertexai=False,
             messages=messages,
             model=model,
             temperature=temperature,
             stream=stream,
-            api_key=api_keys["GEMINI_PAID_API_KEY"],
-            vertexai=False,
+            thought=thought,
+            code_execution=code_execution,
         )
     elif api_type == 'Gemini-Vertex':
         return await create_gemini_client(
+            api_key=api_keys["GEMINI_VERTEX_API_KEY"],
             messages=messages,
             model=model,
             temperature=temperature,
             stream=stream,
-            api_key=api_keys["GEMINI_VERTEX_API_KEY"],
+            thought=thought,
+            code_execution=code_execution,
             vertexai=True,
         )
     elif api_type == 'Claude':
         return await create_claude_client(
+            api_key=api_keys["ANTHROPIC_API_KEY"],
             messages=messages,
             model=model,
             temperature=temperature,
             stream=stream,
-            api_key=api_keys["ANTHROPIC_API_KEY"]
+            thought=thought,
+            code_execution=code_execution,
         )
     else:
         raise HTTPException(status_code=400, detail="Invalid API type")

{llm_bridge-1.11.9 → llm_bridge-1.12.0}/llm_bridge/logic/chat_generate/model_client_factory/claude_client_factory.py

@@ -11,11 +11,13 @@ from llm_bridge.type.message import Message
 
 
 async def create_claude_client(
+        api_key: str,
         messages: list[Message],
         model: str,
         temperature: float,
         stream: bool,
-        api_key: str,
+        thought: bool,
+        code_execution: bool,
 ):
     client = anthropic.AsyncAnthropic(
         api_key=api_key,
@@ -38,11 +40,13 @@ async def create_claude_client(
         32_000,  # Max output: Claude 4.5 64K; Claude 4.1 32K
         200_000 - input_tokens,  # Context window: Claude Sonnet 4.5 beta: 1M; otherwise 200K
     )
-    thinking = ThinkingConfigEnabledParam(
-        type="enabled",
-        budget_tokens=max(1024, max_tokens // 2), # Minimum budget tokens: 1024
-    )
-    temperature = 1
+    thinking = None
+    if thought:
+        thinking = ThinkingConfigEnabledParam(
+            type="enabled",
+            budget_tokens=max(1024, max_tokens // 2),  # Minimum budget tokens: 1024
+        )
+        temperature = 1
     betas: list[AnthropicBetaParam] = [
         "context-1m-2025-08-07",
         "output-128k-2025-02-19",
@@ -53,11 +57,14 @@ async def create_claude_client(
             type="web_search_20250305",
             name="web_search",
         ),
-        BetaCodeExecutionTool20250825Param(
-            type="code_execution_20250825",
-            name="code_execution",
-        )
     ]
+    if code_execution:
+        tools.append(
+            BetaCodeExecutionTool20250825Param(
+                type="code_execution_20250825",
+                name="code_execution",
+            )
+        )
 
     if stream:
         return StreamClaudeClient(
@@ -67,10 +74,10 @@ async def create_claude_client(
             system=system,
             client=client,
             max_tokens=max_tokens,
-            thinking=thinking,
             betas=betas,
             input_tokens=input_tokens,
             tools=tools,
+            thinking=thinking,
         )
     else:
         return NonStreamClaudeClient(
@@ -80,8 +87,8 @@ async def create_claude_client(
             system=system,
             client=client,
             max_tokens=max_tokens,
-            thinking=thinking,
             betas=betas,
             input_tokens=input_tokens,
             tools=tools,
+            thinking=thinking,
         )

{llm_bridge-1.11.9 → llm_bridge-1.12.0}/llm_bridge/logic/chat_generate/model_client_factory/gemini_client_factory.py

@@ -1,6 +1,5 @@
 from google import genai
 from google.genai import types
-from google.genai._api_client import HttpOptions
 from google.genai.types import Modality
 
 from llm_bridge.client.implementations.gemini.non_stream_gemini_client import NonStreamGeminiClient
@@ -11,25 +10,26 @@ from llm_bridge.type.message import Message
 
 
 async def create_gemini_client(
+        api_key: str,
+        vertexai: bool,
         messages: list[Message],
         model: str,
         temperature: float,
         stream: bool,
-        api_key: str,
-        vertexai: bool,
+        thought: bool,
+        code_execution: bool,
 ):
     client = genai.Client(
         vertexai=vertexai,
         api_key=api_key,
     )
 
-    system_instruction = None
+    system_instruction = extract_system_messages(messages) or " "
     tools = []
     thinking_config = None
     response_modalities = [Modality.TEXT]
 
-    system_instruction = extract_system_messages(messages) or " "
-    if "image" not in model and not vertexai:
+    if "image" not in model:
         tools.append(
             types.Tool(
                 google_search=types.GoogleSearch()
@@ -40,27 +40,18 @@ async def create_gemini_client(
                 url_context=types.UrlContext()
             )
         )
-        tools.append(
-            types.Tool(
-                code_execution=types.ToolCodeExecution()
-            )
-        )
-    if "image" not in model and vertexai:
-        tools.append(
-            types.Tool(
-                google_search=types.GoogleSearch()
+        if thought:
+            thinking_config = types.ThinkingConfig(
+                include_thoughts=True,
+                thinking_budget=-1,
             )
-        )
-        tools.append(
-            types.Tool(
-                url_context=types.UrlContext()
-            )
-        )
-    if "image" not in model:
-        thinking_config = types.ThinkingConfig(
-            include_thoughts=True,
-            thinking_budget=-1,
-        )
+        if not vertexai:
+            if code_execution:
+                tools.append(
+                    types.Tool(
+                        code_execution=types.ToolCodeExecution()
+                    )
+                )
     if "image" in model:
         response_modalities = [Modality.TEXT, Modality.IMAGE]
 

{llm_bridge-1.11.9 → llm_bridge-1.12.0}/llm_bridge/logic/chat_generate/model_client_factory/openai_client_factory.py

@@ -17,12 +17,14 @@ from llm_bridge.type.message import Message
 
 
 async def create_openai_client(
+        api_keys: dict,
         messages: list[Message],
         model: str,
         api_type: str,
         temperature: float,
         stream: bool,
-        api_keys: dict
+        thought: bool,
+        code_execution: bool,
 ):
     if api_type == "OpenAI":
         client = openai.AsyncOpenAI(
@@ -47,7 +49,7 @@ async def create_openai_client(
     else:
         raise HTTPException(status_code=500, detail="API Type not matched")
 
-    if api_type in ("OpenAI", "OpenAI-Azure", "Grok"):
+    if api_type in ("OpenAI", "OpenAI-Azure"):
         use_responses_api = True
     else:
         use_responses_api = False
@@ -61,12 +63,13 @@ async def create_openai_client(
     reasoning = None
 
     if model not in ["gpt-5-chat-latest", "gpt-5-pro"]:
-        tools.append(
-            CodeInterpreter(
-                type="code_interpreter",
-                container=CodeInterpreterContainerCodeInterpreterToolAuto(type="auto")
+        if code_execution:
+            tools.append(
+                CodeInterpreter(
+                    type="code_interpreter",
+                    container=CodeInterpreterContainerCodeInterpreterToolAuto(type="auto")
+                )
             )
-        )
     if model not in ["gpt-5-chat-latest"]:
         tools.append(
             WebSearchToolParam(
@@ -77,10 +80,11 @@ async def create_openai_client(
     if re.match(r"gpt-5.*", model) and model != "gpt-5-chat-latest":
         temperature = 1
     if re.match(r"gpt-5.*", model) and model != "gpt-5-chat-latest":
-        reasoning = Reasoning(
-            effort="high",
-            summary="auto",
-        )
+        if thought:
+            reasoning = Reasoning(
+                effort="high",
+                summary="auto",
+            )
         tools.append(
             ImageGeneration(
                 type="image_generation",