LLM-Bridge 1.7.21__tar.gz → 1.8.0a0__tar.gz

{llm_bridge-1.7.21 → llm_bridge-1.8.0a0/LLM_Bridge.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: LLM-Bridge
-Version: 1.7.21
+Version: 1.8.0a0
 Summary: A Bridge for LLMs
 Author-email: windsnow1025 <windsnow1025@gmail.com>
 License-Expression: MIT

{llm_bridge-1.7.21/LLM_Bridge.egg-info → llm_bridge-1.8.0a0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: LLM-Bridge
-Version: 1.7.21
+Version: 1.8.0a0
 Summary: A Bridge for LLMs
 Author-email: windsnow1025 <windsnow1025@gmail.com>
 License-Expression: MIT

{llm_bridge-1.7.21 → llm_bridge-1.8.0a0}/llm_bridge/client/implementations/claude/claude_stream_response_handler.py

@@ -1,14 +1,10 @@
 from anthropic import BetaMessageStreamEvent, AsyncAnthropic
 
 from llm_bridge.client.implementations.claude.claude_token_counter import count_claude_output_tokens
-from llm_bridge.client.implementations.printing_status import PrintingStatus
 from llm_bridge.type.chat_response import ChatResponse
 
 
 class ClaudeStreamResponseHandler:
-    def __init__(self):
-        self.printing_status = None
-
     async def process_claude_stream_response(
             self,
             event: BetaMessageStreamEvent,
@@ -17,21 +13,21 @@ class ClaudeStreamResponseHandler:
             model: str,
     ) -> ChatResponse:
         text = ""
+        thought = ""
+
         if event.type == "content_block_delta":
             if event.delta.type == "thinking_delta":
-                if not self.printing_status:
-                    text += "# Model Thought:\n\n"
-                    self.printing_status = PrintingStatus.Thought
-                text += event.delta.thinking
+                thought += event.delta.thinking
             elif event.delta.type == "text_delta":
-                if self.printing_status == PrintingStatus.Thought:
-                    text += "\n\n# Model Response:\n\n"
-                    self.printing_status = PrintingStatus.Response
                 text += event.delta.text
         elif event.type == "citation":
             citation = event.citation
             text += f"([{citation.title}]({citation.url})) "
-        chat_response = ChatResponse(text=text)
+
+        chat_response = ChatResponse(
+            text=text,
+            thought=thought,
+        )
         output_tokens = await count_claude_output_tokens(
             client=client,
             model=model,
@@ -39,6 +35,7 @@ class ClaudeStreamResponseHandler:
         )
         return ChatResponse(
             text=text,
+            thought=thought,
             input_tokens=input_tokens,
             output_tokens=output_tokens,
         )

{llm_bridge-1.7.21 → llm_bridge-1.8.0a0}/llm_bridge/client/implementations/claude/non_stream_claude_client.py

@@ -19,20 +19,22 @@ async def process_claude_non_stream_response(
         model: str,
 ) -> ChatResponse:
     text = ""
+    thought = ""
 
     for content in message.content:
         if content.type == "thinking":
-            text += "# Model Thought:\n\n"
-            text += content.thinking
+            thought += content.thinking
         if content.type == "text":
-            text += "\n\n# Model Response:\n\n"
             text += content.text
-            # Unable to test since streaming Claude is currently not allowed
+            # Unable to test: non-streaming Claude is currently not allowed
             if citations := content.citations:
                 for citation in citations:
                     text += f"([{citation.title}]({citation.url})) "
 
-    chat_response = ChatResponse(text=text)
+    chat_response = ChatResponse(
+        text=text,
+        thought=thought,
+    )
     output_tokens = await count_claude_output_tokens(
         client=client,
         model=model,
@@ -40,6 +42,7 @@ async def process_claude_non_stream_response(
     )
     return ChatResponse(
         text=text,
+        thought=thought,
         input_tokens=input_tokens,
         output_tokens=output_tokens,
     )

{llm_bridge-1.7.21 → llm_bridge-1.8.0a0}/llm_bridge/client/implementations/claude/stream_claude_client.py

@@ -3,13 +3,9 @@ import re
 from typing import AsyncGenerator
 
 import httpx
-from anthropic import AsyncAnthropic, BetaMessageStreamEvent
-from anthropic.types.beta import BetaMessage
 from fastapi import HTTPException
 
 from llm_bridge.client.implementations.claude.claude_stream_response_handler import ClaudeStreamResponseHandler
-from llm_bridge.client.implementations.claude.claude_token_counter import count_claude_input_tokens, \
-    count_claude_output_tokens
 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

{llm_bridge-1.7.21 → llm_bridge-1.8.0a0}/llm_bridge/client/implementations/gemini/gemini_response_handler.py

@@ -1,4 +1,5 @@
 import base64
+from typing import Optional
 
 from google.genai import types
 
@@ -9,32 +10,33 @@ from llm_bridge.type.chat_response import Citation, ChatResponse
 
 class GeminiResponseHandler:
     def __init__(self):
-        self.printing_status = None
-        self.prev_output_tokens = 0
-        self.prev_printing_status = None
+        self.printing_status: Optional[PrintingStatus] = None
+        self.prev_output_tokens: int = 0
+        self.prev_printing_status: Optional[PrintingStatus] = None
 
     async def process_gemini_response(
             self,
             response: types.GenerateContentResponse,
     ) -> ChatResponse:
         text = ""
+        thought = ""
         display = None
         image_base64 = None
         citations = extract_citations(response)
         input_tokens, stage_output_tokens = await count_gemini_tokens(response)
 
+        printing_status = None
         if candidates := response.candidates:
             if candidates[0].content.parts:
                 for part in response.candidates[0].content.parts:
                     # Thought Output
                     if part.text:
-                        if part.thought and not self.printing_status:
-                            text += "# Model Thought:\n\n"
-                            self.printing_status = PrintingStatus.Thought
-                        elif not part.thought and self.printing_status == PrintingStatus.Thought:
-                            text += f"\n\n# Model Response:\n\n"
-                            self.printing_status = PrintingStatus.Response
-                        text += part.text
+                        if part.thought:
+                            printing_status = PrintingStatus.Thought
+                            thought += part.text
+                        elif not part.thought:
+                            printing_status = PrintingStatus.Response
+                            text += part.text
                     # Image Output
                     elif part.inline_data:
                         image_base64 = base64.b64encode(part.inline_data.data).decode('utf-8')
@@ -50,16 +52,17 @@ class GeminiResponseHandler:
                         if chunk.web:
                             text += f"{i}. [{chunk.web.title}]({chunk.web.uri})\n"
 
-        if self.printing_status == self.prev_printing_status:
+        if printing_status == self.prev_printing_status and printing_status == PrintingStatus.Response:
             output_tokens = stage_output_tokens - self.prev_output_tokens
         else:
             output_tokens = stage_output_tokens
 
         self.prev_output_tokens = stage_output_tokens
-        self.prev_printing_status = self.printing_status
+        self.prev_printing_status = printing_status
 
         return ChatResponse(
             text=text,
+            thought=thought,
             image=image_base64,
             display=display,
             citations=citations,

{llm_bridge-1.7.21 → llm_bridge-1.8.0a0}/llm_bridge/type/chat_response.py

@@ -11,6 +11,7 @@ class Citation:
 @dataclass
 class ChatResponse:
     text: Optional[str] = None
+    thought: Optional[str] = None
     image: Optional[str] = None
     display: Optional[str] = None
     citations: Optional[list[Citation]] = None

{llm_bridge-1.7.21 → llm_bridge-1.8.0a0}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "LLM-Bridge"
-version = "1.7.21"
+version = "1.8.0-alpha.0"
 authors = [
     {name = "windsnow1025", email = "windsnow1025@gmail.com"}
 ]