auto-coder 0.1.345__py3-none-any.whl → 0.1.347__py3-none-any.whl

autocoder/rag/cache/file_monitor_cache.py

@@ -48,7 +48,7 @@ class AutoCoderRAGDocListener(BaseCacheManager):
         r"^test.*$",
     ]
 
-    def __init__(self, path: str, ignore_spec, required_exts: List) -> None:
+    def __init__(self, path: str, ignore_spec, required_exts: List, args=None, llm=None) -> None:
         """
         初始化文件监控缓存管理器。
         
@@ -89,6 +89,8 @@ class AutoCoderRAGDocListener(BaseCacheManager):
         self.path = path
         self.ignore_spec = ignore_spec
         self.required_exts = required_exts
+        self.args = args
+        self.llm = llm
         self.stop_event = threading.Event()
 
         # connect list

autocoder/rag/cache/local_byzer_storage_cache.py

@@ -30,6 +30,7 @@ from typing import Union
 from byzerllm import SimpleByzerLLM, ByzerLLM
 from autocoder.rag.cache.cache_result_merge import CacheResultMerger, MergeStrategy
 import time
+from .failed_files_utils import save_failed_files, load_failed_files
 
 if platform.system() != "Windows":
     import fcntl
@@ -70,71 +71,17 @@ class LocalByzerStorageCache(BaseCacheManager):
         emb_llm: Union[ByzerLLM, SimpleByzerLLM] = None,
         host: str = "127.0.0.1",
         port: int = 33333,
+        args=None,
+        llm=None,
     ):
         """
         初始化基于 Byzer Storage 的 RAG 缓存管理器。
-        
-        参数:
-            path: 需要索引的代码库根目录
-            ignore_spec: 指定哪些文件/目录应被忽略的规则
-            required_exts: 需要处理的文件扩展名列表
-            extra_params: 额外的配置参数，包含向量索引相关设置
-            emb_llm: 用于生成文本向量嵌入的 ByzerLLM 实例
-            host: Byzer Storage 服务的主机地址
-            port: Byzer Storage 服务的端口
-            
-        缓存结构 (self.cache):
-            self.cache 是一个字典，键为文件路径，值为 CacheItem 对象:
-            {
-                "file_path1": CacheItem(
-                    file_path: str,              # 文件的绝对路径
-                    relative_path: str,          # 相对于项目根目录的路径
-                    content: List[Dict],         # 文件内容的结构化表示，每个元素是 SourceCode 对象的序列化
-                    modify_time: float,          # 文件最后修改时间的时间戳
-                    md5: str                     # 文件内容的 MD5 哈希值，用于检测变更
-                ),
-                "file_path2": CacheItem(...),
-                ...
-            }
-            
-            这个缓存有两层存储:
-            1. 本地文件缓存: 保存在项目根目录的 .cache/byzer_storage_speedup.jsonl 文件中
-               - 用于跟踪文件变更和快速加载
-               - 使用 JSONL 格式存储，每行是一个 CacheItem 的 JSON 表示
-            
-            2. Byzer Storage 向量数据库:
-               - 存储文件内容的分块和向量嵌入
-               - 每个文件被分割成大小为 chunk_size 的文本块
-               - 每个块都会生成向量嵌入，用于语义搜索
-               - 存储结构包含: 文件路径、内容块、原始内容、向量嵌入、修改时间
-        
-        源代码处理流程:
-            在缓存更新过程中使用了两个关键函数:
-            
-            1. process_file_in_multi_process: 在多进程环境中处理文件
-               - 参数: file_info (文件信息元组)
-               - 返回值: List[SourceCode] 或 None
-               - 用途: 在初始构建缓存时并行处理多个文件
-            
-            2. process_file_local: 在当前进程中处理单个文件
-               - 参数: file_path (文件路径)
-               - 返回值: List[SourceCode] 或 None
-               - 用途: 在检测到文件更新时处理单个文件
-            
-            文件处理后，会:
-            1. 更新内存中的缓存 (self.cache)
-            2. 将缓存持久化到本地文件
-            3. 将内容分块并更新到 Byzer Storage 向量数据库
-        
-        更新机制:
-            - 通过单独的线程异步处理文件变更
-            - 使用 MD5 哈希值检测文件是否发生变化
-            - 支持文件添加、更新和删除事件
-            - 使用向量数据库进行语义检索，支持相似度搜索
         """
         self.path = path
         self.ignore_spec = ignore_spec
         self.required_exts = required_exts
+        self.args = args
+        self.llm = llm
         self.rag_build_name = extra_params.rag_build_name
         self.storage = LocalByzerStorage("byzerai_store",
             "rag_test", self.rag_build_name, host=host, port=port,emb_llm=emb_llm)
@@ -153,16 +100,20 @@ class LocalByzerStorageCache(BaseCacheManager):
             self.cache_dir, "byzer_storage_speedup.jsonl")
         self.cache: Dict[str, CacheItem] = {}
 
+        # 创建缓存目录
+        if not os.path.exists(self.cache_dir):
+            os.makedirs(self.cache_dir)
+
+        # failed files support
+        self.failed_files_path = os.path.join(self.cache_dir, "failed_files.json")
+        self.failed_files = load_failed_files(self.failed_files_path)
+
         self.lock = threading.Lock()
         self.stop_event = threading.Event()
         self.thread = threading.Thread(target=self.process_queue)
         self.thread.daemon = True
         self.thread.start()
 
-        # 创建缓存目录
-        if not os.path.exists(self.cache_dir):
-            os.makedirs(self.cache_dir)
-
         # 加载缓存
         self.cache = self._load_cache()
 
@@ -485,6 +436,10 @@ class LocalByzerStorageCache(BaseCacheManager):
                 for item in file_list.file_paths:
                     logger.info(f"[QUEUE PROCESSING] Processing file deletion: {item}")
                     del self.cache[item]
+                    # remove from failed files if present
+                    if item in self.failed_files:
+                        self.failed_files.remove(item)                        
+                        save_failed_files(self.failed_files_path, self.failed_files)
                     # Create a temporary FileInfo object
                     file_info = FileInfo(
                         file_path=item, relative_path="", modify_time=0, file_md5="")
@@ -495,17 +450,30 @@ class LocalByzerStorageCache(BaseCacheManager):
                 for file_info in file_list.file_infos:
                     logger.info(
                         f"[QUEUE PROCESSING] Processing file update: {file_info.file_path}")
-                    # Process file and create CacheItem
-                    content = process_file_local(
-                        self.fileinfo_to_tuple(file_info))
-                    self.cache[file_info.file_path] = CacheItem(
-                        file_path=file_info.file_path,
-                        relative_path=file_info.relative_path,
-                        content=[c.model_dump() for c in content],
-                        modify_time=file_info.modify_time,
-                        md5=file_info.file_md5,
-                    )
-                    self.update_storage(file_info, is_delete=False)
+                    try:
+                        content = process_file_local(
+                            self.fileinfo_to_tuple(file_info))
+                        if content:
+                            self.cache[file_info.file_path] = CacheItem(
+                                file_path=file_info.file_path,
+                                relative_path=file_info.relative_path,
+                                content=[c.model_dump() for c in content],
+                                modify_time=file_info.modify_time,
+                                md5=file_info.file_md5,
+                            )
+                            self.update_storage(file_info, is_delete=False)
+                            # remove from failed files if present
+                            if file_info.file_path in self.failed_files:
+                                self.failed_files.remove(file_info.file_path)                                
+                                save_failed_files(self.failed_files_path, self.failed_files)
+                        else:
+                            logger.warning(f"Empty result for file: {file_info.file_path}, treat as parse failed, skipping cache update")
+                            self.failed_files.add(file_info.file_path)                            
+                            save_failed_files(self.failed_files_path, self.failed_files)
+                    except Exception as e:
+                        logger.error(f"Error in process_queue: {e}")
+                        self.failed_files.add(file_info.file_path)                        
+                        save_failed_files(self.failed_files_path, self.failed_files)
             self.write_cache()
         
         elapsed = time.time() - start_time
@@ -519,6 +487,10 @@ class LocalByzerStorageCache(BaseCacheManager):
         current_files = set()
         for file_info in self.get_all_files():
             current_files.add(file_info.file_path)
+            # skip failed files
+            if file_info.file_path in self.failed_files:
+                logger.info(f"文件 {file_info.file_path} 之前解析失败，跳过此次更新")
+                continue
             if (
                 file_info.file_path not in self.cache
                 or self.cache[file_info.file_path].md5 != file_info.file_md5

autocoder/rag/cache/local_duckdb_storage_cache.py

@@ -28,6 +28,7 @@ from autocoder.rag.cache.base_cache import (
 from autocoder.rag.utils import process_file_in_multi_process, process_file_local
 from autocoder.rag.variable_holder import VariableHolder
 from byzerllm import SimpleByzerLLM, ByzerLLM
+from .failed_files_utils import save_failed_files, load_failed_files
 
 if platform.system() != "Windows":
     import fcntl
@@ -300,12 +301,16 @@ class LocalDuckDBStorageCache(BaseCacheManager):
             ignore_spec,
             required_exts,
             extra_params: Optional[AutoCoderArgs] = None,
-            emb_llm: Union[ByzerLLM, SimpleByzerLLM] = None
+            emb_llm: Union[ByzerLLM, SimpleByzerLLM] = None,
+            args=None,
+            llm=None
     ):
         self.path = path
         self.ignore_spec = ignore_spec
         self.required_exts = required_exts
         self.extra_params = extra_params
+        self.args = args
+        self.llm = llm
 
         self.storage = LocalDuckdbStorage(
             llm=emb_llm,
@@ -325,6 +330,11 @@ class LocalDuckDBStorageCache(BaseCacheManager):
         if not os.path.exists(self.cache_dir):
             os.makedirs(self.cache_dir)
 
+        # failed files support
+        from .failed_files_utils import load_failed_files
+        self.failed_files_path = os.path.join(self.cache_dir, "failed_files.json")
+        self.failed_files = load_failed_files(self.failed_files_path)
+
         self.lock = threading.Lock()
         self.stop_event = threading.Event()
         self.thread = threading.Thread(target=self.process_queue)
@@ -569,6 +579,10 @@ class LocalDuckDBStorageCache(BaseCacheManager):
                 for item in file_list.file_paths:
                     logger.info(f"{item} is detected to be removed")
                     del self.cache[item]
+                    # remove from failed files if present
+                    if item in self.failed_files:
+                        self.failed_files.remove(item)
+                        save_failed_files(self.failed_files_path, self.failed_files)
                     # 创建一个临时的 FileInfo 对象
                     file_info = FileInfo(
                         file_path=item, relative_path="", modify_time=0, file_md5="")
@@ -578,18 +592,30 @@ class LocalDuckDBStorageCache(BaseCacheManager):
                 for file_info in file_list.file_infos:
                     logger.info(
                         f"{file_info.file_path} is detected to be updated")
-                    # 处理文件并创建 CacheItem
-                    # content = process_file_local(
-                    #     self.fileinfo_to_tuple(file_info))
-                    content = process_file_local(file_info.file_path)
-                    self.cache[file_info.file_path] = CacheItem(
-                        file_path=file_info.file_path,
-                        relative_path=file_info.relative_path,
-                        content=[c.model_dump() for c in content],
-                        modify_time=file_info.modify_time,
-                        md5=file_info.file_md5,
-                    )
-                    self.update_storage(file_info, is_delete=False)
+                    try:
+                        content = process_file_local(file_info.file_path)
+                        if content:
+                            self.cache[file_info.file_path] = CacheItem(
+                                file_path=file_info.file_path,
+                                relative_path=file_info.relative_path,
+                                content=[c.model_dump() for c in content],
+                                modify_time=file_info.modify_time,
+                                md5=file_info.file_md5,
+                            )
+                            self.update_storage(file_info, is_delete=False)
+                            # remove from failed files if present
+                            if file_info.file_path in self.failed_files:
+                                self.failed_files.remove(file_info.file_path)
+                                save_failed_files(self.failed_files_path, self.failed_files)
+                        else:
+                            logger.warning(f"Empty result for file: {file_info.file_path}, treat as parse failed, skipping cache update")
+                            self.failed_files.add(file_info.file_path)
+                            save_failed_files(self.failed_files_path, self.failed_files)
+                    except Exception as e:
+                        logger.error(f"Error in process_queue: {e}")
+                        self.failed_files.add(file_info.file_path)
+                        save_failed_files(self.failed_files_path, self.failed_files)
+
             self.write_cache()
 
     def trigger_update(self):
@@ -598,6 +624,10 @@ class LocalDuckDBStorageCache(BaseCacheManager):
         current_files = set()
         for file_info in self.get_all_files():
             current_files.add(file_info.file_path)
+            # skip failed files
+            if file_info.file_path in self.failed_files:
+                logger.info(f"文件 {file_info.file_path} 之前解析失败，跳过此次更新")
+                continue
             if (
                     file_info.file_path not in self.cache
                     or self.cache[file_info.file_path].md5 != file_info.file_md5

autocoder/rag/cache/simple_cache.py

@@ -19,6 +19,7 @@ from loguru import logger
 from autocoder.rag.utils import process_file_in_multi_process, process_file_local
 from autocoder.rag.variable_holder import VariableHolder
 import hashlib
+from .failed_files_utils import load_failed_files, save_failed_files
 
 
 default_ignore_dirs = [
@@ -45,7 +46,7 @@ def generate_content_md5(content: Union[str, bytes]) -> str:
 
 
 class AutoCoderRAGAsyncUpdateQueue(BaseCacheManager):
-    def __init__(self, path: str, ignore_spec, required_exts: list, update_interval: int = 5):
+    def __init__(self, path: str, ignore_spec, required_exts: list, update_interval: int = 5, args=None, llm=None):
         """
         初始化异步更新队列，用于管理代码文件的缓存。
         
@@ -91,24 +92,31 @@ class AutoCoderRAGAsyncUpdateQueue(BaseCacheManager):
         self.path = path
         self.ignore_spec = ignore_spec
         self.required_exts = required_exts
+        self.args = args
+        self.llm = llm
         self.update_interval = update_interval
         self.queue = []
         self.cache = {}  # 初始化为空字典，稍后通过 read_cache() 填充
         self.lock = threading.Lock()
         self.stop_event = threading.Event()
-        
+
+        # 用于存放解析失败的文件路径集合
+        self.failed_files_path = os.path.join(self.path, ".cache", "failed_files.json")
+        self.failed_files = load_failed_files(self.failed_files_path)
+
         # 启动处理队列的线程
         self.queue_thread = threading.Thread(target=self._process_queue)
         self.queue_thread.daemon = True
         self.queue_thread.start()
-        
+
         # 启动定时触发更新的线程
         self.update_thread = threading.Thread(target=self._periodic_update)
         self.update_thread.daemon = True
         self.update_thread.start()
-        
+
         self.cache = self.read_cache()
 
+
     def _process_queue(self):
         while not self.stop_event.is_set():
             try:
@@ -183,13 +191,18 @@ class AutoCoderRAGAsyncUpdateQueue(BaseCacheManager):
         files_to_process = []
         current_files = set()
         for file_info in self.get_all_files():
-            file_path, _, _, file_md5 = file_info
+            file_path, relative_path, modify_time, file_md5 = file_info
             current_files.add(file_path)
+            # 如果文件曾经解析失败，跳过本次增量更新
+            if file_path in self.failed_files:
+                logger.info(f"文件 {file_path} 之前解析失败，跳过此次更新")
+                continue
+            # 变更检测
             if (
                 file_path not in self.cache
-                or self.cache[file_path].get("md5","") != file_md5
+                or self.cache[file_path].get("md5", "") != file_md5
             ):
-                files_to_process.append(file_info)
+                files_to_process.append((file_path, relative_path, modify_time, file_md5))
 
         deleted_files = set(self.cache.keys()) - current_files
         logger.info(f"files_to_process: {files_to_process}")
@@ -213,19 +226,34 @@ class AutoCoderRAGAsyncUpdateQueue(BaseCacheManager):
             if isinstance(file_list, DeleteEvent):
                 for item in file_list.file_paths:
                     logger.info(f"{item} is detected to be removed")
-                    del self.cache[item]
+                    if item in self.cache:
+                        del self.cache[item]
+                    # 删除时也从失败列表中移除（防止文件已修复）
+                    if item in self.failed_files:
+                        self.failed_files.remove(item)
+                        save_failed_files(self.failed_files_path, self.failed_files)
             elif isinstance(file_list, AddOrUpdateEvent):
                 for file_info in file_list.file_infos:
                     logger.info(f"{file_info.file_path} is detected to be updated")
                     try:
                         result = process_file_local(file_info.file_path)
-                        if result:  # 只有当result不为空时才更新缓存
+                        if result:
+                            # 解析成功且非空
                             self.update_cache(self.fileinfo_to_tuple(file_info), result)
+                            # 如果之前失败过且本次成功，移除失败记录
+                            if file_info.file_path in self.failed_files:
+                                self.failed_files.remove(file_info.file_path)
+                                save_failed_files(self.failed_files_path, self.failed_files)
                         else:
-                            logger.warning(f"Empty result for file: {file_info.file_path}, skipping cache update")
+                            # 只要为空也认为解析失败，加入失败列表
+                            logger.warning(f"Empty result for file: {file_info.file_path}, treat as parse failed, skipping cache update")
+                            self.failed_files.add(file_info.file_path)
+                            save_failed_files(self.failed_files_path, self.failed_files)
                     except Exception as e:
-                        logger.error(
-                            f"SimpleCache Error in process_queue: {e}")
+                        logger.error(f"SimpleCache Error in process_queue: {e}")
+                        # 解析失败则加入失败列表
+                        self.failed_files.add(file_info.file_path)
+                        save_failed_files(self.failed_files_path, self.failed_files)
 
             self.write_cache()
 

autocoder/rag/document_retriever.py

@@ -2,7 +2,6 @@ import threading
 from typing import Dict, Generator, List, Tuple, Any, Optional,Union
 
 from byzerllm import ByzerLLM, SimpleByzerLLM
-
 from loguru import logger
 from autocoder.common import SourceCode
 from uuid import uuid4
@@ -37,6 +36,8 @@ class LocalDocumentRetriever(BaseDocumentRetriever):
 
     def __init__(
         self,
+        args: AutoCoderArgs,
+        llm: Union[ByzerLLM,SimpleByzerLLM],
         path: str,
         ignore_spec,
         required_exts: list,
@@ -45,9 +46,12 @@ class LocalDocumentRetriever(BaseDocumentRetriever):
         single_file_token_limit: int = 60000,
         disable_auto_window: bool = False,
         enable_hybrid_index: bool = False,
-        extra_params: Optional[AutoCoderArgs] = None,
-        emb_llm: Union[ByzerLLM, SimpleByzerLLM] = None,
+        extra_params: Optional['AutoCoderArgs'] = None,
+        emb_llm: Union['ByzerLLM', 'SimpleByzerLLM'] = None,
     ) -> None:
+        self.args = args
+        self.llm = llm
+
         self.path = path
         self.ignore_spec = ignore_spec
         self.required_exts = required_exts
@@ -65,27 +69,32 @@ class LocalDocumentRetriever(BaseDocumentRetriever):
         if self.enable_hybrid_index:
             if self.on_ray:
                 self.cacher = ByzerStorageCache(
-                    path, ignore_spec, required_exts, extra_params
+                    path, ignore_spec, required_exts, extra_params,
+                    args=self.args, llm=self.llm
                 )
             else:
                 if extra_params.rag_storage_type == "duckdb":
                     self.cacher = LocalDuckDBStorageCache(
                         path, ignore_spec, required_exts, extra_params,
-                        emb_llm=emb_llm
+                        emb_llm=emb_llm,
+                        args=self.args, llm=self.llm
                     )
                 elif extra_params.rag_storage_type in ["byzer-storage", "byzer_storage"]:
                     self.cacher = LocalByzerStorageCache(
                         path, ignore_spec, required_exts, extra_params,
-                        emb_llm=emb_llm
+                        emb_llm=emb_llm,
+                        args=self.args, llm=self.llm
                     )
         else:
             if self.monitor_mode:
                 self.cacher = AutoCoderRAGDocListener(
-                    path, ignore_spec, required_exts
+                    path, ignore_spec, required_exts,
+                    args=self.args, llm=self.llm
                 )
             else:
                 self.cacher = AutoCoderRAGAsyncUpdateQueue(
-                    path, ignore_spec, required_exts
+                    path, ignore_spec, required_exts,
+                    args=self.args, llm=self.llm
                 )
 
         logger.info(f"DocumentRetriever initialized with:")

autocoder/rag/long_context_rag.py

@@ -183,6 +183,8 @@ class LongContextRAG:
                     "emb_llm is required for local byzer storage cache")
 
         self.document_retriever = retriever_class(
+            self.args,
+            self.llm,
             self.path,
             self.ignore_spec,
             self.required_exts,
@@ -841,8 +843,7 @@ class LongContextRAG:
                 self._print_rag_stats(rag_stat)
             else:
 
-                qa_strategy = get_qa_strategy(
-                    self.args.rag_qa_conversation_strategy)
+                qa_strategy = get_qa_strategy(self.args)
                 new_conversations = qa_strategy.create_conversation(
                     documents=[doc.source_code for doc in relevant_docs],
                     conversations=conversations, local_image_host=self.args.local_image_host