File size: 6,436 Bytes
aea1ea5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
import gradio as gr
import os
import torch
from transformers import AutoProcessor, AutoModel
from langchain_community.embeddings import HuggingFaceEmbeddings
from langchain_community.vectorstores import FAISS
from langchain.chains import RetrievalQA
from langchain_community.llms import HuggingFacePipeline
from PIL import Image
from langchain_community.document_loaders import PyPDFLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter

class MultimodalRAG:
    def __init__(self, pdf_path=None):
        self.processor = AutoProcessor.from_pretrained("openai/clip-vit-base-patch32")
        self.vision_model = AutoModel.from_pretrained("openai/clip-vit-base-patch32")
        self.text_embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2")
        self.pdf_path = pdf_path
        self.documents = []
        self.vector_store = None
        self.retriever = None
        self.qa_chain = None
        
        try:
            self.llm = HuggingFacePipeline.from_model_id(
                model_id="google/flan-t5-large",
                task="text2text-generation",
                model_kwargs={"temperature": 0.7, "max_length": 512}
            )
        except Exception as e:
            print(f"Error loading flan-t5 model: {e}")
            from langchain.llms import OpenAI
            self.llm = OpenAI(temperature=0.7)
            
        if pdf_path and os.path.exists(pdf_path):
            self.load_pdf(pdf_path)

    def load_pdf(self, pdf_path):
        if not os.path.exists(pdf_path):
            raise FileNotFoundError(f"PDF file not found: {pdf_path}")

        loader = PyPDFLoader(pdf_path)
        self.documents = loader.load()

        text_splitter = RecursiveCharacterTextSplitter(
            chunk_size=1000,
            chunk_overlap=200
        )
        self.documents = text_splitter.split_documents(self.documents)
        
        self.vector_store = FAISS.from_documents(self.documents, self.text_embeddings)
        
        self.retriever = self.vector_store.as_retriever(search_kwargs={"k": 2})
        
        self.qa_chain = RetrievalQA.from_chain_type(
            llm=self.llm,
            chain_type="stuff",
            retriever=self.retriever,
            return_source_documents=True
        )
        
        return f"Successfully loaded and processed PDF: {pdf_path}"

    def process_image(self, image_path):
        if not os.path.exists(image_path):
            print(f"Warning: Image path {image_path} does not exist")
            return None

        image = Image.open(image_path)
        inputs = self.processor(images=image, return_tensors="pt")
        with torch.no_grad():
            image_features = self.vision_model.get_image_features(**inputs)
        return image_features

    def generate_image_description(self, image_features):
        return "a photo"

    def retrieve_related_documents(self, query_text, image_path=None):
        if image_path:
            image_features = self.process_image(image_path)

            if image_features is not None:
                image_query = self.generate_image_description(image_features)

                enhanced_query = f"{query_text} {image_query}"
            else:
                enhanced_query = query_text
        else:
            enhanced_query = query_text

        docs = self.retriever.get_relevant_documents(enhanced_query)
        return docs

    def answer_query(self, query_text, image_path=None):
        if not self.vector_store or not self.qa_chain:
            return "Please upload a PDF document first."
            
        if image_path:
            docs = self.retrieve_related_documents(query_text, image_path)
        else:
            docs = self.retrieve_related_documents(query_text)

        result = self.qa_chain({"query": query_text})
        
        answer = result["result"]
        sources = [doc.page_content[:1000] + "..." for doc in result["source_documents"]]
        
        return answer, sources

rag_system = MultimodalRAG()

def upload_pdf(pdf_file):
    if pdf_file is None:
        return "No file uploaded"
    
    file_path = pdf_file.name
    try:
        result = rag_system.load_pdf(file_path)
        return result
    except Exception as e:
        return f"Error processing PDF: {str(e)}"

def save_image(image):
    if image is None:
        return None
    
    temp_path = "temp_image.jpg"
    image.save(temp_path)
    return temp_path

def process_query(query, pdf_file, image=None):
    if not query.strip():
        return "Please enter a question", []
    
    if pdf_file is None:
        return "Please upload a PDF document first", []
    
    image_path = None
    if image is not None:
        image_path = save_image(image)
    
    try:
        answer, sources = rag_system.answer_query(query, image_path)
        if image_path and os.path.exists(image_path):
            os.remove(image_path)
        return answer, sources
    except Exception as e:
        if image_path and os.path.exists(image_path):
            os.remove(image_path)
        return f"Error processing query: {str(e)}", []

# Create Gradio interface
with gr.Blocks(title="Multimodal RAG System") as demo:
    gr.Markdown("# Multimodal RAG System")
    gr.Markdown("Upload a PDF document and ask questions about it. You can also add an image for multimodal context.")
    
    with gr.Row():
        with gr.Column(scale=1):
            pdf_input = gr.File(label="Upload PDF Document")
            upload_button = gr.Button("Process PDF")
            status_output = gr.Textbox(label="Status")
            
            upload_button.click(
                fn=upload_pdf,
                inputs=[pdf_input],
                outputs=[status_output]
            )
        
        with gr.Column(scale=2):
            image_input = gr.Image(label="Optional: Upload an Image", type="pil")
            query_input = gr.Textbox(label="Ask a question")
            submit_button = gr.Button("Submit Question")
            
            answer_output = gr.Textbox(label="Answer")
            sources_output = gr.JSON(label="Sources")
            
            submit_button.click(
                fn=process_query,
                inputs=[query_input, pdf_input, image_input],
                outputs=[answer_output, sources_output]
            )

if __name__ == "__main__":
    demo.launch(share=True, server_name="0.0.0.0")