auto-coder 0.1.289__tar.gz → 0.1.290__tar.gz

{auto_coder-0.1.289 → auto_coder-0.1.290}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: auto-coder
-Version: 0.1.289
+Version: 0.1.290
 Summary: AutoCoder: AutoCoder
 Author: allwefantasy
 Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
@@ -26,7 +26,7 @@ Requires-Dist: tabulate
 Requires-Dist: jupyter_client
 Requires-Dist: prompt-toolkit
 Requires-Dist: tokenizers
-Requires-Dist: byzerllm[saas]>=0.1.171
+Requires-Dist: byzerllm[saas]>=0.1.174
 Requires-Dist: patch
 Requires-Dist: diff_match_patch
 Requires-Dist: GitPython

{auto_coder-0.1.289 → auto_coder-0.1.290}/src/auto_coder.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: auto-coder
-Version: 0.1.289
+Version: 0.1.290
 Summary: AutoCoder: AutoCoder
 Author: allwefantasy
 Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
@@ -26,7 +26,7 @@ Requires-Dist: tabulate
 Requires-Dist: jupyter_client
 Requires-Dist: prompt-toolkit
 Requires-Dist: tokenizers
-Requires-Dist: byzerllm[saas]>=0.1.171
+Requires-Dist: byzerllm[saas]>=0.1.174
 Requires-Dist: patch
 Requires-Dist: diff_match_patch
 Requires-Dist: GitPython

{auto_coder-0.1.289 → auto_coder-0.1.290}/src/auto_coder.egg-info/SOURCES.txt

@@ -124,6 +124,7 @@ src/autocoder/privacy/model_filter.py
 src/autocoder/pyproject/__init__.py
 src/autocoder/rag/__init__.py
 src/autocoder/rag/api_server.py
+src/autocoder/rag/conversation_to_queries.py
 src/autocoder/rag/doc_filter.py
 src/autocoder/rag/document_retriever.py
 src/autocoder/rag/lang.py
@@ -147,8 +148,10 @@ src/autocoder/rag/variable_holder.py
 src/autocoder/rag/cache/__init__.py
 src/autocoder/rag/cache/base_cache.py
 src/autocoder/rag/cache/byzer_storage_cache.py
+src/autocoder/rag/cache/cache_result_merge.py
 src/autocoder/rag/cache/file_monitor_cache.py
 src/autocoder/rag/cache/local_byzer_storage_cache.py
+src/autocoder/rag/cache/rag_file_meta.py
 src/autocoder/rag/cache/simple_cache.py
 src/autocoder/rag/loaders/__init__.py
 src/autocoder/rag/loaders/docx_loader.py

{auto_coder-0.1.289 → auto_coder-0.1.290}/src/auto_coder.egg-info/requires.txt

@@ -16,7 +16,7 @@ tabulate
 jupyter_client
 prompt-toolkit
 tokenizers
-byzerllm[saas]>=0.1.171
+byzerllm[saas]>=0.1.174
 patch
 diff_match_patch
 GitPython

{auto_coder-0.1.289 → auto_coder-0.1.290}/src/autocoder/auto_coder_rag.py

@@ -262,6 +262,7 @@ def main(input_args: Optional[List[str]] = None):
     serve_parser.add_argument("--source_dir", default=".", help="")
     serve_parser.add_argument("--host", default="", help="")
     serve_parser.add_argument("--port", type=int, default=8000, help="")
+    serve_parser.add_argument("--workers", type=int, default=4, help="")
     serve_parser.add_argument("--uvicorn_log_level", default="info", help="")
     serve_parser.add_argument("--allow_credentials", action="store_true", help="")
     serve_parser.add_argument("--allowed_origins", default=["*"], help="")
@@ -274,6 +275,7 @@ def main(input_args: Optional[List[str]] = None):
     serve_parser.add_argument("--ssl_certfile", default="", help="")
     serve_parser.add_argument("--response_role", default="assistant", help="")
     serve_parser.add_argument("--doc_dir", default="", help="")
+    serve_parser.add_argument("--enable_local_image_host", action="store_true", help=" enable local image host for local Chat app")
     serve_parser.add_argument("--tokenizer_path", default=tokenizer_path, help="")
     serve_parser.add_argument(
         "--collections", default="", help="Collection name for indexing"
@@ -516,6 +518,14 @@ def main(input_args: Optional[List[str]] = None):
                 if hasattr(args, arg)
             }
         )
+        # 设置本地图床的地址
+        if args.enable_local_image_host:
+            host = server_args.host or "127.0.0.1"
+            if host == "0.0.0.0":
+                host = "127.0.0.1"
+            port = str(server_args.port)
+            auto_coder_args.local_image_host = f"{host}:{port}"
+
 
         # Generate unique name for RAG build if doc_dir exists
         if server_args.doc_dir:

{auto_coder-0.1.289 → auto_coder-0.1.290}/src/autocoder/common/__init__.py

@@ -296,6 +296,10 @@ class AutoCoderArgs(pydantic.BaseModel):
     rag_params_max_tokens: Optional[int] = 4096 
     rag_doc_filter_relevance: Optional[int] = 5
     rag_context_window_limit: Optional[int] = 120000 
+    # rag 本地图床地址
+    local_image_host: Optional[str] = ""
+    rag_recall_max_queries: Optional[int] = 5
+
     
     # 回答用户问题时，使用哪种对话历史策略
     # single_round: 单轮对话

{auto_coder-0.1.289 → auto_coder-0.1.290}/src/autocoder/rag/api_server.py

@@ -1,10 +1,14 @@
 import os
 import time
+import aiofiles
 import uvicorn
 from fastapi import Request
 from fastapi import FastAPI
 from fastapi.middleware.cors import CORSMiddleware
 from fastapi.responses import JSONResponse, StreamingResponse, Response
+from fastapi import HTTPException
+import mimetypes
+from urllib.parse import unquote
 
 from byzerllm.log import init_logger
 from byzerllm.utils import random_uuid
@@ -165,9 +169,52 @@ async def embed(body: EmbeddingCompletionRequest):
         id=embedding_id
     )
 
+@router_app.get("/static/{full_path:path}")
+async def serve_image(full_path: str, request: Request):
+    
+    allowed_file_type = ['.png', '.jpg', '.jpeg', '.gif', '.bmp', '.webp']
+    
+    if any(full_path.endswith(ext) for ext in allowed_file_type):
+        try:
+            # 获取文件的完整路径，并进行URL解码
+            file_path = unquote(full_path)
+            # 使用 os.path.normpath 来标准化路径，自动处理不同操作系统的路径分隔符
+            file_path = os.path.normpath(file_path)
+            if not os.path.isabs(file_path):
+                file_path = os.path.join("/", file_path)
+            
+            # 检查文件是否存在
+            if not os.path.exists(file_path):
+                raise FileNotFoundError(f"File not found: {file_path}")
+                
+            # 异步读取文件内容
+            async with aiofiles.open(file_path, "rb") as f:
+                content = await f.read()
+            
+            # 获取文件的 MIME 类型
+            content_type = mimetypes.guess_type(file_path)[0]
+            if not content_type:
+                content_type = "application/octet-stream"
+                
+            # 返回文件内容
+            return Response(content=content, media_type=content_type)
+        except FileNotFoundError as e:
+            logger.error(f"Image not found: {str(e)}")
+            raise HTTPException(status_code=404, detail=f"Image not found: {str(e)}")
+        except PermissionError as e:
+            logger.error(f"Permission denied: {str(e)}")
+            raise HTTPException(status_code=403, detail=f"Permission denied: {str(e)}")
+        except Exception as e:
+            logger.error(f"Error serving image: {str(e)}")
+            raise HTTPException(status_code=500, detail=f"Error serving image: {str(e)}")
+    
+    # 如果路径中没有图片, 返回 404
+    raise HTTPException(status_code=404, detail="Only images are supported")
+
 class ServerArgs(BaseModel):
     host: str = None
     port: int = 8000
+    workers: int = 4
     uvicorn_log_level: str = "info"
     allow_credentials: bool = False
     allowed_origins: List[str] = ["*"]  
@@ -227,6 +274,7 @@ def serve(llm:ByzerLLM, args: ServerArgs):
         prompt_template=args.prompt_template
     )
 
+    # 如果使用workers>1或reload=True，必须使用导入字符串而不是应用实例    
     uvicorn.run(
         router_app,
         host=args.host,

{auto_coder-0.1.289 → auto_coder-0.1.290}/src/autocoder/rag/cache/byzer_storage_cache.py

@@ -28,6 +28,8 @@ import platform
 import hashlib
 from typing import Union
 from pydantic import BaseModel
+from autocoder.rag.cache.cache_result_merge import CacheResultMerger, MergeStrategy
+import time
 
 if platform.system() != "Windows":
     import fcntl
@@ -64,6 +66,69 @@ class ByzerStorageCache(BaseCacheManager):
         required_exts,
         extra_params: Optional[AutoCoderArgs] = None,
     ):
+        """
+        初始化基于云端 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
+            
+        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. 云端 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
+            
+        与 LocalByzerStorageCache 的区别:
+            - 使用云端 ByzerStorage 而非本地存储
+            - 适用于需要远程访问和共享索引的场景
+            - 支持大规模分布式检索和更高级的查询功能
+        """
         self.path = path
         self.ignore_spec = ignore_spec
         self.required_exts = required_exts
@@ -185,7 +250,7 @@ class ByzerStorageCache(BaseCacheManager):
 
     def build_cache(self):
         """Build the cache by reading files and storing in Byzer Storage"""
-        logger.info(f"Building cache for path: {self.path}")
+        logger.info(f"[BUILD CACHE] Starting cache build for path: {self.path}")
 
         files_to_process = []
         for file_info in self.get_all_files():            
@@ -195,11 +260,15 @@ class ByzerStorageCache(BaseCacheManager):
             ):
                 files_to_process.append(file_info)
                 
+        logger.info(f"[BUILD CACHE] Found {len(files_to_process)} files to process")
         if not files_to_process:
+            logger.info("[BUILD CACHE] No files to process, cache build completed")
             return
 
         from autocoder.rag.token_counter import initialize_tokenizer
 
+        logger.info("[BUILD CACHE] Starting parallel file processing...")
+        start_time = time.time()
         with Pool(
             processes=os.cpu_count(),
             initializer=initialize_tokenizer,
@@ -209,6 +278,8 @@ class ByzerStorageCache(BaseCacheManager):
             for file_info in files_to_process:
                 target_files_to_process.append(self.fileinfo_to_tuple(file_info))
             results = pool.map(process_file_in_multi_process, target_files_to_process)
+        processing_time = time.time() - start_time
+        logger.info(f"[BUILD CACHE] File processing completed, time elapsed: {processing_time:.2f}s")
 
         items = []
         for file_info, result in zip(files_to_process, results):            
@@ -222,109 +293,176 @@ class ByzerStorageCache(BaseCacheManager):
             )
 
             for doc in content:
-                logger.info(f"Processing file: {doc.module_name}")
+                logger.info(f"[BUILD CACHE] Processing file: {doc.module_name}")
                 doc.module_name
                 chunks = self._chunk_text(doc.source_code, self.chunk_size)
+                logger.info(f"[BUILD CACHE] File {doc.module_name} chunking completed, total chunks: {len(chunks)}")
                 for chunk_idx, chunk in enumerate(chunks):
                     chunk_item = {
                         "_id": f"{doc.module_name}_{chunk_idx}",
                         "file_path": file_info.file_path,
-                        "content": chunk,
-                        "raw_content": chunk,
-                        "vector": chunk,
+                        "content": chunk[0:self.chunk_size*2],
+                        "raw_content": chunk[0:self.chunk_size*2],
+                        "vector": chunk[0:self.chunk_size*2],
                         "mtime": file_info.modify_time,
                     }
                     items.append(chunk_item)
 
         # Save to local cache
-        logger.info("Saving cache to local file")
+        logger.info("[BUILD CACHE] Saving cache to local file")
         self.write_cache()
         
         if items:
-            logger.info("Clear cache from Byzer Storage")
+            logger.info("[BUILD CACHE] Clearing existing cache from Byzer Storage")
             self.storage.truncate_table()
-            logger.info("Save new cache to Byzer Storage")            
-            max_workers = 5            
-            chunk_size = max(1, len(items) // max_workers)
-            item_chunks = [items[i:i + chunk_size] for i in range(0, len(items), chunk_size)]
+            logger.info(f"[BUILD CACHE] Preparing to write to Byzer Storage, total chunks: {len(items)}, total files: {len(files_to_process)}")
+            
+            # Use a fixed optimal batch size instead of dividing by worker count
+            batch_size = 100  # Optimal batch size for Byzer Storage
+            item_batches = [items[i:i + batch_size] for i in range(0, len(items), batch_size)]
             
-            total_chunks = len(item_chunks)
-            completed_chunks = 0
+            total_batches = len(item_batches)
+            completed_batches = 0
 
-            logger.info(f"Progress: {0}/{total_chunks} chunks completed")
+            logger.info(f"[BUILD CACHE] Starting to write to Byzer Storage using {batch_size} items per batch, "
+                        f"total batches: {total_batches}")
+            start_time = time.time()
+            
+            # Use more workers to process the smaller batches efficiently
+            max_workers = min(10, total_batches)  # Cap at 10 workers or total batch count
+            logger.info(f"[BUILD CACHE] Using {max_workers} parallel workers for processing")
             
             with ThreadPoolExecutor(max_workers=max_workers) as executor:
                 futures = []
-                for chunk in item_chunks:
+                # Submit all batches to the executor upfront (non-blocking)
+                for batch in item_batches:
                     futures.append(
                         executor.submit(
                             lambda x: self.storage.write_builder().add_items(
                                 x, vector_fields=["vector"], search_fields=["content"]
                             ).execute(),
-                            chunk
+                            batch
                         )
                     )
-                # Wait for all futures to complete
+                # Wait for futures to complete
                 for future in as_completed(futures):
                     try:
                         future.result()
-                        completed_chunks += 1
-                        logger.info(f"Progress: {completed_chunks}/{total_chunks} chunks completed")
+                        completed_batches += 1
+                        elapsed = time.time() - start_time
+                        estimated_total = elapsed / completed_batches * total_batches if completed_batches > 0 else 0
+                        remaining = estimated_total - elapsed
+                        
+                        # Only log progress at reasonable intervals to reduce log spam
+                        if completed_batches == 1 or completed_batches == total_batches or completed_batches % max(1, total_batches // 10) == 0:
+                            logger.info(
+                                f"[BUILD CACHE] Progress: {completed_batches}/{total_batches} batches completed "
+                                f"({(completed_batches/total_batches*100):.1f}%) "
+                                f"Estimated time remaining: {remaining:.1f}s"
+                            )
                     except Exception as e:
-                        logger.error(f"Error in saving chunk: {str(e)}")
+                        logger.error(f"[BUILD CACHE] Error saving batch: {str(e)}")
+                        # Add more detailed error information
+                        logger.error(f"[BUILD CACHE] Error details: batch size: {len(batch) if 'batch' in locals() else 'unknown'}")
             
+            total_time = time.time() - start_time
+            logger.info(f"[BUILD CACHE] All chunks written, total time: {total_time:.2f}s")
             self.storage.commit()
+            logger.info("[BUILD CACHE] Changes committed to Byzer Storage")
 
     def update_storage(self, file_info: FileInfo, is_delete: bool):
+        """
+        Updates file content in the Byzer Storage vector database.
+        
+        Parameters:
+            file_info: FileInfo object containing file path, relative path, modify time, and MD5 hash
+            is_delete: Whether this is a delete operation, True means all records for this file will be removed
+        """
+        logger.info(f"[UPDATE STORAGE] Starting update for file: {file_info.file_path}, is delete: {is_delete}")
+        
         query = self.storage.query_builder()
         query.and_filter().add_condition("file_path", file_info.file_path).build()
         results = query.execute()
         if results:
+            logger.info(f"[UPDATE STORAGE] Deleting existing records from Byzer Storage: {len(results)} records")
             for result in results:
                 self.storage.delete_by_ids([result["_id"]])
         items = []
 
         if not is_delete:
+            logger.info(f"[UPDATE STORAGE] Getting file content from cache and preparing update")
             content = [
                 SourceCode.model_validate(doc)
                 for doc in self.cache[file_info.file_path].content
             ]
             modify_time = self.cache[file_info.file_path].modify_time
             for doc in content:
-                logger.info(f"Processing file: {doc.module_name}")
+                logger.info(f"[UPDATE STORAGE] Processing file: {doc.module_name}")
                 doc.module_name
                 chunks = self._chunk_text(doc.source_code, self.chunk_size)
+                logger.info(f"[UPDATE STORAGE] File {doc.module_name} chunking completed, total chunks: {len(chunks)}")
                 for chunk_idx, chunk in enumerate(chunks):
                     chunk_item = {
                         "_id": f"{doc.module_name}_{chunk_idx}",
                         "file_path": file_info.file_path,
-                        "content": chunk,
-                        "raw_content": chunk,
-                        "vector": chunk,
+                        "content": chunk[0:self.chunk_size*2],
+                        "raw_content": chunk[0:self.chunk_size*2],
+                        "vector": chunk[0:self.chunk_size*2],
                         "mtime": modify_time,
                     }
                     items.append(chunk_item)
         if items:
-            self.storage.write_builder().add_items(
-                items, vector_fields=["vector"], search_fields=["content"]
-            ).execute()
+            logger.info(f"[UPDATE STORAGE] Starting to write {len(items)} chunks to Byzer Storage")
+            start_time = time.time()
+            
+            # Use optimal batch size for larger updates
+            batch_size = 100
+            if len(items) > batch_size:
+                logger.info(f"[UPDATE STORAGE] Using batched writes with {batch_size} items per batch")
+                batches = [items[i:i + batch_size] for i in range(0, len(items), batch_size)]
+                total_batches = len(batches)
+                
+                for i, batch in enumerate(batches):
+                    self.storage.write_builder().add_items(
+                        batch, vector_fields=["vector"], search_fields=["content"]
+                    ).execute()
+                    logger.info(f"[UPDATE STORAGE] Progress: {i+1}/{total_batches} batches written")
+            else:
+                # For small item counts, just use a single write operation
+                self.storage.write_builder().add_items(
+                    items, vector_fields=["vector"], search_fields=["content"]
+                ).execute()
+                
             self.storage.commit()
+            elapsed = time.time() - start_time
+            logger.info(f"[UPDATE STORAGE] Write completed, time elapsed: {elapsed:.2f}s")
+        else:
+            logger.info(f"[UPDATE STORAGE] No content to write")
 
     def process_queue(self):
+        if not self.queue:
+            logger.info("[QUEUE PROCESSING] Queue is empty, nothing to process")
+            return
+            
+        logger.info(f"[QUEUE PROCESSING] Starting queue processing, queue length: {len(self.queue)}")
+        start_time = time.time()
+        
         while self.queue:
             file_list = self.queue.pop(0)
             if isinstance(file_list, DeleteEvent):
+                logger.info(f"[QUEUE PROCESSING] Processing delete event, total files: {len(file_list.file_paths)}")
                 for item in file_list.file_paths:
-                    logger.info(f"{item} is detected to be removed")
+                    logger.info(f"[QUEUE PROCESSING] Processing file deletion: {item}")
                     del self.cache[item]
-                    # 创建一个临时的 FileInfo 对象
+                    # Create a temporary FileInfo object
                     file_info = FileInfo(file_path=item, relative_path="", modify_time=0, file_md5="")
                     self.update_storage(file_info, is_delete=True)
 
             elif isinstance(file_list, AddOrUpdateEvent):
+                logger.info(f"[QUEUE PROCESSING] Processing add/update event, total files: {len(file_list.file_infos)}")
                 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
+                    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,
@@ -335,9 +473,14 @@ class ByzerStorageCache(BaseCacheManager):
                     )
                     self.update_storage(file_info, is_delete=False)
             self.write_cache()
+        
+        elapsed = time.time() - start_time
+        logger.info(f"[QUEUE PROCESSING] Queue processing completed, time elapsed: {elapsed:.2f}s")
 
     def trigger_update(self):
-        logger.info("检查文件是否有更新.....")
+        logger.info("[TRIGGER UPDATE] Starting file update check...")
+        start_time = time.time()
+        
         files_to_process = []
         current_files = set()
         for file_info in self.get_all_files():            
@@ -349,26 +492,26 @@ class ByzerStorageCache(BaseCacheManager):
                 files_to_process.append(file_info)
 
         deleted_files = set(self.cache.keys()) - current_files
-        logger.info(f"files_to_process: {files_to_process}")
-        logger.info(f"deleted_files: {deleted_files}")
+        
+        logger.info(f"[TRIGGER UPDATE] Files to process: {len(files_to_process)}")
+        logger.info(f"[TRIGGER UPDATE] Files deleted: {len(deleted_files)}")
+        
         if deleted_files:
+            logger.info(f"[TRIGGER UPDATE] Adding delete event to queue")
             with self.lock:
                 self.queue.append(DeleteEvent(file_paths=deleted_files))
         if files_to_process:
+            logger.info(f"[TRIGGER UPDATE] Adding update event to queue")
             with self.lock:
                 self.queue.append(AddOrUpdateEvent(file_infos=files_to_process))
+        
+        elapsed = time.time() - start_time
+        logger.info(f"[TRIGGER UPDATE] Check completed, time elapsed: {elapsed:.2f}s")
 
-    def get_cache(self, options: Dict[str, Any]) -> Dict[str, Dict]:
+    def get_single_cache(self, query: str,options: Dict[str, Any]) -> Dict[str, Dict]:
         """Search cached documents using query"""
-
-        self.trigger_update()
-
-        if options is None or "query" not in options:
-            return {file_path: self.cache[file_path].model_dump() for file_path in self.cache}
-        
-        query = options.get("query", "")
         total_tokens = 0    
-
+        logger.info(f"查询缓存 query: {query}")
         # Build query with both vector search and text search
         query_builder = self.storage.query_builder()
         query_builder.set_limit(100000)
@@ -383,6 +526,14 @@ class ByzerStorageCache(BaseCacheManager):
 
         results = query_builder.execute()
 
+        return results
+
+
+    def _process_search_results(self, results):
+        """处理搜索结果，提取文件路径并构建结果字典"""
+        # 记录被处理的总tokens数
+        total_tokens = 0
+        
         # Group results by file_path and reconstruct documents while preserving order
         # 这里还可以有排序优化，综合考虑一篇内容出现的次数以及排序位置
         file_paths = []
@@ -400,11 +551,70 @@ class ByzerStorageCache(BaseCacheManager):
                 cached_data = self.cache[file_path]
                 for doc in cached_data.content:                    
                     if total_tokens + doc["tokens"] > self.max_output_tokens:
+                        logger.info(f"用户tokens设置为:{self.max_output_tokens}，累计tokens: {total_tokens} 当前文件: {file_path} tokens: {doc['tokens']}，数据条数变化: {len(results)} -> {len(result)}")
                         return result
                     total_tokens += doc["tokens"]
                 result[file_path] = cached_data.model_dump()
-
+        
+        logger.info(f"用户tokens设置为:{self.max_output_tokens}，累计tokens: {total_tokens}，数据条数变化: {len(results)} -> {len(result)}")
         return result
+        
+
+    def get_cache(self, options: Dict[str, Any]) -> Dict[str, Dict]:
+        """
+        获取缓存中的文档信息
+        
+        如果options中包含query，则根据query搜索；否则返回所有缓存
+        """
+        # options是一个词典，词典的key是搜索参数，value是具体值
+
+        # 触发更新
+        self.trigger_update()
+
+        if options is None or "queries" not in options:
+            return {file_path: self.cache[file_path].model_dump() for file_path in self.cache}
+        
+        queries = options.get("queries", [])
+        
+        # 如果没有查询或只有一个查询，使用原来的方法
+        if not queries:
+            return {file_path: self.cache[file_path].model_dump() for file_path in self.cache}
+        elif len(queries) == 1:
+            results = self.get_single_cache(queries[0], options)
+            return self._process_search_results(results)
+        
+        # 获取合并策略
+        merge_strategy_name = options.get("merge_strategy", MergeStrategy.WEIGHTED_RANK.value)
+        try:
+            merge_strategy = MergeStrategy(merge_strategy_name)
+        except ValueError:
+            logger.warning(f"未知的合并策略: {merge_strategy_name}，使用默认策略")
+            merge_strategy = MergeStrategy.WEIGHTED_RANK
+        
+        # 限制最大结果数
+        max_results = options.get("max_results", None)
+        merger = CacheResultMerger(max_results=max_results)
+        
+        # 并发处理多个查询
+        query_results = []
+        with ThreadPoolExecutor(max_workers=min(len(queries), 10)) as executor:
+            future_to_query = {executor.submit(self.get_single_cache, query, options): query for query in queries}
+            for future in as_completed(future_to_query):
+                query = future_to_query[future]
+                try:
+                    query_result = future.result()
+                    logger.info(f"查询 '{query}' 返回 {len(query_result)} 条结果")
+                    query_results.append((query, query_result))
+                except Exception as e:
+                    logger.error(f"处理查询 '{query}' 时出错: {str(e)}")
+        
+        logger.info(f"所有查询共返回 {sum(len(r) for _, r in query_results)} 条结果")
+        logger.info(f"使用合并策略: {merge_strategy}")
+        
+        # 使用策略合并结果
+        merged_results = merger.merge(query_results, strategy=merge_strategy)
+        
+        return self._process_search_results(merged_results)