Update rag_server.py

rag_server.py

@@ -1,3 +1,14 @@
+import os
+import re
+import glob
+import time
+from collections import defaultdict
+
+from langchain.text_splitter import RecursiveCharacterTextSplitter
+from langchain_core.documents import Document
+from langchain_community.embeddings import HuggingFaceEmbeddings
+from langchain_community.vectorstores import FAISS
+
 from fastapi import FastAPI, Request
 from fastapi.responses import JSONResponse, FileResponse, HTMLResponse
 from fastapi.staticfiles import StaticFiles
@@ -6,8 +17,6 @@ from rag_system import build_rag_chain, ask_question
 from vector_store import get_embeddings, load_vector_store
 from llm_loader import load_llama_model
 import uuid
-import os
-import shutil
 from urllib.parse import urljoin, quote
 
 from fastapi.responses import StreamingResponse
@@ -16,17 +25,17 @@ import time
 
 app = FastAPI()
 
-# 정적 파일 서빙을 위한 설정
+# Configuration for serving static files
 os.makedirs("static/documents", exist_ok=True)
 app.mount("/static", StaticFiles(directory="static"), name="static")
 
-# 전역 객체 준비
+# Prepare global objects
 embeddings = get_embeddings(device="cpu")
 vectorstore = load_vector_store(embeddings, load_path="vector_db")
 llm = load_llama_model()
-qa_chain = build_rag_chain(llm, vectorstore, language="ko", k=7)
+qa_chain = build_rag_chain(llm, vectorstore, language="en", k=7)
 
-# 서버 URL 설정 (실제 환경에 맞게 수정 필요)
+# Server URL configuration (adjust to match your actual environment)
 BASE_URL = "http://220.124.155.35:8500"
 
 class Question(BaseModel):
@@ -37,7 +46,7 @@ def get_document_url(source_path):
         return None
     filename = os.path.basename(source_path)
     dataset_root = os.path.join(os.getcwd(), "dataset")
-    # dataset 전체 하위 폴더에서 파일명 일치하는 파일 찾기
+    # Find file matching filename in the entire dataset subdirectory
     found_path = None
     for root, dirs, files in os.walk(dataset_root):
         if filename in files:
@@ -51,13 +60,13 @@ def get_document_url(source_path):
     return urljoin(BASE_URL, f"/static/documents/{encoded_filename}")
 
 def create_download_link(url, filename):
-    return f'출처: [{filename}]({url})'
+    return f'Source: [{filename}]({url})'
 
 @app.post("/ask")
 def ask(question: Question):
     result = ask_question(qa_chain, question.question)
     
-    # 소스 문서 정보 처리
+    # Process source document information
     sources = []
     for doc in result["source_documents"]:
         source_path = doc.metadata.get('source', 'N/A')
@@ -100,7 +109,7 @@ async def openai_compatible_chat(request: Request):
     result = ask_question(qa_chain, user_input)
     answer = result['result']
     
-    # 소스 문서 정보 처리
+    # Process source document information
     sources = []
     for doc in result["source_documents"]:
         source_path = doc.metadata.get('source', 'N/A')
@@ -116,13 +125,13 @@ async def openai_compatible_chat(request: Request):
         }
         sources.append(source_info)
 
-    # 소스 정보를 한 줄씩만 출력
-    sources_md = "\n참고 문서:\n"
+    # Output source information one line at a time
+    sources_md = "\nReferences Documents:\n"
     seen = set()
     for source in sources:
         key = (source['filename'], source['document_url'])
         if source['document_url'] and source['filename'] and key not in seen:
-            sources_md += f"출처: [{source['filename']}]({source['document_url']})\n"
+            sources_md += f"Source: [{source['filename']}]({source['document_url']})\n"
             seen.add(key)
 
     final_answer = answer.split("A:")[-1].strip() if "A:" in answer else answer.strip()
@@ -143,9 +152,9 @@ async def openai_compatible_chat(request: Request):
             "model": "rag",
         })
 
-    # 스트리밍 응답을 위한 generator
+    # Generator for streaming response
     def event_stream():
-        # 답변 본문만 먼저 스트리밍
+        # Stream only the answer body first
         answer_main = answer.split("A:")[-1].strip() if "A:" in answer else answer.strip()
         for char in answer_main:
             chunk = {
@@ -161,15 +170,15 @@ async def openai_compatible_chat(request: Request):
             }
             yield f"data: {json.dumps(chunk)}\n\n"
             time.sleep(0.005)
-        # 참고 문서(다운로드 링크)는 마지막에 한 번에 붙여서 전송
-        sources_md = "\n참고 문서:\n"
+        # Send reference documents (download links) all at once at the end
+        sources_md = "\nReferences Documents:\n"
         seen = set()
         for source in sources:
             key = (source['filename'], source['document_url'])
             if source['document_url'] and source['filename'] and key not in seen:
-                sources_md += f"출처: [{source['filename']}]({source['document_url']})\n"
+                sources_md += f"Source: [{source['filename']}]({source['document_url']})\n"
                 seen.add(key)
-        if sources_md.strip() != "참고 문서:":
+        if sources_md.strip() != "References Documents:":
             chunk = {
                 "id": f"chatcmpl-{uuid.uuid4()}",
                 "object": "chat.completion.chunk",
@@ -194,4 +203,4 @@ async def openai_compatible_chat(request: Request):
         yield f"data: {json.dumps(done)}\n\n"
         return
 
-    return StreamingResponse(event_stream(), media_type="text/event-stream")
+    return StreamingResponse(event_stream(), media_type="text/event-stream")