Update main.py

main.py

@@ -88,7 +88,7 @@ pdf_indexer = PDFIndexer(
 )
 print("✓ PDF Indexer initialized")
 
-# Initialize Multimodal PDF Indexer (for PDFs with images)
+# Initialize Multimodal PDF Indexer
 multimodal_pdf_indexer = MultimodalPDFIndexer(
     embedding_service=embedding_service,
     qdrant_service=qdrant_service,
@@ -143,12 +143,6 @@ Quy tắc tuyệt đối:
     use_reranking: bool = True
     use_compression: bool = True
     score_threshold: float = 0.5
-    # Advanced RAG options
-    use_advanced_rag: bool = True
-    use_query_expansion: bool = True
-    use_reranking: bool = True
-    use_compression: bool = True
-    score_threshold: float = 0.5
 
 
 class ChatResponse(BaseModel):
@@ -156,7 +150,6 @@ class ChatResponse(BaseModel):
     context_used: List[Dict]
     timestamp: str
     rag_stats: Optional[Dict] = None  # Stats from advanced RAG pipeline
-    rag_stats: Optional[Dict] = None  # Stats from advanced RAG pipeline
 
 
 class AddDocumentRequest(BaseModel):
@@ -170,724 +163,274 @@ class AddDocumentResponse(BaseModel):
     message: str
 
 
-class UploadPDFResponse(BaseModel):
-    success: bool
-    document_id: str
-    filename: str
-    chunks_indexed: int
-    message: str
-
-
-class UploadPDFResponse(BaseModel):
-    success: bool
-    document_id: str
-    filename: str
-    chunks_indexed: int
-    message: str
-
-
 @app.get("/")
 async def root():
     """Health check endpoint with comprehensive API documentation"""
     return {
         "status": "running",
-        "service": "ChatbotRAG API - Advanced RAG with Multimodal Support",
-        "version": "3.0.0",
-        "service": "ChatbotRAG API - Advanced RAG with Multimodal Support",
-        "version": "3.0.0",
+        "service": "ChatbotRAG API",
+        "version": "2.0.0",
         "vector_db": "Qdrant",
-        "document_db": "MongoDB",
-        "features": {
-            "multiple_inputs": "Index up to 10 texts + 10 images per request",
-            "advanced_rag": "Query expansion, reranking, contextual compression",
-            "pdf_support": "Upload PDFs and chat about their content",
-            "multimodal_pdf": "PDFs with text and image URLs - perfect for user guides",
-            "chat_history": "Track conversation history",
-            "hybrid_search": "Text + image search with Jina CLIP v2"
-        },
-        "document_db": "MongoDB",
-        "features": {
-            "multiple_inputs": "Index up to 10 texts + 10 images per request",
-            "advanced_rag": "Query expansion, reranking, contextual compression",
-            "pdf_support": "Upload PDFs and chat about their content",
-            "multimodal_pdf": "PDFs with text and image URLs - perfect for user guides",
-            "chat_history": "Track conversation history",
-            "hybrid_search": "Text + image search with Jina CLIP v2"
-        },
+        "document_db": "MongoDB",        
         "endpoints": {
-            "indexing": {
-                "POST /index": {
-                    "description": "Index multiple texts and images (NEW: up to 10 each)",
-                    "content_type": "multipart/form-data",
-                    "body": {
-                        "id": "string (required) - Document ID (primary)",
-                        "texts": "List[string] (optional) - Up to 10 texts",
-                        "images": "List[UploadFile] (optional) - Up to 10 images",
-                        "id_use": "string (optional) - ID của SocialMedia hoặc EventCode",
-                        "id_user": "string (optional) - ID của User"
-                    },
-                    "example": "curl -X POST '/index' -F 'id=doc1' -F 'id_use=social_123' -F 'id_user=user_789' -F 'texts=Text 1' -F 'images=@img1.jpg'",
-            "indexing": {
-                "POST /index": {
-                    "description": "Index multiple texts and images (NEW: up to 10 each)",
-                    "content_type": "multipart/form-data",
-                    "body": {
-                        "id": "string (required) - Document ID (primary)",
-                        "texts": "List[string] (optional) - Up to 10 texts",
-                        "images": "List[UploadFile] (optional) - Up to 10 images",
-                        "id_use": "string (optional) - ID của SocialMedia hoặc EventCode",
-                        "id_user": "string (optional) - ID của User"
+            "chatbot_rag": {
+                "API endpoint": "https://minhvtt-ChatbotRAG.hf.space/",
+                "POST /chat": {
+                    "description": "Chat với AI sử dụng RAG (Retrieval-Augmented Generation)",
+                    "request": {
+                        "method": "POST",
+                        "content_type": "application/json",
+                        "body": {
+                            "message": "string (required) - User message/question",
+                            "use_rag": "boolean (optional, default: true) - Enable RAG context retrieval",
+                            "top_k": "integer (optional, default: 3) - Number of context documents to retrieve",
+                            "system_message": "string (optional) - Custom system prompt",
+                            "max_tokens": "integer (optional, default: 512) - Max response length",
+                            "temperature": "float (optional, default: 0.7, range: 0-1) - Creativity level",
+                            "top_p": "float (optional, default: 0.95) - Nucleus sampling",
+                            "hf_token": "string (optional) - Hugging Face token (fallback to env)"
+                        }
                     },
-                    "example": "curl -X POST '/index' -F 'id=doc1' -F 'id_use=social_123' -F 'id_user=user_789' -F 'texts=Text 1' -F 'images=@img1.jpg'",
                     "response": {
-                        "success": True,
-                        "id": "doc1",
-                        "message": "Indexed successfully with 2 texts and 1 images"
-                        "success": True,
-                        "id": "doc1",
-                        "message": "Indexed successfully with 2 texts and 1 images"
+                        "response": "string - AI generated response",
+                        "context_used": [
+                            {
+                                "id": "string - Document ID",
+                                "confidence": "float - Relevance score",
+                                "metadata": {
+                                    "text": "string - Retrieved context"
+                                }
+                            }
+                        ],
+                        "timestamp": "string - ISO 8601 timestamp"
                     },
-                    "use_cases": {
-                        "social_media_post": {
-                            "id": "post_uuid_123",
-                            "id_use": "social_media_456",
-                            "id_user": "user_789",
-                            "description": "Link post to social media account and user"
-                        },
-                        "event_post": {
-                            "id": "post_uuid_789",
-                            "id_use": "event_code_ABC123",
-                            "id_user": "user_101",
-                            "description": "Link post to event and user"
-                        }
-                    }
-                    "use_cases": {
-                        "social_media_post": {
-                            "id": "post_uuid_123",
-                            "id_use": "social_media_456",
-                            "id_user": "user_789",
-                            "description": "Link post to social media account and user"
-                        },
-                        "event_post": {
-                            "id": "post_uuid_789",
-                            "id_use": "event_code_ABC123",
-                            "id_user": "user_101",
-                            "description": "Link post to event and user"
-                        }
-                    }
-                },
-                "POST /documents": {
-                    "description": "Add text document to knowledge base",
-                    "content_type": "application/json",
-                    "body": {
-                        "text": "string (required) - Document content",
-                        "metadata": "object (optional) - Additional metadata"
+                    "example_request": {
+                        "message": "Dao có nguy hiểm không?",
+                        "use_rag": True,
+                        "top_k": 3,
+                        "temperature": 0.7
                     },
-                    "example": {
-                        "text": "How to create event: Click 'Create Event' button...",
-                        "metadata": {"category": "tutorial", "source": "user_guide"}
-                    }
-                },
-                "POST /upload-pdf": {
-                    "description": "Upload PDF file (text only)",
-                    "content_type": "multipart/form-data",
-                    "body": {
-                        "file": "UploadFile (required) - PDF file",
-                        "title": "string (optional) - Document title",
-                        "category": "string (optional) - Category",
-                        "description": "string (optional) - Description"
+                    "example_response": {
+                        "response": "Dựa trên thông tin trong database, dao được phân loại là vũ khí nguy hiểm. Dao sắc có thể gây thương tích nghiêm trọng nếu không sử dụng đúng cách. Cần tuân thủ các quy định an toàn khi sử dụng.",
+                        "context_used": [
+                            {
+                                "id": "68a3fc14c853d7621e8977b5",
+                                "confidence": 0.92,
+                                "metadata": {
+                                    "text": "Vũ khí"
+                                }
+                            },
+                            {
+                                "id": "68a3fc4cc853d7621e8977b6",
+                                "confidence": 0.85,
+                                "metadata": {
+                                    "text": "Con dao sắc"
+                                }
+                            }
+                        ],
+                        "timestamp": "2025-10-13T10:30:45.123456"
                     },
-                    "example": "curl -X POST '/upload-pdf' -F 'file=@guide.pdf' -F 'title=User Guide'"
+                    "notes": [
+                        "RAG retrieves relevant context from vector DB before generating response",
+                        "LLM uses context to provide accurate, grounded answers",
+                        "Requires HUGGINGFACE_TOKEN environment variable or hf_token in request"
+                    ]
                 },
-                "POST /upload-pdf-multimodal": {
-                    "description": "Upload PDF with text and image URLs (RECOMMENDED for user guides)",
-                    "content_type": "multipart/form-data",
-                    "features": [
-                        "Extracts text from PDF",
-                        "Detects image URLs (http://, https://)",
-                        "Supports markdown: ![alt](url)",
-                        "Supports HTML: <img src='url'>",
-                        "Links images to text chunks",
-                        "Returns images with context in chat"
-                    ],
-                    "body": {
-                        "file": "UploadFile (required) - PDF file with image URLs",
-                        "title": "string (optional) - Document title",
-                        "category": "string (optional) - e.g. 'user_guide', 'tutorial'",
-                        "description": "string (optional)"
-                    },
-                    "example": "curl -X POST '/upload-pdf-multimodal' -F 'file=@guide_with_images.pdf' -F 'category=user_guide'",
-                    "description": "Add text document to knowledge base",
-                    "content_type": "application/json",
-                    "body": {
-                        "text": "string (required) - Document content",
-                        "metadata": "object (optional) - Additional metadata"
+                "POST /documents": {
+                    "description": "Add document to knowledge base for RAG",
+                    "request": {
+                        "method": "POST",
+                        "content_type": "application/json",
+                        "body": {
+                            "text": "string (required) - Document text content",
+                            "metadata": "object (optional) - Additional metadata (source, category, etc.)"
+                        }
                     },
-                    "example": {
-                        "text": "How to create event: Click 'Create Event' button...",
-                        "metadata": {"category": "tutorial", "source": "user_guide"}
-                    }
-                },
-                "POST /upload-pdf": {
-                    "description": "Upload PDF file (text only)",
-                    "content_type": "multipart/form-data",
-                    "body": {
-                        "file": "UploadFile (required) - PDF file",
-                        "title": "string (optional) - Document title",
-                        "category": "string (optional) - Category",
-                        "description": "string (optional) - Description"
+                    "response": {
+                        "success": "boolean",
+                        "doc_id": "string - MongoDB ObjectId",
+                        "message": "string - Status message"
                     },
-                    "example": "curl -X POST '/upload-pdf' -F 'file=@guide.pdf' -F 'title=User Guide'"
-                },
-                "POST /upload-pdf-multimodal": {
-                    "description": "Upload PDF with text and image URLs (RECOMMENDED for user guides)",
-                    "content_type": "multipart/form-data",
-                    "features": [
-                        "Extracts text from PDF",
-                        "Detects image URLs (http://, https://)",
-                        "Supports markdown: ![alt](url)",
-                        "Supports HTML: <img src='url'>",
-                        "Links images to text chunks",
-                        "Returns images with context in chat"
-                    ],
-                    "body": {
-                        "file": "UploadFile (required) - PDF file with image URLs",
-                        "title": "string (optional) - Document title",
-                        "category": "string (optional) - e.g. 'user_guide', 'tutorial'",
-                        "description": "string (optional)"
+                    "example_request": {
+                        "text": "Để tạo event mới: Click nút 'Tạo Event' ở góc trên bên phải màn hình. Điền thông tin sự kiện bao gồm tên, ngày giờ, địa điểm. Click Lưu để hoàn tất.",
+                        "metadata": {
+                            "source": "user_guide.pdf",
+                            "section": "create_event",
+                            "page": 5,
+                            "category": "tutorial"
+                        }
                     },
-                    "example": "curl -X POST '/upload-pdf-multimodal' -F 'file=@guide_with_images.pdf' -F 'category=user_guide'",
-                    "response": {
+                    "example_response": {
                         "success": True,
-                        "document_id": "pdf_multimodal_20251029_150000",
-                        "chunks_indexed": 25,
-                        "message": "PDF indexed with 25 chunks and 15 images"
-                        "success": True,
-                        "document_id": "pdf_multimodal_20251029_150000",
-                        "chunks_indexed": 25,
-                        "message": "PDF indexed with 25 chunks and 15 images"
-                    },
-                    "use_case": "Perfect for user guides with screenshots, tutorials with diagrams"
-                }
-            },
-            "search": {
-                "POST /search": {
-                    "description": "Hybrid search with text and/or image",
-                    "body": {
-                        "text": "string (optional) - Query text",
-                        "image": "UploadFile (optional) - Query image",
-                        "limit": "int (default: 10)",
-                        "score_threshold": "float (optional, 0-1)",
-                        "text_weight": "float (default: 0.5)",
-                        "image_weight": "float (default: 0.5)"
-                    }
-                },
-                "POST /search/text": {
-                    "description": "Text-only search",
-                    "body": {"text": "string", "limit": "int", "score_threshold": "float"}
-                },
-                "POST /search/image": {
-                    "description": "Image-only search",
-                    "body": {"image": "UploadFile", "limit": "int", "score_threshold": "float"}
-                    "use_case": "Perfect for user guides with screenshots, tutorials with diagrams"
-                }
-            },
-            "search": {
-                "POST /search": {
-                    "description": "Hybrid search with text and/or image",
-                    "body": {
-                        "text": "string (optional) - Query text",
-                        "image": "UploadFile (optional) - Query image",
-                        "limit": "int (default: 10)",
-                        "score_threshold": "float (optional, 0-1)",
-                        "text_weight": "float (default: 0.5)",
-                        "image_weight": "float (default: 0.5)"
+                        "doc_id": "67a9876543210fedcba98765",
+                        "message": "Document added successfully with ID: 67a9876543210fedcba98765"
                     }
                 },
-                "POST /search/text": {
-                    "description": "Text-only search",
-                    "body": {"text": "string", "limit": "int", "score_threshold": "float"}
-                },
-                "POST /search/image": {
-                    "description": "Image-only search",
-                    "body": {"image": "UploadFile", "limit": "int", "score_threshold": "float"}
-                },
                 "POST /rag/search": {
-                    "description": "Search in RAG knowledge base",
-                    "body": {"query": "string", "top_k": "int (default: 5)", "score_threshold": "float (default: 0.5)"}
-                }
-            },
-            "chat": {
-                "POST /chat": {
-                    "description": "Chat với Advanced RAG (Query expansion + Reranking + Compression)",
-                    "content_type": "application/json",
-                    "body": {
-                        "message": "string (required) - User question",
-                        "use_rag": "bool (default: true) - Enable RAG retrieval",
-                        "use_advanced_rag": "bool (default: true) - Use advanced RAG pipeline (RECOMMENDED)",
-                        "use_query_expansion": "bool (default: true) - Expand query with variations",
-                        "use_reranking": "bool (default: true) - Rerank results for accuracy",
-                        "use_compression": "bool (default: true) - Compress context to relevant parts",
-                        "top_k": "int (default: 3) - Number of documents to retrieve",
-                        "score_threshold": "float (default: 0.5) - Min relevance score (0-1)",
-                        "max_tokens": "int (default: 512) - Max response tokens",
-                        "temperature": "float (default: 0.7) - Creativity (0-1)",
-                        "hf_token": "string (optional) - Hugging Face token"
-                    },
-                    "response": {
-                        "response": "string - AI answer",
-                        "context_used": "array - Retrieved documents with metadata",
-                        "timestamp": "string",
-                        "rag_stats": "object - RAG pipeline statistics (query variants, retrieval counts)"
+                    "description": "Search in knowledge base (similar to /search/text but for RAG documents)",
+                    "request": {
+                        "method": "POST",
+                        "content_type": "multipart/form-data",
+                        "body": {
+                            "query": "string (required) - Search query",
+                            "top_k": "integer (optional, default: 5) - Number of results",
+                            "score_threshold": "float (optional, default: 0.5) - Minimum relevance score"
+                        }
                     },
-                    "example_advanced": {
-                        "message": "Làm sao để upload PDF có hình ảnh?",
-                        "use_advanced_rag": True,
-                        "use_reranking": True,
-                        "top_k": 5,
-                        "score_threshold": 0.5
-                    "description": "Search in RAG knowledge base",
-                    "body": {"query": "string", "top_k": "int (default: 5)", "score_threshold": "float (default: 0.5)"}
-                }
-            },
-            "chat": {
-                "POST /chat": {
-                    "description": "Chat với Advanced RAG (Query expansion + Reranking + Compression)",
-                    "content_type": "application/json",
-                    "body": {
-                        "message": "string (required) - User question",
-                        "use_rag": "bool (default: true) - Enable RAG retrieval",
-                        "use_advanced_rag": "bool (default: true) - Use advanced RAG pipeline (RECOMMENDED)",
-                        "use_query_expansion": "bool (default: true) - Expand query with variations",
-                        "use_reranking": "bool (default: true) - Rerank results for accuracy",
-                        "use_compression": "bool (default: true) - Compress context to relevant parts",
-                        "top_k": "int (default: 3) - Number of documents to retrieve",
-                        "score_threshold": "float (default: 0.5) - Min relevance score (0-1)",
-                        "max_tokens": "int (default: 512) - Max response tokens",
-                        "temperature": "float (default: 0.7) - Creativity (0-1)",
-                        "hf_token": "string (optional) - Hugging Face token"
+                    "response": [
+                        {
+                            "id": "string",
+                            "confidence": "float",
+                            "metadata": {
+                                "text": "string",
+                                "source": "string"
+                            }
+                        }
+                    ],
+                    "example_request": {
+                        "query": "cách tạo sự kiện mới",
+                        "top_k": 3,
+                        "score_threshold": 0.6
+                    }
+                },
+                "GET /history": {
+                    "description": "Get chat conversation history",
+                    "request": {
+                        "method": "GET",
+                        "query_params": {
+                            "limit": "integer (optional, default: 10) - Number of messages",
+                            "skip": "integer (optional, default: 0) - Pagination offset"
+                        }
                     },
                     "response": {
-                        "response": "string - AI answer",
-                        "context_used": "array - Retrieved documents with metadata",
-                        "timestamp": "string",
-                        "rag_stats": "object - RAG pipeline statistics (query variants, retrieval counts)"
-                    },
-                    "example_advanced": {
-                        "message": "Làm sao để upload PDF có hình ảnh?",
-                        "use_advanced_rag": True,
-                        "use_reranking": True,
-                        "top_k": 5,
-                        "score_threshold": 0.5
-                    },
-                    "example_response_with_images": {
-                        "response": "Để upload PDF có hình ảnh, sử dụng endpoint /upload-pdf-multimodal...",
-                        "context_used": [
+                        "history": [
                             {
-                                "id": "pdf_multimodal_...._p2_c1",
-                                "confidence": 0.89,
-                                "metadata": {
-                                    "text": "Bước 1: Chuẩn bị PDF với image URLs...",
-                                    "has_images": True,
-                                    "image_urls": [
-                                        "https://example.com/screenshot1.png",
-                                        "https://example.com/diagram.jpg"
-                                    ],
-                                    "num_images": 2,
-                                    "page": 2
-                                }
+                                "user_message": "string",
+                                "assistant_response": "string",
+                                "context_used": "array",
+                                "timestamp": "string - ISO 8601"
                             }
                         ],
-                        "rag_stats": {
-                            "original_query": "Làm sao để upload PDF có hình ảnh?",
-                            "expanded_queries": ["upload PDF hình ảnh", "PDF có ảnh"],
-                            "initial_results": 10,
-                            "after_rerank": 5,
-                            "after_compression": 5
-                        }
+                        "total": "integer - Total messages count"
                     },
-                    "notes": [
-                        "Advanced RAG significantly improves answer quality",
-                        "When multimodal PDF is used, images are returned in metadata",
-                        "Requires HUGGINGFACE_TOKEN for actual LLM generation"
-                    ]
-                    "example_response_with_images": {
-                        "response": "Để upload PDF có hình ảnh, sử dụng endpoint /upload-pdf-multimodal...",
-                        "context_used": [
+                    "example_request": "GET /history?limit=5&skip=0",
+                    "example_response": {
+                        "history": [
                             {
-                                "id": "pdf_multimodal_...._p2_c1",
-                                "confidence": 0.89,
-                                "metadata": {
-                                    "text": "Bước 1: Chuẩn bị PDF với image URLs...",
-                                    "has_images": True,
-                                    "image_urls": [
-                                        "https://example.com/screenshot1.png",
-                                        "https://example.com/diagram.jpg"
-                                    ],
-                                    "num_images": 2,
-                                    "page": 2
-                                }
+                                "user_message": "Dao có nguy hiểm không?",
+                                "assistant_response": "Dao được phân loại là vũ khí...",
+                                "context_used": [],
+                                "timestamp": "2025-10-13T10:30:45.123456"
                             }
                         ],
-                        "rag_stats": {
-                            "original_query": "Làm sao để upload PDF có hình ảnh?",
-                            "expanded_queries": ["upload PDF hình ảnh", "PDF có ảnh"],
-                            "initial_results": 10,
-                            "after_rerank": 5,
-                            "after_compression": 5
+                        "total": 15
+                    }
+                },
+                "DELETE /documents/{doc_id}": {
+                    "description": "Delete document from knowledge base",
+                    "request": {
+                        "method": "DELETE",
+                        "path_params": {
+                            "doc_id": "string - MongoDB ObjectId"
                         }
                     },
-                    "notes": [
-                        "Advanced RAG significantly improves answer quality",
-                        "When multimodal PDF is used, images are returned in metadata",
-                        "Requires HUGGINGFACE_TOKEN for actual LLM generation"
-                    ]
-                },
-                "GET /history": {
-                    "description": "Get chat history",
-                    "query_params": {"limit": "int (default: 10)", "skip": "int (default: 0)"},
-                    "response": {"history": "array", "total": "int"}
-                }
-            },
-            "management": {
-                "GET /documents/pdf": {
-                    "description": "List all PDF documents",
-                    "response": {"documents": "array", "total": "int"}
-                },
-                "DELETE /documents/pdf/{document_id}": {
-                    "description": "Delete PDF and all its chunks",
-                    "response": {"success": "bool", "message": "string"}
-                },
-                "GET /document/{doc_id}": {
-                    "description": "Get document by ID",
-                    "response": {"success": "bool", "data": "object"}
-                },
-                "DELETE /delete/{doc_id}": {
-                    "description": "Delete document by ID",
-                    "response": {"success": "bool", "message": "string"}
-                },
-                "GET /stats": {
-                    "description": "Get Qdrant collection statistics",
-                    "response": {"vectors_count": "int", "segments": "int", "indexed_vectors_count": "int"}
+                    "response": {
+                        "success": "boolean",
+                        "message": "string"
+                    }
                 }
             }
         },
-        "quick_start": {
-            "1_upload_multimodal_pdf": "curl -X POST '/upload-pdf-multimodal' -F 'file=@user_guide.pdf' -F 'title=Guide'",
-            "2_verify_upload": "curl '/documents/pdf'",
-            "3_chat_with_rag": "curl -X POST '/chat' -H 'Content-Type: application/json' -d '{\"message\": \"How to...?\", \"use_advanced_rag\": true}'",
-            "4_see_images_in_context": "response['context_used'][0]['metadata']['image_urls']"
-        },
-        "use_cases": {
-            "user_guide_with_screenshots": {
-                "endpoint": "/upload-pdf-multimodal",
-                "description": "PDFs with text instructions + image URLs for visual guidance",
-                "benefits": ["Images linked to text chunks", "Chatbot returns relevant screenshots", "Perfect for step-by-step guides"]
-            },
-            "simple_text_docs": {
-                "endpoint": "/upload-pdf",
-                "description": "Simple PDFs with text only (FAQ, policies, etc.)"
-            },
-            "social_media_posts": {
-                "endpoint": "/index",
-                "description": "Index multiple posts with texts (up to 10) and images (up to 10)"
-            },
-            "complex_queries": {
-                "endpoint": "/chat",
-                "description": "Use advanced RAG for better accuracy on complex questions",
-                "settings": {"use_advanced_rag": True, "use_reranking": True, "use_compression": True}
-            }
-        },
-        "best_practices": {
-            "pdf_format": [
-                "Include image URLs in text (http://, https://)",
-                "Use markdown format: ![alt](url) or HTML: <img src='url'>",
-                "Clear structure with headings and sections",
-                "Link images close to their related text"
-            ],
-            "chat_settings": {
-                "for_accuracy": {"temperature": 0.3, "use_advanced_rag": True, "use_reranking": True},
-                "for_creativity": {"temperature": 0.8, "use_advanced_rag": False},
-                "for_factual_answers": {"temperature": 0.3, "use_compression": True, "score_threshold": 0.6}
-            },
-            "retrieval_tuning": {
-                "not_finding_info": "Lower score_threshold to 0.3-0.4, increase top_k to 7-10",
-                "too_much_context": "Increase score_threshold to 0.6-0.7, decrease top_k to 3-5",
-                "slow_responses": "Disable compression, use basic RAG, decrease top_k"
-            }
-                    "description": "Get chat history",
-                    "query_params": {"limit": "int (default: 10)", "skip": "int (default: 0)"},
-                    "response": {"history": "array", "total": "int"}
-                }
-            },
-            "management": {
-                "GET /documents/pdf": {
-                    "description": "List all PDF documents",
-                    "response": {"documents": "array", "total": "int"}
-                },
-                "DELETE /documents/pdf/{document_id}": {
-                    "description": "Delete PDF and all its chunks",
-                    "response": {"success": "bool", "message": "string"}
-                },
-                "GET /document/{doc_id}": {
-                    "description": "Get document by ID",
-                    "response": {"success": "bool", "data": "object"}
-                },
-                "DELETE /delete/{doc_id}": {
-                    "description": "Delete document by ID",
-                    "response": {"success": "bool", "message": "string"}
-                },
-                "GET /stats": {
-                    "description": "Get Qdrant collection statistics",
-                    "response": {"vectors_count": "int", "segments": "int", "indexed_vectors_count": "int"}
-                }
-            }
+        "usage_examples": {
+            "curl_chat": "curl -X POST 'http://localhost:8000/chat' -H 'Content-Type: application/json' -d '{\"message\": \"Dao có nguy hiểm không?\", \"use_rag\": true}'",
+            "python_chat": """
+import requests
+
+response = requests.post(
+    'http://localhost:8000/chat',
+    json={
+        'message': 'Nút tạo event ở đâu?',
+        'use_rag': True,
+        'top_k': 3
+    }
+)
+print(response.json()['response'])
+            """
         },
-        "quick_start": {
-            "1_upload_multimodal_pdf": "curl -X POST '/upload-pdf-multimodal' -F 'file=@user_guide.pdf' -F 'title=Guide'",
-            "2_verify_upload": "curl '/documents/pdf'",
-            "3_chat_with_rag": "curl -X POST '/chat' -H 'Content-Type: application/json' -d '{\"message\": \"How to...?\", \"use_advanced_rag\": true}'",
-            "4_see_images_in_context": "response['context_used'][0]['metadata']['image_urls']"
+        "authentication": {
+            "embeddings_apis": "No authentication required",
+            "chat_api": "Requires HUGGINGFACE_TOKEN (env variable or request body)"
         },
-        "use_cases": {
-            "user_guide_with_screenshots": {
-                "endpoint": "/upload-pdf-multimodal",
-                "description": "PDFs with text instructions + image URLs for visual guidance",
-                "benefits": ["Images linked to text chunks", "Chatbot returns relevant screenshots", "Perfect for step-by-step guides"]
-            },
-            "simple_text_docs": {
-                "endpoint": "/upload-pdf",
-                "description": "Simple PDFs with text only (FAQ, policies, etc.)"
-            },
-            "social_media_posts": {
-                "endpoint": "/index",
-                "description": "Index multiple posts with texts (up to 10) and images (up to 10)"
-            },
-            "complex_queries": {
-                "endpoint": "/chat",
-                "description": "Use advanced RAG for better accuracy on complex questions",
-                "settings": {"use_advanced_rag": True, "use_reranking": True, "use_compression": True}
-            }
+        "rate_limits": {
+            "embeddings": "No limit",
+            "chat_with_llm": "Limited by Hugging Face API (free tier: ~1000 requests/hour)"
         },
-        "best_practices": {
-            "pdf_format": [
-                "Include image URLs in text (http://, https://)",
-                "Use markdown format: ![alt](url) or HTML: <img src='url'>",
-                "Clear structure with headings and sections",
-                "Link images close to their related text"
-            ],
-            "chat_settings": {
-                "for_accuracy": {"temperature": 0.3, "use_advanced_rag": True, "use_reranking": True},
-                "for_creativity": {"temperature": 0.8, "use_advanced_rag": False},
-                "for_factual_answers": {"temperature": 0.3, "use_compression": True, "score_threshold": 0.6}
-            },
-            "retrieval_tuning": {
-                "not_finding_info": "Lower score_threshold to 0.3-0.4, increase top_k to 7-10",
-                "too_much_context": "Increase score_threshold to 0.6-0.7, decrease top_k to 3-5",
-                "slow_responses": "Disable compression, use basic RAG, decrease top_k"
-            }
+        "error_codes": {
+            "400": "Bad Request - Missing required fields or invalid input",
+            "401": "Unauthorized - Invalid Hugging Face token",
+            "404": "Not Found - Document ID not found",
+            "500": "Internal Server Error - Server or database error"
         },
         "links": {
             "docs": "http://localhost:8000/docs",
             "redoc": "http://localhost:8000/redoc",
-            "openapi": "http://localhost:8000/openapi.json",
-            "guides": {
-                "multimodal_pdf": "See MULTIMODAL_PDF_GUIDE.md",
-                "advanced_rag": "See ADVANCED_RAG_GUIDE.md",
-                "pdf_general": "See PDF_RAG_GUIDE.md",
-                "quick_start": "See QUICK_START_PDF.md"
-            }
-        },
-        "system_info": {
-            "embedding_model": "Jina CLIP v2 (multimodal)",
-            "vector_db": "Qdrant with HNSW index",
-            "document_db": "MongoDB",
-            "rag_pipeline": "Advanced RAG with query expansion, reranking, compression",
-            "pdf_parser": "pypdfium2 with URL extraction",
-            "max_inputs": "10 texts + 10 images per /index request"
-            "openapi": "http://localhost:8000/openapi.json",
-            "guides": {
-                "multimodal_pdf": "See MULTIMODAL_PDF_GUIDE.md",
-                "advanced_rag": "See ADVANCED_RAG_GUIDE.md",
-                "pdf_general": "See PDF_RAG_GUIDE.md",
-                "quick_start": "See QUICK_START_PDF.md"
-            }
-        },
-        "system_info": {
-            "embedding_model": "Jina CLIP v2 (multimodal)",
-            "vector_db": "Qdrant with HNSW index",
-            "document_db": "MongoDB",
-            "rag_pipeline": "Advanced RAG with query expansion, reranking, compression",
-            "pdf_parser": "pypdfium2 with URL extraction",
-            "max_inputs": "10 texts + 10 images per /index request"
+            "openapi": "http://localhost:8000/openapi.json"
         }
     }
 
 @app.post("/index", response_model=IndexResponse)
 async def index_data(
     id: str = Form(...),
-    texts: Optional[List[str]] = Form(None),
-    images: Optional[List[UploadFile]] = File(None),
-    id_use: Optional[str] = Form(None),
-    id_user: Optional[str] = Form(None)
-    texts: Optional[List[str]] = Form(None),
-    images: Optional[List[UploadFile]] = File(None),
-    id_use: Optional[str] = Form(None),
-    id_user: Optional[str] = Form(None)
+    text: str = Form(...),
+    image: Optional[UploadFile] = File(None)
 ):
     """
-    Index data vào vector database (hỗ trợ nhiều texts và images)
-    Index data vào vector database (hỗ trợ nhiều texts và images)
+    Index data vào vector database
 
     Body:
-    - id: Document ID (primary ID)
-    - texts: List of text contents (tiếng Việt supported) - Tối đa 10 texts
-    - images: List of image files (optional) - Tối đa 10 images
-    - id_use: ID của SocialMedia hoặc EventCode (optional)
-    - id_user: ID của User (optional)
-    - id: Document ID (primary ID)
-    - texts: List of text contents (tiếng Việt supported) - Tối đa 10 texts
-    - images: List of image files (optional) - Tối đa 10 images
-    - id_use: ID của SocialMedia hoặc EventCode (optional)
-    - id_user: ID của User (optional)
+    - id: Document ID (event ID, post ID, etc.)
+    - text: Text content (tiếng Việt supported)
+    - image: Image file (optional)
 
     Returns:
     - success: True/False
     - id: Document ID
     - message: Status message
-
-    Example:
-    ```bash
-    curl -X POST '/index' \
-      -F 'id=doc123' \
-      -F 'id_use=social_media_456' \
-      -F 'id_user=user_789' \
-      -F 'texts=Post content 1' \
-      -F 'texts=Post content 2' \
-      -F 'images=@image1.jpg'
-    ```
-
-    Example:
-    ```bash
-    curl -X POST '/index' \
-      -F 'id=doc123' \
-      -F 'id_use=social_media_456' \
-      -F 'id_user=user_789' \
-      -F 'texts=Post content 1' \
-      -F 'texts=Post content 2' \
-      -F 'images=@image1.jpg'
-    ```
     """
     try:
-        # Validation
-        if texts is None and images is None:
-            raise HTTPException(status_code=400, detail="Phải cung cấp ít nhất texts hoặc images")
-
-        if texts and len(texts) > 10:
-            raise HTTPException(status_code=400, detail="Tối đa 10 texts")
-
-        if images and len(images) > 10:
-            raise HTTPException(status_code=400, detail="Tối đa 10 images")
-
-        # Validation
-        if texts is None and images is None:
-            raise HTTPException(status_code=400, detail="Phải cung cấp ít nhất texts hoặc images")
-
-        if texts and len(texts) > 10:
-            raise HTTPException(status_code=400, detail="Tối đa 10 texts")
-
-        if images and len(images) > 10:
-            raise HTTPException(status_code=400, detail="Tối đa 10 images")
-
         # Prepare embeddings
-        text_embeddings = []
-        image_embeddings = []
-        text_embeddings = []
-        image_embeddings = []
-
-        # Encode multiple texts (tiếng Việt)
-        if texts:
-            for text in texts:
-                if text and text.strip():
-                    text_emb = embedding_service.encode_text(text)
-                    text_embeddings.append(text_emb)
-        # Encode multiple texts (tiếng Việt)
-        if texts:
-            for text in texts:
-                if text and text.strip():
-                    text_emb = embedding_service.encode_text(text)
-                    text_embeddings.append(text_emb)
-
-        # Encode multiple images
-        if images:
-            for image in images:
-                if image.filename:  # Check if image is provided
-                    image_bytes = await image.read()
-                    pil_image = Image.open(io.BytesIO(image_bytes)).convert('RGB')
-                    image_emb = embedding_service.encode_image(pil_image)
-                    image_embeddings.append(image_emb)
-        # Encode multiple images
-        if images:
-            for image in images:
-                if image.filename:  # Check if image is provided
-                    image_bytes = await image.read()
-                    pil_image = Image.open(io.BytesIO(image_bytes)).convert('RGB')
-                    image_emb = embedding_service.encode_image(pil_image)
-                    image_embeddings.append(image_emb)
-
-        # Combine embeddings
-        all_embeddings = []
-
-        if text_embeddings:
-            # Average all text embeddings
-            avg_text_embedding = np.mean(text_embeddings, axis=0)
-            all_embeddings.append(avg_text_embedding)
-
-        if image_embeddings:
-            # Average all image embeddings
-            avg_image_embedding = np.mean(image_embeddings, axis=0)
-            all_embeddings.append(avg_image_embedding)
-
-        if not all_embeddings:
-            raise HTTPException(status_code=400, detail="Không có embedding nào được tạo từ texts hoặc images")
-
-        # Final combined embedding
-        combined_embedding = np.mean(all_embeddings, axis=0)
-        all_embeddings = []
-
-        if text_embeddings:
-            # Average all text embeddings
-            avg_text_embedding = np.mean(text_embeddings, axis=0)
-            all_embeddings.append(avg_text_embedding)
+        text_embedding = None
+        image_embedding = None
 
-        if image_embeddings:
-            # Average all image embeddings
-            avg_image_embedding = np.mean(image_embeddings, axis=0)
-            all_embeddings.append(avg_image_embedding)
+        # Encode text (tiếng Việt)
+        if text and text.strip():
+            text_embedding = embedding_service.encode_text(text)
 
-        if not all_embeddings:
-            raise HTTPException(status_code=400, detail="Không có embedding nào được tạo từ texts hoặc images")
+        # Encode image nếu có
+        if image:
+            image_bytes = await image.read()
+            pil_image = Image.open(io.BytesIO(image_bytes)).convert('RGB')
+            image_embedding = embedding_service.encode_image(pil_image)
 
-        # Final combined embedding
-        combined_embedding = np.mean(all_embeddings, axis=0)
+        # Combine embeddings
+        if text_embedding is not None and image_embedding is not None:
+            # Average của text và image embeddings
+            combined_embedding = np.mean([text_embedding, image_embedding], axis=0)
+        elif text_embedding is not None:
+            combined_embedding = text_embedding
+        elif image_embedding is not None:
+            combined_embedding = image_embedding
+        else:
+            raise HTTPException(status_code=400, detail="Phải cung cấp ít nhất text hoặc image")
 
         # Normalize
         combined_embedding = combined_embedding / np.linalg.norm(combined_embedding, axis=1, keepdims=True)
 
         # Index vào Qdrant
         metadata = {
-            "texts": texts if texts else [],
-            "text_count": len(texts) if texts else 0,
-            "image_count": len(images) if images else 0,
-            "image_filenames": [img.filename for img in images] if images else [],
-            "id_use": id_use if id_use else None,  # ID của SocialMedia hoặc EventCode
-            "id_user": id_user if id_user else None  # ID của User
-            "texts": texts if texts else [],
-            "text_count": len(texts) if texts else 0,
-            "image_count": len(images) if images else 0,
-            "image_filenames": [img.filename for img in images] if images else [],
-            "id_use": id_use if id_use else None,  # ID của SocialMedia hoặc EventCode
-            "id_user": id_user if id_user else None  # ID của User
+            "text": text,
+            "has_image": image is not None,
+            "image_filename": image.filename if image else None
         }
 
         result = qdrant_service.index_data(
@@ -899,14 +442,9 @@ async def index_data(
         return IndexResponse(
             success=True,
             id=result["original_id"],  # Trả về MongoDB ObjectId
-            message=f"Đã index thành công document {result['original_id']} với {len(texts) if texts else 0} texts và {len(images) if images else 0} images (Qdrant UUID: {result['qdrant_id']})"
-            message=f"Đã index thành công document {result['original_id']} với {len(texts) if texts else 0} texts và {len(images) if images else 0} images (Qdrant UUID: {result['qdrant_id']})"
+            message=f"Đã index thành công document {result['original_id']} (Qdrant UUID: {result['qdrant_id']})"
         )
 
-    except HTTPException:
-        raise
-    except HTTPException:
-        raise
     except Exception as e:
         raise HTTPException(status_code=500, detail=f"Lỗi khi index: {str(e)}")
 
@@ -1129,8 +667,7 @@ async def get_stats():
 @app.post("/chat", response_model=ChatResponse)
 async def chat(request: ChatRequest):
     """
-    Chat endpoint với Advanced RAG
-    Chat endpoint với Advanced RAG
+    Chat endpoint với RAG
 
     Body:
     - message: User message
@@ -1140,137 +677,38 @@ async def chat(request: ChatRequest):
     - max_tokens: Max tokens for response (default: 512)
     - temperature: Temperature for generation (default: 0.7)
     - hf_token: Hugging Face token (optional, sẽ dùng env nếu không truyền)
-    - use_advanced_rag: Use advanced RAG pipeline (default: true)
-    - use_query_expansion: Enable query expansion (default: true)
-    - use_reranking: Enable reranking (default: true)
-    - use_compression: Enable context compression (default: true)
-    - score_threshold: Minimum relevance score (default: 0.5)
-    - use_advanced_rag: Use advanced RAG pipeline (default: true)
-    - use_query_expansion: Enable query expansion (default: true)
-    - use_reranking: Enable reranking (default: true)
-    - use_compression: Enable context compression (default: true)
-    - score_threshold: Minimum relevance score (default: 0.5)
 
     Returns:
     - response: Generated response
     - context_used: Retrieved context documents
     - timestamp: Response timestamp
-    - rag_stats: Statistics from RAG pipeline
-    - rag_stats: Statistics from RAG pipeline
     """
     try:
-        # ============================================
-        # CAG Layer: Check Semantic Cache First
-        # ============================================
-        cache_hit = None
-        if cag_service and request.use_rag:
-            cache_hit = cag_service.check_cache(request.message)
-            
-            if cache_hit:
-                # Cache hit! Return cached response immediately
-                return ChatResponse(
-                    response=cache_hit["response"],
-                    context_used=cache_hit["context_used"],
-                    timestamp=datetime.utcnow().isoformat(),
-                    rag_stats={
-                        **cache_hit.get("rag_stats", {}),
-                        "cache_hit": True,
-                        "cached_query": cache_hit["cached_query"],
-                        "similarity_score": cache_hit["similarity_score"],
-                        "cached_at": cache_hit["cached_at"]
-                    }
-                )
-        
-        # ============================================
-        # RAG Pipeline (if cache miss)
-        # ============================================
         # Retrieve context if RAG enabled
         context_used = []
-        rag_stats = None
-
-        rag_stats = None
-
         if request.use_rag:
-            if request.use_advanced_rag:
-                # Initialize LLM client for query expansion
-                hf_client = None
-                if request.hf_token or hf_token:
-                    hf_client = InferenceClient(token=request.hf_token or hf_token)
-                
-                # Use Advanced RAG Pipeline (Best Case 2025)
-                documents, stats = advanced_rag.hybrid_rag_pipeline(
-                    query=request.message,
-                    top_k=request.top_k,
-                    score_threshold=request.score_threshold,
-                    use_reranking=request.use_reranking,
-                    use_compression=request.use_compression,
-                    use_query_expansion=request.use_query_expansion,
-                    max_context_tokens=500,
-                    hf_client=hf_client
-                )
-
-                # Convert to dict format for compatibility
-                context_used = [
-                    {
-                        "id": doc.id,
-                        "confidence": doc.confidence,
-                        "metadata": doc.metadata
-                    }
-                    for doc in documents
-                ]
-                rag_stats = stats
-
-                # Format context using advanced RAG formatter
-                context_text = advanced_rag.format_context_for_llm(documents)
-
-            else:
-                # Use basic RAG (original implementation)
-                query_embedding = embedding_service.encode_text(request.message)
-
-                results = qdrant_service.search(
-                    query_embedding=query_embedding,
-                    limit=request.top_k,
-                    score_threshold=request.score_threshold
-                )
-                context_used = results
-
-                # Build context text (basic format)
-                context_text = "\n\nRelevant Context:\n"
-                for i, doc in enumerate(context_used, 1):
-                    doc_text = doc["metadata"].get("text", "")
-                    confidence = doc["confidence"]
-                    context_text += f"\n[{i}] (Confidence: {confidence:.2f})\n{doc_text}\n"
-                # Build context text (basic format)
-                context_text = "\n\nRelevant Context:\n"
-                for i, doc in enumerate(context_used, 1):
-                    doc_text = doc["metadata"].get("text", "")
-                    confidence = doc["confidence"]
-                    context_text += f"\n[{i}] (Confidence: {confidence:.2f})\n{doc_text}\n"
-
-        # Build system message with context
-        if request.use_rag and context_used:
-            if request.use_advanced_rag:
-                # Use advanced prompt builder
-                system_message = advanced_rag.build_rag_prompt(
-                    query=request.message,
-                    context=context_text,
-                    system_message=request.system_message
-                )
-            else:
-                # Basic prompt
-                system_message = f"{request.system_message}\n{context_text}\n\nPlease use the above context to answer the user's question when relevant."
-        # Build system message with context
-        if request.use_rag and context_used:
-            if request.use_advanced_rag:
-                # Use advanced prompt builder
-                system_message = advanced_rag.build_rag_prompt(
-                    query=request.message,
-                    context=context_text,
-                    system_message=request.system_message
-                )
-            else:
-                # Basic prompt
-                system_message = f"{request.system_message}\n{context_text}\n\nPlease use the above context to answer the user's question when relevant."
+            # Generate query embedding
+            query_embedding = embedding_service.encode_text(request.message)
+            
+            # Search in Qdrant
+            results = qdrant_service.search(
+                query_embedding=query_embedding,
+                limit=request.top_k,
+                score_threshold=0.5
+            )
+            context_used = results
+
+        # Build context text
+        context_text = ""
+        if context_used:
+            context_text = "\n\nRelevant Context:\n"
+            for i, doc in enumerate(context_used, 1):
+                doc_text = doc["metadata"].get("text", "")
+                confidence = doc["confidence"]
+                context_text += f"\n[{i}] (Confidence: {confidence:.2f})\n{doc_text}\n"
+
+            # Add context to system message
+            system_message = f"{request.system_message}\n{context_text}\n\nPlease use the above context to answer the user's question when relevant."
         else:
             system_message = request.system_message
 
@@ -1330,28 +768,11 @@ Example:
             "timestamp": datetime.utcnow()
         }
         chat_history_collection.insert_one(chat_data)
-        
-        # ============================================
-        # CAG: Save to Cache (if RAG was used)
-        # ============================================
-        if cag_service and request.use_rag and context_used and response:
-            try:
-                cag_service.save_to_cache(
-                    query=request.message,
-                    response=response,
-                    context_used=context_used,
-                    rag_stats=rag_stats
-                )
-            except Exception as cache_error:
-                print(f"Warning: Failed to save to cache: {cache_error}")
 
         return ChatResponse(
             response=response,
             context_used=context_used,
-            timestamp=datetime.utcnow().isoformat(),
-            rag_stats=rag_stats
-            timestamp=datetime.utcnow().isoformat(),
-            rag_stats=rag_stats
+            timestamp=datetime.utcnow().isoformat()
         )
 
     except Exception as e:
@@ -1511,484 +932,6 @@ async def delete_document_from_kb(doc_id: str):
         raise HTTPException(status_code=500, detail=f"Error: {str(e)}")
 
 
-@app.post("/upload-pdf", response_model=UploadPDFResponse)
-async def upload_pdf(
-    file: UploadFile = File(...),
-    document_id: Optional[str] = Form(None),
-    title: Optional[str] = Form(None),
-    description: Optional[str] = Form(None),
-    category: Optional[str] = Form(None)
-):
-    """
-    Upload and index PDF file into knowledge base
-
-    Body (multipart/form-data):
-    - file: PDF file (required)
-    - document_id: Custom document ID (optional, auto-generated if not provided)
-    - title: Document title (optional)
-    - description: Document description (optional)
-    - category: Document category (optional, e.g., "user_guide", "faq")
-
-    Returns:
-    - success: True/False
-    - document_id: Document ID
-    - filename: Original filename
-    - chunks_indexed: Number of chunks created
-    - message: Status message
-
-    Example:
-    ```bash
-    curl -X POST "http://localhost:8000/upload-pdf" \
-      -F "file=@user_guide.pdf" \
-      -F "title=Hướng dẫn sử dụng ChatbotRAG" \
-      -F "category=user_guide"
-    ```
-    """
-    try:
-        # Validate file type
-        if not file.filename.endswith('.pdf'):
-            raise HTTPException(status_code=400, detail="Only PDF files are allowed")
-
-        # Generate document ID if not provided
-        if not document_id:
-            from datetime import datetime
-            timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
-            document_id = f"pdf_{timestamp}"
-
-        # Read PDF bytes
-        pdf_bytes = await file.read()
-
-        # Prepare metadata
-        metadata = {}
-        if title:
-            metadata['title'] = title
-        if description:
-            metadata['description'] = description
-        if category:
-            metadata['category'] = category
-
-        # Index PDF
-        result = pdf_indexer.index_pdf_bytes(
-            pdf_bytes=pdf_bytes,
-            document_id=document_id,
-            filename=file.filename,
-            document_metadata=metadata
-        )
-
-        return UploadPDFResponse(
-            success=True,
-            document_id=result['document_id'],
-            filename=result['filename'],
-            chunks_indexed=result['chunks_indexed'],
-            message=f"PDF '{file.filename}' đã được index thành công với {result['chunks_indexed']} chunks"
-        )
-
-    except HTTPException:
-        raise
-    except Exception as e:
-        raise HTTPException(status_code=500, detail=f"Error uploading PDF: {str(e)}")
-
-
-@app.get("/documents/pdf")
-async def list_pdf_documents():
-    """
-    List all PDF documents in knowledge base
-
-    Returns:
-    - documents: List of PDF documents with metadata
-    """
-    try:
-        docs = list(documents_collection.find(
-            {"type": "pdf"},
-            {"_id": 0}
-        ))
-        return {"documents": docs, "total": len(docs)}
-    except Exception as e:
-        raise HTTPException(status_code=500, detail=f"Error: {str(e)}")
-
-
-@app.delete("/documents/pdf/{document_id}")
-async def delete_pdf_document(document_id: str):
-    """
-    Delete PDF document and all its chunks from knowledge base
-
-    Args:
-    - document_id: Document ID
-
-    Returns:
-    - success: True/False
-    - message: Status message
-    """
-    try:
-        # Get document info
-        doc = documents_collection.find_one({"document_id": document_id, "type": "pdf"})
-
-        if not doc:
-            raise HTTPException(status_code=404, detail=f"PDF document {document_id} not found")
-
-        # Delete all chunks from Qdrant
-        chunk_ids = doc.get('chunk_ids', [])
-        for chunk_id in chunk_ids:
-            try:
-                qdrant_service.delete_by_id(chunk_id)
-            except:
-                pass  # Chunk might already be deleted
-
-        # Delete from MongoDB
-        documents_collection.delete_one({"document_id": document_id})
-
-        return {
-            "success": True,
-            "message": f"PDF document {document_id} and {len(chunk_ids)} chunks deleted"
-        }
-
-    except HTTPException:
-        raise
-    except Exception as e:
-        raise HTTPException(status_code=500, detail=f"Error: {str(e)}")
-
-
-@app.post("/upload-pdf-multimodal", response_model=UploadPDFResponse)
-async def upload_pdf_multimodal(
-    file: UploadFile = File(...),
-    document_id: Optional[str] = Form(None),
-    title: Optional[str] = Form(None),
-    description: Optional[str] = Form(None),
-    category: Optional[str] = Form(None)
-):
-    """
-    Upload PDF with text and image URLs (for user guides with screenshots)
-
-    This endpoint is optimized for PDFs containing:
-    - Text instructions
-    - Image URLs (http://... or https://...)
-    - Markdown images: ![alt](url)
-    - HTML images: <img src="url">
-
-    The system will:
-    1. Extract text from PDF
-    2. Detect all image URLs in the text
-    3. Link images to their corresponding text chunks
-    4. Store image URLs in metadata
-    5. Return images along with text during chat
-
-    Body (multipart/form-data):
-    - file: PDF file (required)
-    - document_id: Custom document ID (optional, auto-generated if not provided)
-    - title: Document title (optional)
-    - description: Document description (optional)
-    - category: Document category (optional, e.g., "user_guide", "tutorial")
-
-    Returns:
-    - success: True/False
-    - document_id: Document ID
-    - filename: Original filename
-    - chunks_indexed: Number of chunks created
-    - message: Status message (includes image count)
-
-    Example:
-    ```bash
-    curl -X POST "http://localhost:8000/upload-pdf-multimodal" \
-      -F "file=@user_guide_with_images.pdf" \
-      -F "title=Hướng dẫn có ảnh minh họa" \
-      -F "category=user_guide"
-    ```
-
-    Example Response:
-    ```json
-    {
-      "success": true,
-      "document_id": "pdf_20251029_150000",
-      "filename": "user_guide_with_images.pdf",
-      "chunks_indexed": 25,
-      "message": "PDF 'user_guide_with_images.pdf' indexed with 25 chunks and 15 images"
-    }
-    ```
-    """
-    try:
-        # Validate file type
-        if not file.filename.endswith('.pdf'):
-            raise HTTPException(status_code=400, detail="Only PDF files are allowed")
-
-        # Generate document ID if not provided
-        if not document_id:
-            from datetime import datetime
-            timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
-            document_id = f"pdf_multimodal_{timestamp}"
-
-        # Read PDF bytes
-        pdf_bytes = await file.read()
-
-        # Prepare metadata
-        metadata = {'type': 'multimodal'}
-        if title:
-            metadata['title'] = title
-        if description:
-            metadata['description'] = description
-        if category:
-            metadata['category'] = category
-
-        # Index PDF with multimodal parser
-        result = multimodal_pdf_indexer.index_pdf_bytes(
-            pdf_bytes=pdf_bytes,
-            document_id=document_id,
-            filename=file.filename,
-            document_metadata=metadata
-        )
-
-        return UploadPDFResponse(
-            success=True,
-            document_id=result['document_id'],
-            filename=result['filename'],
-            chunks_indexed=result['chunks_indexed'],
-            message=f"PDF '{file.filename}' indexed successfully with {result['chunks_indexed']} chunks and {result.get('images_found', 0)} images"
-        )
-
-    except HTTPException:
-        raise
-    except Exception as e:
-        raise HTTPException(status_code=500, detail=f"Error uploading multimodal PDF: {str(e)}")
-
-
-@app.post("/upload-pdf", response_model=UploadPDFResponse)
-async def upload_pdf(
-    file: UploadFile = File(...),
-    document_id: Optional[str] = Form(None),
-    title: Optional[str] = Form(None),
-    description: Optional[str] = Form(None),
-    category: Optional[str] = Form(None)
-):
-    """
-    Upload and index PDF file into knowledge base
-
-    Body (multipart/form-data):
-    - file: PDF file (required)
-    - document_id: Custom document ID (optional, auto-generated if not provided)
-    - title: Document title (optional)
-    - description: Document description (optional)
-    - category: Document category (optional, e.g., "user_guide", "faq")
-
-    Returns:
-    - success: True/False
-    - document_id: Document ID
-    - filename: Original filename
-    - chunks_indexed: Number of chunks created
-    - message: Status message
-
-    Example:
-    ```bash
-    curl -X POST "http://localhost:8000/upload-pdf" \
-      -F "file=@user_guide.pdf" \
-      -F "title=Hướng dẫn sử dụng ChatbotRAG" \
-      -F "category=user_guide"
-    ```
-    """
-    try:
-        # Validate file type
-        if not file.filename.endswith('.pdf'):
-            raise HTTPException(status_code=400, detail="Only PDF files are allowed")
-
-        # Generate document ID if not provided
-        if not document_id:
-            from datetime import datetime
-            timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
-            document_id = f"pdf_{timestamp}"
-
-        # Read PDF bytes
-        pdf_bytes = await file.read()
-
-        # Prepare metadata
-        metadata = {}
-        if title:
-            metadata['title'] = title
-        if description:
-            metadata['description'] = description
-        if category:
-            metadata['category'] = category
-
-        # Index PDF
-        result = pdf_indexer.index_pdf_bytes(
-            pdf_bytes=pdf_bytes,
-            document_id=document_id,
-            filename=file.filename,
-            document_metadata=metadata
-        )
-
-        return UploadPDFResponse(
-            success=True,
-            document_id=result['document_id'],
-            filename=result['filename'],
-            chunks_indexed=result['chunks_indexed'],
-            message=f"PDF '{file.filename}' đã được index thành công với {result['chunks_indexed']} chunks"
-        )
-
-    except HTTPException:
-        raise
-    except Exception as e:
-        raise HTTPException(status_code=500, detail=f"Error uploading PDF: {str(e)}")
-
-
-@app.get("/documents/pdf")
-async def list_pdf_documents():
-    """
-    List all PDF documents in knowledge base
-
-    Returns:
-    - documents: List of PDF documents with metadata
-    """
-    try:
-        docs = list(documents_collection.find(
-            {"type": "pdf"},
-            {"_id": 0}
-        ))
-        return {"documents": docs, "total": len(docs)}
-    except Exception as e:
-        raise HTTPException(status_code=500, detail=f"Error: {str(e)}")
-
-
-@app.delete("/documents/pdf/{document_id}")
-async def delete_pdf_document(document_id: str):
-    """
-    Delete PDF document and all its chunks from knowledge base
-
-    Args:
-    - document_id: Document ID
-
-    Returns:
-    - success: True/False
-    - message: Status message
-    """
-    try:
-        # Get document info
-        doc = documents_collection.find_one({"document_id": document_id, "type": "pdf"})
-
-        if not doc:
-            raise HTTPException(status_code=404, detail=f"PDF document {document_id} not found")
-
-        # Delete all chunks from Qdrant
-        chunk_ids = doc.get('chunk_ids', [])
-        for chunk_id in chunk_ids:
-            try:
-                qdrant_service.delete_by_id(chunk_id)
-            except:
-                pass  # Chunk might already be deleted
-
-        # Delete from MongoDB
-        documents_collection.delete_one({"document_id": document_id})
-
-        return {
-            "success": True,
-            "message": f"PDF document {document_id} and {len(chunk_ids)} chunks deleted"
-        }
-
-    except HTTPException:
-        raise
-    except Exception as e:
-        raise HTTPException(status_code=500, detail=f"Error: {str(e)}")
-
-
-@app.post("/upload-pdf-multimodal", response_model=UploadPDFResponse)
-async def upload_pdf_multimodal(
-    file: UploadFile = File(...),
-    document_id: Optional[str] = Form(None),
-    title: Optional[str] = Form(None),
-    description: Optional[str] = Form(None),
-    category: Optional[str] = Form(None)
-):
-    """
-    Upload PDF with text and image URLs (for user guides with screenshots)
-
-    This endpoint is optimized for PDFs containing:
-    - Text instructions
-    - Image URLs (http://... or https://...)
-    - Markdown images: ![alt](url)
-    - HTML images: <img src="url">
-
-    The system will:
-    1. Extract text from PDF
-    2. Detect all image URLs in the text
-    3. Link images to their corresponding text chunks
-    4. Store image URLs in metadata
-    5. Return images along with text during chat
-
-    Body (multipart/form-data):
-    - file: PDF file (required)
-    - document_id: Custom document ID (optional, auto-generated if not provided)
-    - title: Document title (optional)
-    - description: Document description (optional)
-    - category: Document category (optional, e.g., "user_guide", "tutorial")
-
-    Returns:
-    - success: True/False
-    - document_id: Document ID
-    - filename: Original filename
-    - chunks_indexed: Number of chunks created
-    - message: Status message (includes image count)
-
-    Example:
-    ```bash
-    curl -X POST "http://localhost:8000/upload-pdf-multimodal" \
-      -F "file=@user_guide_with_images.pdf" \
-      -F "title=Hướng dẫn có ảnh minh họa" \
-      -F "category=user_guide"
-    ```
-
-    Example Response:
-    ```json
-    {
-      "success": true,
-      "document_id": "pdf_20251029_150000",
-      "filename": "user_guide_with_images.pdf",
-      "chunks_indexed": 25,
-      "message": "PDF 'user_guide_with_images.pdf' indexed with 25 chunks and 15 images"
-    }
-    ```
-    """
-    try:
-        # Validate file type
-        if not file.filename.endswith('.pdf'):
-            raise HTTPException(status_code=400, detail="Only PDF files are allowed")
-
-        # Generate document ID if not provided
-        if not document_id:
-            from datetime import datetime
-            timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
-            document_id = f"pdf_multimodal_{timestamp}"
-
-        # Read PDF bytes
-        pdf_bytes = await file.read()
-
-        # Prepare metadata
-        metadata = {'type': 'multimodal'}
-        if title:
-            metadata['title'] = title
-        if description:
-            metadata['description'] = description
-        if category:
-            metadata['category'] = category
-
-        # Index PDF with multimodal parser
-        result = multimodal_pdf_indexer.index_pdf_bytes(
-            pdf_bytes=pdf_bytes,
-            document_id=document_id,
-            filename=file.filename,
-            document_metadata=metadata
-        )
-
-        return UploadPDFResponse(
-            success=True,
-            document_id=result['document_id'],
-            filename=result['filename'],
-            chunks_indexed=result['chunks_indexed'],
-            message=f"PDF '{file.filename}' indexed successfully with {result['chunks_indexed']} chunks and {result.get('images_found', 0)} images"
-        )
-
-    except HTTPException:
-        raise
-    except Exception as e:
-        raise HTTPException(status_code=500, detail=f"Error uploading multimodal PDF: {str(e)}")
-
-
 if __name__ == "__main__":
     import uvicorn
     uvicorn.run(