auto-coder 0.1.231__tar.gz → 0.1.232__tar.gz

{auto_coder-0.1.231 → auto_coder-0.1.232}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: auto-coder
-Version: 0.1.231
+Version: 0.1.232
 Summary: AutoCoder: AutoCoder
 Author: allwefantasy
 Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence

{auto_coder-0.1.231 → auto_coder-0.1.232}/src/auto_coder.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: auto-coder
-Version: 0.1.231
+Version: 0.1.232
 Summary: AutoCoder: AutoCoder
 Author: allwefantasy
 Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence

{auto_coder-0.1.231 → auto_coder-0.1.232}/src/autocoder/auto_coder.py

@@ -1272,10 +1272,13 @@ def main(input_args: Optional[List[str]] = None):
                     )
                 )
                 v = [[response.result,None]]
-            else:                
-                v = chat_llm.stream_chat_oai(
-                    conversations=loaded_conversations, delta_mode=True
-                )                
+            else:
+                from autocoder.common.utils_code_auto_generate import stream_chat_with_continue
+                v = stream_chat_with_continue(
+                    llm=chat_llm,
+                    conversations=loaded_conversations,
+                    llm_config={}
+                )              
             
             assistant_response, last_meta = stream_out(
                     v, 

auto_coder-0.1.232/src/autocoder/common/utils_code_auto_generate.py

@@ -0,0 +1,90 @@
+from byzerllm import ByzerLLM,SimpleByzerLLM
+from typing import Generator, List, Any, Union
+from pydantic import BaseModel
+from loguru import logger
+
+class ChatWithContinueResult(BaseModel):
+    content: str
+    input_tokens_count: int
+    generated_tokens_count: int
+    
+
+def chat_with_continue(llm: Union[ByzerLLM,SimpleByzerLLM], conversations: List[dict], llm_config: dict) -> ChatWithContinueResult:
+    final_result = ChatWithContinueResult(content="", input_tokens_count=0, generated_tokens_count=0)
+    v = llm.chat_oai(
+        conversations=conversations, llm_config=llm_config)
+                
+    single_result = v[0].output
+    metadata = v[0].metadata
+
+    final_result.input_tokens_count += metadata.get("input_tokens_count", 0)
+    final_result.generated_tokens_count += metadata.get("generated_tokens_count", 0)
+
+    temp_conversations = conversations + \
+        [{"role": "assistant", "content": single_result}]
+    
+    count = 1
+    while (metadata.get("finish_reason", "stop") == "length" and count < 6):        
+        v = llm.chat_oai(
+            conversations=temp_conversations, llm_config={**llm_config, "gen.response_prefix": True})
+        metadata = v[0].metadata
+        single_result += v[0].output
+        final_result.input_tokens_count += metadata.get("input_tokens_count", 0)
+        final_result.generated_tokens_count += metadata.get("generated_tokens_count", 0)
+        count += 1
+    
+    # if count >= 2:
+    #   logger.info(f"The code generation is exceed the max length, continue to generate the code {count -1 } times")
+    final_result.content = single_result    
+    return final_result
+
+def stream_chat_with_continue(
+    llm: Union[ByzerLLM, SimpleByzerLLM], 
+    conversations: List[dict], 
+    llm_config: dict
+) -> Generator[Any, None, None]:
+    """
+    流式处理并继续生成内容，直到完成。
+    
+    Args:
+        llm (Union[ByzerLLM, SimpleByzerLLM]): LLM实例
+        conversations (List[dict]): 对话历史
+        llm_config (dict): LLM配置参数
+        
+
+    """
+    
+    count = 0
+    temp_conversations = conversations
+    current_metadata = None
+    
+    while True:
+        # 使用流式接口获取生成内容
+        stream_generator = llm.stream_chat_oai(
+            conversations=temp_conversations,
+            delta_mode=True,
+            llm_config={**llm_config, "gen.response_prefix": True if count > 0 else False}
+        )
+        
+        current_content = ""        
+        for res in stream_generator:
+            content = res[0]
+            current_content += content 
+            if current_metadata is None:
+                current_metadata = res[1]            
+            else:
+                current_metadata.generated_tokens_count += res[1].generated_tokens_count
+                current_metadata.input_tokens_count += res[1].input_tokens_count
+                current_metadata.finish_reason = res[1].finish_reason                
+            
+            # Yield 当前的 StreamChatWithContinueResult
+            yield (content,current_metadata)
+        
+        # 更新对话历史
+        temp_conversations.append({"role": "assistant", "content": current_content})
+        
+        # 检查是否需要继续生成
+        if current_metadata.finish_reason != "length" or count >= 5:
+            break
+        
+        count += 1

{auto_coder-0.1.231 → auto_coder-0.1.232}/src/autocoder/utils/auto_coder_utils/chat_stream_out.py

@@ -39,7 +39,7 @@ def stream_out(
             console=console
         ) as live:
             for res in stream_generator:
-                last_meta = res[1]
+                last_meta = res[1]                
                 content = res[0]
                 assistant_response += content
                 

auto_coder-0.1.232/src/autocoder/version.py

@@ -0,0 +1 @@
+__version__ = "0.1.232"

auto_coder-0.1.231/src/autocoder/common/utils_code_auto_generate.py

@@ -1,38 +0,0 @@
-from byzerllm import ByzerLLM
-from typing import List,Any,Union
-from pydantic import BaseModel
-from loguru import logger
-class ChatWithContinueResult(BaseModel):
-    content: str
-    input_tokens_count: int
-    generated_tokens_count: int
-    
-
-def chat_with_continue(llm: ByzerLLM, conversations: List[dict], llm_config: dict) -> ChatWithContinueResult:
-    final_result = ChatWithContinueResult(content="", input_tokens_count=0, generated_tokens_count=0)
-    v = llm.chat_oai(
-        conversations=conversations, llm_config=llm_config)
-                
-    single_result = v[0].output
-    metadata = v[0].metadata
-
-    final_result.input_tokens_count += metadata.get("input_tokens_count", 0)
-    final_result.generated_tokens_count += metadata.get("generated_tokens_count", 0)
-
-    temp_conversations = conversations + \
-        [{"role": "assistant", "content": single_result}]
-    
-    count = 1
-    while (metadata.get("finish_reason", "stop") == "length" and count < 6):        
-        v = llm.chat_oai(
-            conversations=temp_conversations, llm_config={**llm_config, "gen.response_prefix": True})
-        metadata = v[0].metadata
-        single_result += v[0].output
-        final_result.input_tokens_count += metadata.get("input_tokens_count", 0)
-        final_result.generated_tokens_count += metadata.get("generated_tokens_count", 0)
-        count += 1
-    
-    # if count >= 2:
-    #   logger.info(f"The code generation is exceed the max length, continue to generate the code {count -1 } times")
-    final_result.content = single_result    
-    return final_result

auto_coder-0.1.231/src/autocoder/version.py

@@ -1 +0,0 @@
-__version__ = "0.1.231"