Hugging Face's logo

Spaces:
NaikPriyank
/
ConvoTrack
Sleeping

App Files Community
ConvoTrack
File size: 11,292 Bytes 
3f1425d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
 
# import streamlit as st
# import json
# import torch
# from transformers import AutoTokenizer, AutoModel
# import faiss
# import google.generativeai as genai
# from flashrank.Ranker import Ranker, RerankRequest

# # Configure Google Generative AI API Key
# genai.configure(api_key="AIzaSyArG3gnpZHnzi10mMSnyOMhzYJBeAZEJUs")  # Replace with your API key

# # Load and preprocess the uploaded file
# def load_and_preprocess(uploaded_file):
#     data = json.load(uploaded_file)
#     passages = [f"Speaker: {item['speaker']}. Text: {item['text']}" 
#                 for item in data if item["text"].strip()]
#     return data, passages

# # Load embedding model
# def load_model(model_name="BAAI/bge-m3"):
#     tokenizer = AutoTokenizer.from_pretrained(model_name)
#     model = AutoModel.from_pretrained(model_name)
#     return tokenizer, model

# # Generate embeddings
# def generate_embeddings(passages, tokenizer, model, batch_size=10, device="cuda" if torch.cuda.is_available() else "cpu"):
#     model.to(device)
#     embeddings = []
#     for i in range(0, len(passages), batch_size):
#         batch = passages[i:i + batch_size]
#         inputs = tokenizer(batch, return_tensors="pt", padding=True, truncation=True, max_length=512).to(device)
#         with torch.no_grad():
#             outputs = model(**inputs).last_hidden_state.mean(dim=1) 
#         embeddings.append(outputs.cpu())
#     embeddings = torch.cat(embeddings, dim=0) 
#     return embeddings.numpy()

# # Store embeddings in FAISS
# def store_in_faiss(embeddings):
#     dimension = embeddings.shape[1]
#     index = faiss.IndexFlatL2(dimension)  
#     index.add(embeddings) 
#     return index

# # Retrieve top-k passages
# def retrieve_top_k(query, tokenizer, model, faiss_index, passages, k=5, device="cuda" if torch.cuda.is_available() else "cpu"):
#     model.to(device)
#     inputs = tokenizer([query], return_tensors="pt", padding=True, truncation=True, max_length=512).to(device)
#     with torch.no_grad():
#         query_embedding = model(**inputs).last_hidden_state.mean(dim=1).cpu().numpy()
#     distances, indices = faiss_index.search(query_embedding, k)
#     retrieved_passages = [passages[i] for i in indices[0]]
#     return retrieved_passages

# # Rerank passages using FlashRank Ranker
# def rerank_passages(query, passages):
#     formatted_passages = [{"text": passage} for passage in passages]
#     ranker = Ranker(model_name="rank-T5-flan", cache_dir="/my_cache_dir")  # Adjust cache directory as needed
#     rerank_request = RerankRequest(query=query, passages=formatted_passages)
#     results = ranker.rerank(rerank_request)
#     return results

# # Generate a response using Gemini 1.5 Flash
# def generate_response(reranked_passages, query):
#     context = " ".join([passage["text"] for passage in reranked_passages])
#     input_text = f"Context: {context}\n\nQuestion: {query}\n\nAnswer:"
#     model = genai.GenerativeModel("gemini-1.5-flash")
#     response = model.generate_content(input_text)
#     return response.text

# # Streamlit app
# def main():
#     st.set_page_config(page_title="Chatbot with Document Upload", layout="wide")
#     st.title("📄 Chatbot for Minutes of Meeting")
    
#     # Initialize session state
#     if "chat_history" not in st.session_state:
#         st.session_state.chat_history = []
#     if "faiss_index" not in st.session_state:
#         st.session_state.faiss_index = None
#     if "passages" not in st.session_state:
#         st.session_state.passages = None
#     if "tokenizer" not in st.session_state or "model" not in st.session_state:
#         st.session_state.tokenizer, st.session_state.model = load_model()

#     # File uploader
#     uploaded_file = st.file_uploader("Upload a JSON file for processing", type=["json"])
#     if uploaded_file:
#         st.write("Processing the file...")
#         data, passages = load_and_preprocess(uploaded_file)
#         st.session_state.passages = passages
        
#         # Generate embeddings and store in FAISS
#         tokenizer, model = st.session_state.tokenizer, st.session_state.model
#         embeddings = generate_embeddings(passages, tokenizer, model)
#         st.session_state.faiss_index = store_in_faiss(embeddings)
#         st.success("File processed and embeddings generated successfully!")

#     # Chat interface
#     if st.session_state.faiss_index:
#         st.header("Ask a Question")
#         user_query = st.text_input("Type your question here:")
#         if user_query:
#             # Retrieve and rerank passages
#             top_k_passages = retrieve_top_k(user_query, st.session_state.tokenizer, st.session_state.model, st.session_state.faiss_index, st.session_state.passages)
#             reranked_passages = rerank_passages(user_query, top_k_passages)
            
#             # Generate response
#             response = generate_response(reranked_passages, user_query)
            
#             # Display response
#             st.markdown(f"**Question:** {user_query}")
#             st.markdown(f"**Answer:** {response}")
            
#             # Update chat history
#             st.session_state.chat_history.append({"question": user_query, "answer": response})

#     # Chat history
#     if st.session_state.chat_history:
#         st.header("Chat History")
#         for chat in st.session_state.chat_history:
#             st.markdown(f"**Q:** {chat['question']}")
#             st.markdown(f"**A:** {chat['answer']}")

# # Run the app
# if __name__ == "__main__":
#     main()

import streamlit as st
from streamlit_chat import message  
import json
import torch
from transformers import AutoTokenizer, AutoModel
import faiss
import google.generativeai as genai
from flashrank.Ranker import Ranker, RerankRequest
from langchain.memory import ConversationBufferMemory
from pydantic import BaseModel,ConfigDict


genai.configure(api_key="AIzaSyArG3gnpZHnzi10mMSnyOMhzYJBeAZEJUs")  

class CustomMemory(ConversationBufferMemory):
    model_config = ConfigDict(arbitrary_types_allowed=True)

def load_and_preprocess(uploaded_file):
    data = json.load(uploaded_file)
    passages = [f"Speaker: {item['speaker']}. Text: {item['text']}" 
                for item in data if item["text"].strip()]
    return data, passages


def load_model(model_name="BAAI/bge-m3"):
    tokenizer = AutoTokenizer.from_pretrained(model_name)
    model = AutoModel.from_pretrained(model_name)
    return tokenizer, model


def generate_embeddings(passages, tokenizer, model, batch_size=10, device="cuda" if torch.cuda.is_available() else "cpu"):
    model.to(device)
    embeddings = []
    for i in range(0, len(passages), batch_size):
        batch = passages[i:i + batch_size]
        inputs = tokenizer(batch, return_tensors="pt", padding=True, truncation=True, max_length=512).to(device)
        with torch.no_grad():
            outputs = model(**inputs).last_hidden_state.mean(dim=1) 
        embeddings.append(outputs.cpu())
    embeddings = torch.cat(embeddings, dim=0) 
    return embeddings.numpy()


def store_in_faiss(embeddings):
    dimension = embeddings.shape[1]
    index = faiss.IndexFlatL2(dimension)  
    index.add(embeddings) 
    return index


def retrieve_top_k(query, tokenizer, model, faiss_index, passages, k=5, device="cuda" if torch.cuda.is_available() else "cpu"):
    model.to(device)
    inputs = tokenizer([query], return_tensors="pt", padding=True, truncation=True, max_length=512).to(device)
    with torch.no_grad():
        query_embedding = model(**inputs).last_hidden_state.mean(dim=1).cpu().numpy()
    distances, indices = faiss_index.search(query_embedding, k)
    retrieved_passages = [passages[i] for i in indices[0]]
    return retrieved_passages


def rerank_passages(query, passages):
    formatted_passages = [{"text": passage} for passage in passages]
    ranker = Ranker(model_name="rank-T5-flan", cache_dir="/my_cache_dir")  # Adjust cache directory as needed
    rerank_request = RerankRequest(query=query, passages=formatted_passages)
    results = ranker.rerank(rerank_request)
    return results


def generate_response(context, query):
    input_text = f"Context: {context}\n\nQuestion: {query}\n\nAnswer:"
    model = genai.GenerativeModel("gemini-1.5-flash")
    response = model.generate_content(input_text)
    return response.text


def handle_userinput(user_question):
    
    top_k_passages = retrieve_top_k(user_question, st.session_state.tokenizer, st.session_state.model, st.session_state.faiss_index, st.session_state.passages)
    reranked_passages = rerank_passages(user_question, top_k_passages)

    
    context = " ".join([passage["text"] for passage in reranked_passages])

   
    response = generate_response(context, user_question)

    
    st.session_state.memory.chat_memory.add_user_message(user_question)
    st.session_state.memory.chat_memory.add_ai_message(response)

    return response


def main():
    st.set_page_config(page_title="Chatbot with MoM Document Upload", layout="wide")
    st.title("📄 Chatbot for Minutes of Meeting ")

    
    if "memory" not in st.session_state:
        st.session_state.memory = CustomMemory(memory_key='chat_history', return_messages=True)
    if "faiss_index" not in st.session_state:
        st.session_state.faiss_index = None
    if "passages" not in st.session_state:
        st.session_state.passages = None
    if "tokenizer" not in st.session_state or "model" not in st.session_state:
        st.session_state.tokenizer, st.session_state.model = load_model()

    
    uploaded_file = st.file_uploader("Upload a JSON file for processing", type=["json"])
    if uploaded_file:
        st.write("Processing the file...")
        data, passages = load_and_preprocess(uploaded_file)
        st.session_state.passages = passages

        
        tokenizer, model = st.session_state.tokenizer, st.session_state.model
        embeddings = generate_embeddings(passages, tokenizer, model)
        st.session_state.faiss_index = store_in_faiss(embeddings)
        st.success("File processed and embeddings generated successfully!")

    
    if st.session_state.faiss_index:
        st.header("Ask a Question")
        user_query = st.text_input("Type your question here:")
        if user_query:
            response = handle_userinput(user_query)

            
            if "chat_history_ui" not in st.session_state:
                st.session_state.chat_history_ui = []  
            
            st.session_state.chat_history_ui.append({"role": "user", "content": user_query})
            st.session_state.chat_history_ui.append({"role": "bot", "content": response})

    
    if "chat_history_ui" in st.session_state:
        for i,chat in enumerate(st.session_state.chat_history_ui):
            if chat["role"] == "user":
                message(chat["content"], is_user=True,key=f"user_{i}")
            else:
                message(chat["content"], is_user=False,key=f"bot_{i}")


if __name__ == "__main__":
    main()