Spaces:
Sleeping
Sleeping
| from fastapi import APIRouter, UploadFile, File, HTTPException, Depends, Header | |
| from PIL import Image | |
| import io | |
| import uuid | |
| from typing import Optional | |
| from auth_deps import get_current_user_id | |
| import shared_resources | |
| from detection_history import add_history_record, add_temp_record # Import hàm mới | |
| from core.config import BEFORE_COLLECTION, DURING_COLLECTION | |
| import torch | |
| router = APIRouter(tags=["Sightseeing Recognition"]) | |
| async def visual_search_endpoint( | |
| file: UploadFile = File(...), | |
| user_id: Optional[str] = Depends(get_current_user_id), | |
| # [NEW] Nhận temp_id từ header (Frontend gửi lên nếu có) | |
| x_temp_id: Optional[str] = Header(None, alias="X-Temp-ID") | |
| ): | |
| if shared_resources.model is None: | |
| raise HTTPException(503, "Model service is currently unavailable.") | |
| if shared_resources.db is None: | |
| raise HTTPException(503, "Database connection failed.") | |
| # 1. Read image | |
| try: | |
| image_data = await file.read() | |
| image = Image.open(io.BytesIO(image_data)).convert("RGB") | |
| except Exception: | |
| raise HTTPException(400, "Invalid image file format or corruption.") | |
| # 2. Embed & Search | |
| try: | |
| query_vector = shared_resources.embed_image(image) | |
| search_results = shared_resources.search_similar_landmark( | |
| query_vector, | |
| collection_name=DURING_COLLECTION, | |
| limit=1 | |
| ) | |
| except Exception as e: | |
| raise HTTPException(500, detail=f"Processing or search failed: {e}") | |
| if not search_results or search_results[0]["score"] < 0.9: | |
| return {"status": "not_found", "message": "Low confidence or no match."} | |
| best_match = search_results[0] | |
| landmark_id = best_match["landmark_id"] | |
| # 3. Lookup metadata | |
| before_col = shared_resources.db[BEFORE_COLLECTION] | |
| landmark_info = before_col.find_one( | |
| {"landmark_id": landmark_id}, | |
| {"_id": 0, "text_chunk": 0, "v_hybrid": 0, "combined_tags_array": 0} | |
| ) | |
| if not landmark_info: | |
| return {"status": "missing_info", "message": "Metadata not found"} | |
| # Chuẩn bị data response | |
| response_data = { | |
| "status": "success", | |
| "similarity_score": best_match["score"], | |
| "landmark_id": landmark_id, | |
| "matched_image_url": best_match.get("image_url"), | |
| "landmark_info": landmark_info, | |
| "temp_id": None # Mặc định | |
| } | |
| # 4. LOGIC LƯU LỊCH SỬ (QUAN TRỌNG) | |
| landmark_name = landmark_info.get("name", "Unknown Landmark") | |
| if user_id: | |
| # A. USER ĐÃ LOGIN -> Lưu thẳng vào User History | |
| add_history_record( | |
| user_id=user_id, | |
| landmark_data=response_data, | |
| uploaded_image_bytes=image_data, | |
| landmark_name=landmark_name | |
| ) | |
| else: | |
| # B. USER CHƯA LOGIN -> Lưu vào Temp History | |
| # Nếu Frontend chưa gửi Temp ID, Backend tự tạo một cái mới | |
| current_temp_id = x_temp_id if x_temp_id else str(uuid.uuid4()) | |
| add_temp_record( | |
| temp_id=current_temp_id, | |
| landmark_data=response_data, | |
| uploaded_image_bytes=image_data, | |
| landmark_name=landmark_name | |
| ) | |
| # Trả về temp_id để Frontend lưu lại (nếu chưa có) | |
| response_data["temp_id"] = current_temp_id | |
| response_data["message"] = "Saved to temporary history. Please login to sync." | |
| return response_data | |