feat(rag): 优化 vllm 作为 LLM 提供商支持

- 在 RAGEngine 中实现 vllm 提供商的支持
- 添加 OpenAILike 集成用于 vllm 兼容性
- 更新 docker-compose.yml 中的默认 LLM 配置为 vllm
- 添加 llama-index-llms-openai-like 依赖
- 优化同步模块中的文件路径处理逻辑
- 修复 SSH 连接和 SFTP 文件操作的路径兼容性问题

docker-compose.yml

@@ -28,6 +28,7 @@ services:
       # ChromaDB 容器默认将数据存储在 /data 目录
       - ./chroma_db_data:/data
     # 将主机端口映射为可配置（从 .env 读取 CHROMA_SERVER_PORT，默认 8002）
+    # 注意以下 8000端口是ChromaDB默认端口，不用更改
     ports:
       - "${CHROMA_SERVER_PORT:-8002}:8000"
     environment:
@@ -94,6 +95,13 @@ services:
       - OLLAMA_MODEL=${OLLAMA_MODEL:-qwen3:1.7b}
       - OLLAMA_EMBEDDING_MODEL=${OLLAMA_EMBEDDING_MODEL:-qwen3-embedding:0.6b}
 
+      # LLM 配置
+      # LLM provider: ollama, openai 等
+      - LLM_PROVIDER=${LLM_PROVIDER:-vllm}
+      - LLM_BASE_URL=${LLM_BASE_URL:-http://172.20.32.236:8000/v1}
+      - LLM_MODEL=${LLM_MODEL:-MiniMax-M2.7}
+      - LLM_API_KEY=${LLM_API_KEY:-none}  # 如使用openai等需要API Key的服务
+
       # RAG 配置
       - EMBEDDING_DIMENSION=${EMBEDDING_DIMENSION:-768}
       - CHUNK_SIZE=${CHUNK_SIZE:-1024}

rag/rag_engine.py

@@ -200,6 +200,17 @@ class RAGEngine:
                 temperature=self._temperature,
                 timeout=self._request_timeout,
             )
+        elif provider == "vllm":
+            from llama_index.llms.openai_like import OpenAILike
+            self.llm = OpenAILike(
+                model=model,
+                api_base=base_url,
+                api_key=api_key or "none",
+                is_chat_model=True,
+                context_window=128000,
+                temperature=self._temperature,
+                timeout=self._request_timeout,
+            )
         else:
             self.llm = Ollama(
                 model=model,

requirements.txt

@@ -43,3 +43,5 @@ markdown2>=2.5.4  # Markdown parsing
 paramiko>=3.4.0  # For SSH/SCP functionality
 
 requests~=2.32.5
+
+llama-index-llms-openai-like>=0.2.0

sync/folder_sync.py

@@ -1,5 +1,6 @@
 """Folder synchronization implementation for local and remote folders"""
 import os
+import posixpath
 import re
 from typing import List, Dict, Any, Set
 from datetime import datetime
@@ -309,7 +310,8 @@ class FolderSync(BaseSync):
             items = self._sftp_client.listdir_attr(folder_path)
 
             for item in items:
-                item_path = os.path.join(folder_path, item.filename)
+                # old:relative_path = os.path.relpath(file_path, self.config.folder_path)
+                item_path = posixpath.join(folder_path, item.filename)
 
                 if item.filename not in ('.', '..'):
                     if item.st_mode & 0o040000:  # Check if it's a directory
@@ -336,7 +338,8 @@ class FolderSync(BaseSync):
             return False
 
         # Get relative path from folder root
-        relative_path = os.path.relpath(file_path, self.config.folder_path)
+        # old: relative_path = os.path.relpath(file_path, self.config.folder_path)
+        relative_path = posixpath.relpath(file_path, self.config.folder_path)
 
         for pattern in self.config.ignore_patterns:
             if self._match_pattern(relative_path, pattern):