File size: 11,211 Bytes
b479e5d
38b607b
 
36698b9
38b607b
 
 
 
 
 
 
 
 
 
 
 
b479e5d
0badbc4
fd15582
38b607b
 
b479e5d
36698b9
38b607b
 
b479e5d
38b607b
 
 
 
b479e5d
38b607b
b479e5d
 
 
38b607b
 
 
 
 
 
 
b479e5d
 
 
 
 
 
 
 
 
38b607b
 
b479e5d
 
 
38b607b
b479e5d
 
38b607b
 
 
 
b479e5d
 
38b607b
 
 
 
 
 
 
 
 
 
 
 
 
 
b479e5d
38b607b
 
 
 
 
 
 
48f37ba
38b607b
 
 
36698b9
 
b479e5d
36698b9
 
38b607b
36698b9
b479e5d
38b607b
 
b479e5d
38b607b
48f37ba
38b607b
 
 
 
 
48f37ba
38b607b
 
 
 
 
 
 
 
 
36698b9
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
38b607b
36698b9
 
38b607b
 
36698b9
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
38b607b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
b479e5d
38b607b
 
 
 
 
b479e5d
38b607b
 
 
 
 
 
 
 
 
 
 
b479e5d
 
542022c
b479e5d
38b607b
40fa7be
 
38b607b
 
 
 
 
 
 
b479e5d
38b607b
 
 
 
 
 
 
 
 
 
 
 
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
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
import os
import re
import uuid
import time  # Add this
import tempfile
import numpy as np
import gradio as gr
import chardet
import fitz  # PyMuPDF
import docx
import gtts
from pptx import Presentation
from typing import TypedDict, List
from langchain_community.tools import DuckDuckGoSearchRun
from langchain_core.messages import BaseMessage, HumanMessage, AIMessage, SystemMessage
from langgraph.graph import StateGraph, END
from langchain_groq import ChatGroq
from langchain_community.embeddings import HuggingFaceEmbeddings
from langchain_community.vectorstores import Chroma
from langchain_text_splitters import RecursiveCharacterTextSplitter
from langchain_core.documents import Document


# --- 1. INITIALIZATION & CORE TOOLS ---
groq_api_key = os.getenv("GROQ_API_KEY")

chat_model = ChatGroq(model_name="llama-3.3-70b-versatile", api_key=groq_api_key)
web_search_tool = DuckDuckGoSearchRun()
embedding_model = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2")
vectorstore = Chroma(embedding_function=embedding_model, persist_directory="chroma_db")

# --- 2. HELPER FUNCTIONS ---


def clean_response(response):
    """Remove <think>...</think> blocks and common markdown artifacts."""
    # Remove think tags and their content (greedily, case-insensitive)
    cleaned = re.sub(r"<think>.*?(?:</think>|$)", "", response, flags=re.DOTALL | re.IGNORECASE)
    # Remove stray closing tags and markdown symbols
    cleaned = re.sub(r"</?think>|\*\*|\*|\[|\]|#", "", cleaned)
    return cleaned.strip()
    #return cleaned_text.strip()


def retrieve_documents(query):
    results = vectorstore.similarity_search(query, k=3)
    return [doc.page_content for doc in results]

def speech_playback(text):
    try:
        unique_id = str(uuid.uuid4())
        audio_file = f"/content/output_audio_{unique_id}.mp3"
        tts = gtts.gTTS(text[:500], lang='en')
        tts.save(audio_file)
        return audio_file
    except Exception as e:
        print(f"TTS error: {e}")
        return None

# --- 3. DOCUMENT INGESTION FUNCTION ---
def extract_and_store_document(file_path: str):
    text = ""
    file_ext = os.path.splitext(file_path)[1].lower()

    try:
        if file_ext == ".pdf":
            doc = fitz.open(file_path)
            for page in doc:
                text += page.get_text()
            doc.close()
        elif file_ext == ".docx":
            doc = docx.Document(file_path)
            text = "\n".join([para.text for para in doc.paragraphs])
        elif file_ext == ".pptx":
            prs = Presentation(file_path)
            for slide in prs.slides:
                for shape in slide.shapes:
                    if hasattr(shape, "text"):
                        text += shape.text + "\n"
        else:
            with open(file_path, 'rb') as f:
                raw_data = f.read()
                encoding = chardet.detect(raw_data)['encoding'] or 'utf-8'
                text = raw_data.decode(encoding, errors='ignore')
        
        if not text.strip():
            return False

        splitter = RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=50)
        chunks = splitter.split_text(text)
        documents = [Document(page_content=chunk, metadata={"source": os.path.basename(file_path)}) for chunk in chunks]
        
        # Chroma auto-persists in version 0.4.x+
        vectorstore.add_documents(documents)
        # REMOVE THIS LINE: vectorstore.persist()  # Delete line 93
        
        return True
  
    except Exception as e:
        print(f"Error processing {file_path}: {e}")
        return False

# --- 4. REFRAG MULTI-AGENT LOGIC (LangGraph) ---

class AgentState(TypedDict):
    messages: List[BaseMessage]
    context: str
    decision: str 
    source: str

def sensing_node(state: AgentState):
    user_query = state["messages"][-1].content
    relevant_docs = retrieve_documents(user_query) 
    context = "\n".join(relevant_docs) if relevant_docs else ""
    
    prompt = f"Docs: {context}\nQuery: {user_query}\nIf docs answer this, reply 'RAG'. Else reply 'WEB'."
    decision = chat_model.invoke([HumanMessage(content=prompt)]).content.strip().upper()
    return {"context": context, "decision": "RAG" if "RAG" in decision else "WEB"}


#Alternative: Better Approach - Add Fallback Search Strategy
#Add this function for more robust searching:

def safe_web_search_with_fallback(query: str):
    """Web search with multiple fallback strategies"""
    global last_web_search_time
    
    strategies = [
        # Strategy 1: Direct search
        lambda: web_search_tool.run(query),
        # Strategy 2: Search with simplified query
        lambda: web_search_tool.run(query.split("?")[0] if "?" in query else query),
        # Strategy 3: Search with keywords only
        lambda: web_search_tool.run(' '.join(query.split()[:10]))
    ]
    
    for i, strategy in enumerate(strategies):
        try:
            # Rate limiting check
            current_time = time.time()
            if current_time - last_web_search_time < 5:  # 5 second cooldown
                time.sleep(5 - (current_time - last_web_search_time))
            
            result = strategy()
            last_web_search_time = time.time()
            
            if result and len(result) > 50:  # Valid result
                return result[:2000]  # Truncate
                
        except Exception as e:
            if i == len(strategies) - 1:  # Last strategy failed
                return f"Web search unavailable. Error: {str(e)[:100]}"
            continue
    
    return "Web search temporarily unavailable."

# Add global variable for rate limiting
last_web_search_time = 0
WEB_SEARCH_COOLDOWN = 10  # 10 seconds between web searches

def expansion_node(state: AgentState):
    global last_web_search_time
    
    if state["decision"] == "WEB":
        user_query = state["messages"][-1].content
        web_data = safe_web_search_with_fallback(user_query)
    
                return {
            "context": f"WEB INFO: {web_data}\nLOCAL: {state['context']}", 
            "source": "Web + Local Documents"
        }
    
    return {"source": "Local Documents Only"}

    
        # Implement rate limiting
        current_time = time.time()
        time_since_last = current_time - last_web_search_time
        
        # If we searched recently, wait or skip web search
        if time_since_last < WEB_SEARCH_COOLDOWN:
            # Option 1: Skip web search and use local docs only
            # return {"context": state['context'], "source": "Local Documents Only (Rate limited)"}
            
            # Option 2: Wait and then search (for demo)
            wait_time = WEB_SEARCH_COOLDOWN - time_since_last
            time.sleep(wait_time)
        
        try:
            web_data = web_search_tool.run(user_query)
            last_web_search_time = time.time()  # Update timestamp
            
            # Truncate web data to avoid context overflow
            if len(web_data) > 1500:
                web_data = web_data[:1500] + "..."
                
            return {
                "context": f"WEB SEARCH RESULTS: {web_data}\nLOCAL DOCUMENTS: {state['context']}", 
                "source": "Web Search + Local Documents"
            }
        except Exception as e:
            # If web search fails, use local docs with explanation
            error_msg = str(e)
            if "Ratelimit" in error_msg:
                return {
                    "context": state['context'], 
                    "source": "Local Documents Only (Search rate limit reached)"
                }
            else:
                return {
                    "context": state['context'], 
                    "source": f"Local Documents Only (Search error: {error_msg[:100]})"
                }
    
    return {"source": "Local Documents Only"}

def generation_node(state: AgentState):
    system_msg = f"You are a Tutor AI. Use this context: {state['context']}"
    response = chat_model.invoke([SystemMessage(content=system_msg)] + state["messages"])
    cleaned = clean_response(response.content)
    return {"messages": [AIMessage(content=f"{cleaned}\n\n*(Verified via: {state['source']})*")]}

workflow = StateGraph(AgentState)
workflow.add_node("sense", sensing_node)
workflow.add_node("expand", expansion_node)
workflow.add_node("generate", generation_node)
workflow.set_entry_point("sense")
workflow.add_edge("sense", "expand")
workflow.add_edge("expand", "generate")
workflow.add_edge("generate", END)
app_agent = workflow.compile()

# --- 5. GRADIO APP WITH MANUAL AUDIO ---

# Store last assistant response globally (simple approach for demo)
last_assistant_response = ""

def chat_handler(user_input, chat_history):
    global last_assistant_response
    if not user_input: 
        return chat_history, "", None
    
    inputs = {"messages": [HumanMessage(content=user_input)], "context": "", "decision": "", "source": ""}
    result = app_agent.invoke(inputs)
    final_msg = result["messages"][-1].content
    
    chat_history.append({"role": "user", "content": user_input})
    chat_history.append({"role": "assistant", "content": final_msg})
    
    # Save clean text for later TTS (without source note)
    last_assistant_response = final_msg.split("*(Verified")[0].strip()
    
    # Return chat history and clear audio (no autoplay)
    return chat_history, "", None

def generate_audio():
    global last_assistant_response
    if not last_assistant_response:
        return None
    return speech_playback(last_assistant_response)

def upload_file(file):
    if file is None:
        return "โŒ No file uploaded."
    try:
        success = extract_and_store_document(file.name)
        if success:
            return f"โœ… **{os.path.basename(file.name)}** successfully parsed and added to knowledge base!"
        else:
            return f"โš ๏ธ Failed to extract text from **{os.path.basename(file.name)}**."
    except Exception as e:
        return f"โŒ Error: {str(e)}"

with gr.Blocks() as demo:
    gr.Markdown("# ๐ŸŽ“ REFRAG Multi-Agent Tutor")
    
    with gr.Tab("AI Chatbot"):
        #chatbot = gr.Chatbot(type="messages", height=400)
        chatbot = gr.Chatbot(value=[], height=400)
        with gr.Row():
            msg = gr.Textbox(placeholder="Ask your tutor...", scale=4)
            submit = gr.Button("Send", variant="primary")
        # Manual audio control
        with gr.Row():
            play_audio_btn = gr.Button("๐Ÿ”Š Play Audio Response", variant="secondary")
            audio_out = gr.Audio(label="Audio Response", autoplay=False)  # autoplay=False
        
        # Chat submission
        submit.click(chat_handler, [msg, chatbot], [chatbot, msg, audio_out])
        msg.submit(chat_handler, [msg, chatbot], [chatbot, msg, audio_out])
        # Manual audio generation
        play_audio_btn.click(generate_audio, None, audio_out)

    with gr.Tab("Upload Notes"):
        file_input = gr.File(label="Upload PDF / DOCX / PPTX / TXT", file_types=[".pdf", ".docx", ".pptx", ".txt"])
        upload_status = gr.Markdown()
        file_input.change(upload_file, file_input, upload_status)

demo.launch(share=True, debug=True)