auto-coder 0.1.383__py3-none-any.whl → 0.1.385__py3-none-any.whl

{auto_coder-0.1.383.dist-info → auto_coder-0.1.385.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: auto-coder
-Version: 0.1.383
+Version: 0.1.385
 Summary: AutoCoder: AutoCoder
 Author: allwefantasy
 Classifier: Programming Language :: Python :: 3.10

{auto_coder-0.1.383.dist-info → auto_coder-0.1.385.dist-info}/RECORD

@@ -1,7 +1,7 @@
 autocoder/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 autocoder/auto_coder.py,sha256=7602L3tG0JErNxh8vkLAmGUgv2c-DGPzPCkmWIQt9bs,69757
 autocoder/auto_coder_rag.py,sha256=tRAKfo3jIhcaQKN_3g7DZRKtDJSZXJxMRdT6Zz8W9nw,41173
-autocoder/auto_coder_runner.py,sha256=IDaCD__NJ7OAStPLI5EfYMq0etiMU-ICTV7xkLD3koo,115220
+autocoder/auto_coder_runner.py,sha256=NhT1fVdV4NnV2SwpXl7G-tTKgduA0D-0lK_p2R6DWew,115569
 autocoder/auto_coder_server.py,sha256=bLORGEclcVdbBVfM140JCI8WtdrU0jbgqdJIVVupiEU,20578
 autocoder/benchmark.py,sha256=Ypomkdzd1T3GE6dRICY3Hj547dZ6_inqJbBJIp5QMco,4423
 autocoder/chat_auto_coder.py,sha256=vNQwbYkdqeMl07Vx8z6x-kSPkHKn9AT3sSkYMTJiWtc,26655
@@ -11,7 +11,7 @@ autocoder/command_parser.py,sha256=fx1g9E6GaM273lGTcJqaFQ-hoksS_Ik2glBMnVltPCE,1
 autocoder/lang.py,sha256=PFtATuOhHRnfpqHQkXr6p4C893JvpsgwTMif3l-GEi0,14321
 autocoder/models.py,sha256=pD5u6gcMKRwWaLxeVin18g25k-ERyeHOFsRpOgO_Ae0,13788
 autocoder/run_context.py,sha256=IUfSO6_gp2Wt1blFWAmOpN0b0nDrTTk4LmtCYUBIoro,1643
-autocoder/version.py,sha256=gj191QDAwvI_4Wfq02TLyc83ZVdeBHTkoS0W9_2_bzo,25
+autocoder/version.py,sha256=iaY60sgwWc9wED3HJ9j7D-Lk6OSNJoA1OS8XwdbqWtI,25
 autocoder/agent/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 autocoder/agent/agentic_filter.py,sha256=zlInIRhawKIYTJjCiJBWqPCOV5UtMbh5VnvszfTy2vo,39824
 autocoder/agent/auto_demand_organizer.py,sha256=URAq0gSEiHeV_W4zwhOI_83kHz0Ryfj1gcfh5jwCv_w,6501
@@ -270,7 +270,7 @@ autocoder/rag/doc_filter.py,sha256=UduVO2mlrngwJICrefjDJTYfdmQ4GcRXrfWDQ7xXksk,1
 autocoder/rag/document_retriever.py,sha256=rFwbAuHTvEFJq16HQNlmRLyJp2ddn2RNFslw_ncU7NI,8847
 autocoder/rag/lang.py,sha256=HvcMeu6jReEJOGxyLMn4rwBoD-myFwmykS3VLceBJLs,3364
 autocoder/rag/llm_wrapper.py,sha256=LsNv8maCnvazyXjjtkO9aN3OT7Br20V1ilHV8Lt45Os,4245
-autocoder/rag/long_context_rag.py,sha256=s00w7Ep6nWjRS0Xy8m2qUCvAaH3CgqLLvh6N5d5ssII,50029
+autocoder/rag/long_context_rag.py,sha256=gL35Y52lCJB3Lf0JgU-kDPCHEf6cR5QgA8uUontyKbo,52822
 autocoder/rag/qa_conversation_strategy.py,sha256=4CiMK88apKbJ2YM4HHq1KGpr5jUkTh0_m_aCyt-JYgc,10568
 autocoder/rag/rag_config.py,sha256=8LwFcTd8OJWWwi1_WY4IzjqgtT6RyE2j4PjxS5cCTDE,802
 autocoder/rag/rag_entry.py,sha256=QOdUX_nd1Qak2NyOW0CYcLRDB26AZ6MeByHJaMMGgqs,2316
@@ -283,7 +283,7 @@ autocoder/rag/token_checker.py,sha256=jc76x6KWmvVxds6W8juZfQGaoErudc2HenG3sNQfSL
 autocoder/rag/token_counter.py,sha256=C-Lwc4oIjJpZDEqp9WLHGOe6hb4yhrdJpMtkrtp_1qc,2125
 autocoder/rag/token_limiter.py,sha256=3VgJF4may3ESyATmBIiOe05oc3VsidJcJTJ5EhoSvH8,18854
 autocoder/rag/token_limiter_utils.py,sha256=FATNEXBnFJy8IK3PWNt1pspIv8wuTgy3F_ACNvqoc4I,404
-autocoder/rag/types.py,sha256=G6A3P5YN-VBUgUX2TFSWRCKRDJq42fUxVIf46DL8fNI,3280
+autocoder/rag/types.py,sha256=XUg47lb9DpJoIM4tcokmRp8bk4_mac0MrNd1GY1EVdg,3383
 autocoder/rag/utils.py,sha256=FPK3Vvk9X9tUuOu4_LctZN5WnRVuEjFiffRtE-pHn0s,6318
 autocoder/rag/variable_holder.py,sha256=PFvBjFcR7-fNDD4Vcsc8CpH2Te057vcpwJMxtrfUgKI,75
 autocoder/rag/cache/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
@@ -336,9 +336,9 @@ autocoder/utils/types.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 autocoder/utils/auto_coder_utils/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 autocoder/utils/auto_coder_utils/chat_stream_out.py,sha256=t902pKxQ5xM7zgIHiAOsTPLwxhE6VuvXAqPy751S7fg,14096
 autocoder/utils/chat_auto_coder_utils/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-auto_coder-0.1.383.dist-info/LICENSE,sha256=HrhfyXIkWY2tGFK11kg7vPCqhgh5DcxleloqdhrpyMY,11558
-auto_coder-0.1.383.dist-info/METADATA,sha256=EvB2XWtEz7GHmHMMWAOC5eGvDLoh3eDI9z0ePFbgCkM,2796
-auto_coder-0.1.383.dist-info/WHEEL,sha256=GV9aMThwP_4oNCtvEC2ec3qUYutgWeAzklro_0m4WJQ,91
-auto_coder-0.1.383.dist-info/entry_points.txt,sha256=0nzHtHH4pNcM7xq4EBA2toS28Qelrvcbrr59GqD_0Ak,350
-auto_coder-0.1.383.dist-info/top_level.txt,sha256=Jqc0_uJSw2GwoFQAa9iJxYns-2mWla-9ok_Y3Gcznjk,10
-auto_coder-0.1.383.dist-info/RECORD,,
+auto_coder-0.1.385.dist-info/LICENSE,sha256=HrhfyXIkWY2tGFK11kg7vPCqhgh5DcxleloqdhrpyMY,11558
+auto_coder-0.1.385.dist-info/METADATA,sha256=CBTR-3ZwyAp_dQkgpXTl3XVwwwUjkQjz6zdkVJnc1WI,2796
+auto_coder-0.1.385.dist-info/WHEEL,sha256=GV9aMThwP_4oNCtvEC2ec3qUYutgWeAzklro_0m4WJQ,91
+auto_coder-0.1.385.dist-info/entry_points.txt,sha256=0nzHtHH4pNcM7xq4EBA2toS28Qelrvcbrr59GqD_0Ak,350
+auto_coder-0.1.385.dist-info/top_level.txt,sha256=Jqc0_uJSw2GwoFQAa9iJxYns-2mWla-9ok_Y3Gcznjk,10
+auto_coder-0.1.385.dist-info/RECORD,,

autocoder/auto_coder_runner.py

@@ -657,6 +657,16 @@ def save_memory():
     load_memory()
 
 
+def save_memory_with_new_memory(new_memory):
+    memory_path = os.path.join(base_persist_dir, "memory.json")
+    lock_path = memory_path + ".lock"
+    
+    with FileLock(lock_path, timeout=30):
+        with open(memory_path, "w", encoding="utf-8") as f:
+            json.dump(new_memory, f, indent=2, ensure_ascii=False)
+    load_memory()        
+
+
 def load_memory():    
     global memory
     memory_path = os.path.join(base_persist_dir, "memory.json")

autocoder/rag/long_context_rag.py

@@ -539,7 +539,7 @@ class LongContextRAG:
                     generated_tokens_count=rag_stat.recall_stat.total_generated_tokens + rag_stat.chunk_stat.total_generated_tokens,
                     reasoning_content=get_message_with_format_and_newline(
                         "context_docs_names",
-                        context_docs_names=",".join(context)
+                        context_docs_names="*****"
                     )
                 ))
                 
@@ -676,6 +676,8 @@ class LongContextRAG:
     def _process_document_retrieval(self, conversations, 
                                     query, rag_stat):
         """第一阶段：文档召回和过滤"""
+        recall_start_time = time.time()  # 记录召回阶段开始时间
+        
         yield ("", SingleOutputMeta(
             input_tokens_count=0,
             generated_tokens_count=0,
@@ -716,6 +718,7 @@ class LongContextRAG:
         rag_stat.recall_stat.total_input_tokens += sum(doc_filter_result.input_tokens_counts)
         rag_stat.recall_stat.total_generated_tokens += sum(doc_filter_result.generated_tokens_counts)
         rag_stat.recall_stat.model_name = doc_filter_result.model_name
+        rag_stat.recall_stat.duration = time.time() - recall_start_time  # 记录召回阶段耗时
 
         relevant_docs = doc_filter_result.docs
         
@@ -724,7 +727,7 @@ class LongContextRAG:
             generated_tokens_count=rag_stat.recall_stat.total_generated_tokens,
             reasoning_content=get_message_with_format_and_newline(
                 "rag_docs_filter_result",
-                filter_time=0,  # 这里实际应该计算时间，但由于重构，我们需要在外部计算
+                filter_time=rag_stat.recall_stat.duration,  # 使用实际耗时
                 docs_num=len(relevant_docs),
                 input_tokens=rag_stat.recall_stat.total_input_tokens,
                 output_tokens=rag_stat.recall_stat.total_generated_tokens,
@@ -743,6 +746,8 @@ class LongContextRAG:
 
     def _process_document_chunking(self, relevant_docs, conversations, rag_stat, filter_time):
         """第二阶段：文档分块与重排序"""
+        chunk_start_time = time.time()  # 记录分块阶段开始时间
+        
         yield ("", SingleOutputMeta(
             generated_tokens_count=0,
             reasoning_content=get_message_with_format_and_newline(
@@ -785,6 +790,8 @@ class LongContextRAG:
             # 如果没有tokenizer，直接限制文档数量
             final_relevant_docs = relevant_docs[: self.args.index_filter_file_num]
         
+        rag_stat.chunk_stat.duration = time.time() - chunk_start_time  # 记录分块阶段耗时
+        
         # 输出分块结果统计
         yield ("", SingleOutputMeta(
             generated_tokens_count=rag_stat.chunk_stat.total_generated_tokens + rag_stat.recall_stat.total_generated_tokens,
@@ -819,6 +826,7 @@ class LongContextRAG:
                                llm_config={}, 
                                extra_request_params={}):
         """第三阶段：大模型问答生成"""
+        answer_start_time = time.time()  # 记录答案生成阶段开始时间
         
         # 使用LLMComputeEngine增强处理（如果可用）
         if LLMComputeEngine is not None and not self.args.disable_inference_enhance:
@@ -853,6 +861,7 @@ class LongContextRAG:
                         rag_stat.chunk_stat.total_generated_tokens + \
                         rag_stat.answer_stat.total_generated_tokens
                 yield chunk
+            rag_stat.answer_stat.duration = time.time() - answer_start_time  # 记录答案生成阶段耗时
         else:
             # 常规QA处理路径
             qa_strategy = get_qa_strategy(self.args)
@@ -892,6 +901,7 @@ class LongContextRAG:
                         rag_stat.chunk_stat.total_generated_tokens + \
                         rag_stat.answer_stat.total_generated_tokens
                 yield chunk
+            rag_stat.answer_stat.duration = time.time() - answer_start_time  # 记录答案生成阶段耗时
 
     def _print_rag_stats(self, rag_stat: RAGStat, conversations: Optional[List[Dict[str, str]]] = None) -> None:
         """打印RAG执行的详细统计信息"""
@@ -906,6 +916,17 @@ class LongContextRAG:
             rag_stat.answer_stat.total_generated_tokens
         )
         total_tokens = total_input_tokens + total_generated_tokens
+        
+        # 计算总耗时
+        total_duration = (
+            rag_stat.recall_stat.duration +
+            rag_stat.chunk_stat.duration +
+            rag_stat.answer_stat.duration
+        )
+        
+        # 添加其他阶段的耗时（如果存在）
+        if rag_stat.other_stats:
+            total_duration += sum(other_stat.duration for other_stat in rag_stat.other_stats)
 
         # 避免除以零错误
         if total_tokens == 0:
@@ -926,6 +947,20 @@ class LongContextRAG:
                                 other_stat.total_generated_tokens) / total_tokens * 100
                 other_percents.append(other_percent)
         
+        # 计算耗时分布百分比
+        if total_duration == 0:
+            recall_duration_percent = chunk_duration_percent = answer_duration_percent = 0
+        else:
+            recall_duration_percent = rag_stat.recall_stat.duration / total_duration * 100
+            chunk_duration_percent = rag_stat.chunk_stat.duration / total_duration * 100
+            answer_duration_percent = rag_stat.answer_stat.duration / total_duration * 100
+            
+        # 计算其他阶段的耗时占比
+        other_duration_percents = []
+        if total_duration > 0 and rag_stat.other_stats:
+            for other_stat in rag_stat.other_stats:
+                other_duration_percents.append(other_stat.duration / total_duration * 100)
+        
         # 计算成本分布百分比
         if rag_stat.cost == 0:
             recall_cost_percent = chunk_cost_percent = answer_cost_percent = 0
@@ -957,6 +992,7 @@ class LongContextRAG:
             f"  * 输入令牌总数: {total_input_tokens}\n"
             f"  * 生成令牌总数: {total_generated_tokens}\n"
             f"  * 总成本: {rag_stat.cost:.6f}\n"
+            f"  * 总耗时: {total_duration:.2f} 秒\n"
             f"\n"
             f"阶段统计:\n"
             f"  1. 文档检索阶段:\n"
@@ -965,6 +1001,7 @@ class LongContextRAG:
             f"     - 生成令牌: {rag_stat.recall_stat.total_generated_tokens}\n"
             f"     - 阶段总计: {rag_stat.recall_stat.total_input_tokens + rag_stat.recall_stat.total_generated_tokens}\n"
             f"     - 阶段成本: {rag_stat.recall_stat.cost:.6f}\n"
+            f"     - 阶段耗时: {rag_stat.recall_stat.duration:.2f} 秒\n"
             f"\n"
             f"  2. 文档分块阶段:\n"
             f"     - 模型: {rag_stat.chunk_stat.model_name}\n"
@@ -972,6 +1009,7 @@ class LongContextRAG:
             f"     - 生成令牌: {rag_stat.chunk_stat.total_generated_tokens}\n"
             f"     - 阶段总计: {rag_stat.chunk_stat.total_input_tokens + rag_stat.chunk_stat.total_generated_tokens}\n"
             f"     - 阶段成本: {rag_stat.chunk_stat.cost:.6f}\n"
+            f"     - 阶段耗时: {rag_stat.chunk_stat.duration:.2f} 秒\n"
             f"\n"
             f"  3. 答案生成阶段:\n"
             f"     - 模型: {rag_stat.answer_stat.model_name}\n"
@@ -979,6 +1017,7 @@ class LongContextRAG:
             f"     - 生成令牌: {rag_stat.answer_stat.total_generated_tokens}\n"
             f"     - 阶段总计: {rag_stat.answer_stat.total_input_tokens + rag_stat.answer_stat.total_generated_tokens}\n"
             f"     - 阶段成本: {rag_stat.answer_stat.cost:.6f}\n"
+            f"     - 阶段耗时: {rag_stat.answer_stat.duration:.2f} 秒\n"
             f"\n"
         )
         
@@ -992,6 +1031,7 @@ class LongContextRAG:
                     f"     - 生成令牌: {other_stat.total_generated_tokens}\n"
                     f"     - 阶段总计: {other_stat.total_input_tokens + other_stat.total_generated_tokens}\n"
                     f"     - 阶段成本: {other_stat.cost:.6f}\n"
+                    f"     - 阶段耗时: {other_stat.duration:.2f} 秒\n"
                     f"\n"
                 )
         
@@ -1009,6 +1049,21 @@ class LongContextRAG:
                 if other_percent > 0:
                     stats_str += f"  - 其他阶段 {i+1}: {other_percent:.1f}%\n"
         
+        # 添加耗时分布百分比
+        stats_str += (
+            f"\n"
+            f"耗时分布百分比:\n"
+            f"  - 文档检索: {recall_duration_percent:.1f}%\n"
+            f"  - 文档分块: {chunk_duration_percent:.1f}%\n"
+            f"  - 答案生成: {answer_duration_percent:.1f}%\n"
+        )
+        
+        # 如果存在 other_stats，添加其耗时占比
+        if rag_stat.other_stats:
+            for i, other_duration_percent in enumerate(other_duration_percents):
+                if other_duration_percent > 0:
+                    stats_str += f"  - 其他阶段 {i+1}: {other_duration_percent:.1f}%\n"
+        
         # 添加成本分布百分比
         stats_str += (
             f"\n"

autocoder/rag/types.py

@@ -1,4 +1,3 @@
-
 import os
 import json
 import time
@@ -13,6 +12,7 @@ class RecallStat(BaseModel):
     total_generated_tokens: int
     model_name: str = "unknown"
     cost:float = 0.0
+    duration: float = 0.0
 
 
 class ChunkStat(BaseModel):
@@ -20,6 +20,7 @@ class ChunkStat(BaseModel):
     total_generated_tokens: int
     model_name: str = "unknown"
     cost:float = 0.0
+    duration: float = 0.0
 
 
 class AnswerStat(BaseModel):
@@ -27,6 +28,7 @@ class AnswerStat(BaseModel):
     total_generated_tokens: int
     model_name: str = "unknown"
     cost:float = 0.0
+    duration: float = 0.0
 
 
 class OtherStat(BaseModel):
@@ -34,6 +36,7 @@ class OtherStat(BaseModel):
     total_generated_tokens: int = 0
     model_name: str = "unknown"
     cost:float = 0.0
+    duration: float = 0.0
 
 
 class RAGStat(BaseModel):

autocoder/version.py

@@ -1,2 +1,2 @@
 
-__version__ = "0.1.383"
+__version__ = "0.1.385"