File size: 12,770 Bytes
5065491
 
 
 
 
 
 
 
 
 
 
 
 
b971859
 
 
 
 
5065491
 
 
b971859
 
 
5065491
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
b971859
 
 
5065491
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4ddb074
5065491
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4ddb074
5065491
 
 
 
 
4ddb074
5065491
 
 
 
b971859
 
 
5065491
 
 
 
 
 
 
 
 
4ddb074
5065491
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
b971859
5065491
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
b971859
 
 
 
5065491
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
"""
Enhanced Conversational WebSocket Handler
Based on friend's implementation with session memory, personalization, and better conversational flow
"""
from fastapi import WebSocket, WebSocketDisconnect
from langchain_core.messages import HumanMessage, SystemMessage, AIMessage
import logging
import json
import asyncio
import uuid
from typing import Dict, Any, List
from session_service import session_service
from conversational_audio_service import conversational_audio_service
from enhanced_search_service import enhanced_search_service
# Temporarily disabled due to protobuf issues
# from llm_service import create_graph, create_basic_graph
from hybrid_llm_service import HybridLLMService
from langchain_core.messages import HumanMessage, AIMessage, SystemMessage

logger = logging.getLogger("conversational_voicebot")

# Initialize hybrid LLM service
hybrid_llm_service = HybridLLMService()

async def handle_conversational_websocket(websocket: WebSocket):
    """
    Enhanced conversational websocket handler similar to friend's implementation
    """
    await websocket.accept()
    logger.info("πŸ”Œ Conversational WebSocket client connected.")

    # Initialize conversation variables
    initial_data = await websocket.receive_json()
    messages = []
    conversation_history = []
    
    # Check if user authentication is provided
    flag = "user_id" in initial_data
    if flag:
        thread_id = initial_data.get("user_id")
        
        # Get user session context and preferences
        conversation_context = await session_service.get_conversation_context(thread_id)
        user_preferences = await session_service.get_user_preferences(thread_id)
        
        # Temporarily disabled due to protobuf issues
        # graph = await create_graph(kb_tool=True, mcp_config=None)
        graph = None
        
        # Set conversational audio service voice based on preferences
        voice_id = user_preferences.get('voice_id', 'en-IN-isha')
        conversational_audio_service.set_default_voice(voice_id)
        
        language_code = user_preferences.get('language', 'english')
        lang_map = {
            'english': 'en',
            'hindi': 'hi',
            'hinglish': 'en'
        }
        lang_code = lang_map.get(language_code, 'en')
        
        config = {
            "configurable": {
                "thread_id": thread_id,
                "knowledge_base": "pension_docs",
            }
        }
        
        # Enhanced conversational system prompt with personalization
        base_prompt = """You are a warm, friendly, and knowledgeable Rajasthan Pension Assistant. Your responses should be:

        - Conversational and natural, as if speaking to a trusted friend
        - Concise and informative - aim for 1-3 sentences unless more detail is specifically requested
        - Clear and easy to understand when spoken aloud
        - Professional yet personable and approachable
        - Avoid overly complex jargon or long lists that are hard to follow in audio format

        When responding about pension policies:
        - Use a warm, reassuring tone appropriate for financial guidance
        - Speak in a natural rhythm suitable for text-to-speech
        - Break complex information into digestible, conversational chunks
        - Ask clarifying questions to better understand their specific situation
        - Remember this is voice interaction - structure responses to be easily understood when heard
        - Reference specific pension documents when relevant
        - If uncertain, clearly state limitations and suggest authoritative sources

        Keep responses short and conversational. Don't use abbreviations or complex numerical content in your responses.
        Focus on being helpful, accurate, and easy to understand through voice."""

        # Add conversation context if available
        if conversation_context:
            system_message = f"{base_prompt}\n\nPersonalization context: {conversation_context}"
        else:
            system_message = base_prompt

        messages.append(SystemMessage(content=system_message))
        
    else:
        # Anonymous user
        # Temporarily disabled due to protobuf issues
        # graph = create_basic_graph()
        graph = None
        thread_id = str(uuid.uuid4())
        voice_id = "en-IN-isha"
        lang_code = "en"
        config = {"configurable": {"thread_id": thread_id}}

    # Generate personalized greeting
    if flag:
        greeting_response = await session_service.generate_personalized_greeting(thread_id, messages)
    else:
        greeting_response = "Hello! I'm your Rajasthan Pension Assistant. I'm here to help you navigate pension policies, calculations, and retirement planning. What pension question can I help you with today?"
    
    # Add greeting to conversation
    messages.append(AIMessage(content=greeting_response))
    conversation_history.append({
        'type': 'assistant',
        'content': greeting_response,
        'timestamp': asyncio.get_event_loop().time()
    })
    
    # Generate and send greeting audio
    try:
        greeting_audio = await conversational_audio_service.synthesize_speech(greeting_response, voice_id)
        await websocket.send_json({"type": "connection_successful"})
        if greeting_audio:
            await websocket.send_bytes(greeting_audio)
    except Exception as e:
        logger.error(f"❌ Error sending greeting: {e}")
        await websocket.send_json({"type": "connection_successful"})

    try:
        while True:
            try:
                # Handle different message types
                data = await websocket.receive_json()
                
                if data.get("type") == "end_call":
                    logger.info("πŸ“ž Call ended by client")
                    await websocket.close()
                    break

                # Get language preference
                lang = data.get("lang", "english").lower()
                
                # Receive audio data
                audio_bytes = await websocket.receive_bytes()

                # --- Enhanced ASR with Groq ---
                if flag:
                    transcription = await conversational_audio_service.transcribe_audio(
                        audio_bytes, language=lang_code
                    )
                else:
                    transcription = await conversational_audio_service.transcribe_audio(audio_bytes)

                if not transcription or not transcription.strip():
                    logger.warning("⚠️ Empty transcription received")
                    continue

                # Send transcription to client
                await websocket.send_json(
                    {"type": "transcription", "text": transcription}
                )
                
                # Add to conversation history
                messages.append(HumanMessage(content=transcription))
                conversation_history.append({
                    'type': 'user',
                    'content': transcription,
                    'timestamp': asyncio.get_event_loop().time()
                })

                # --- Enhanced Document Search ---
                search_results = None
                try:
                    search_results = await enhanced_search_service.enhanced_pension_search(transcription, limit=3)
                    logger.info(f"πŸ” Found {len(search_results) if search_results else 0} relevant documents")
                except Exception as search_error:
                    logger.warning(f"⚠️ Document search failed: {search_error}")

                # --- Enhanced LLM Response with Context ---
                try:
                    if search_results and len(search_results) > 0:
                        # Enrich the message with search context
                        context_message = f"User query: {transcription}\n\nRelevant pension documents found:\n"
                        for i, doc in enumerate(search_results[:2], 1):
                            content_preview = doc.get('content', '')[:300]
                            context_message += f"\n{i}. {doc.get('source', 'Document')}: {content_preview}...\n"
                        
                        context_message += f"\nBased on the above pension documents, provide a helpful and conversational response to: {transcription}"
                        
                        # Replace user message with enriched version for LLM
                        messages[-1] = HumanMessage(content=context_message)
                    
                    # Get LLM response using hybrid service
                    # Extract the user question from messages
                    user_question = messages[-1].content if messages else transcription
                    llm_response = await hybrid_llm_service.generate_response(user_question)
                    
                    # Send LLM response to client
                    await websocket.send_json(
                        {"type": "llm_response", "text": llm_response}
                    )
                    
                    # Add to conversation
                    messages.append(AIMessage(content=llm_response))
                    conversation_history.append({
                        'type': 'assistant',
                        'content': llm_response,
                        'timestamp': asyncio.get_event_loop().time()
                    })
                    
                    # --- Enhanced TTS with Murf ---
                    try:
                        if flag:
                            audio_stream = await conversational_audio_service.synthesize_speech(
                                llm_response, voice_id=voice_id
                            )
                        else:
                            audio_stream = await conversational_audio_service.synthesize_speech(llm_response)
                        
                        await websocket.send_json({"type": "tts_start"})
                        if audio_stream:
                            await websocket.send_bytes(audio_stream)
                        await websocket.send_json({"type": "tts_end"})
                        
                    except Exception as tts_error:
                        logger.error(f"❌ TTS failed: {tts_error}")
                        await websocket.send_json({"type": "tts_error", "message": "Audio generation failed"})

                except Exception as llm_error:
                    logger.error(f"❌ LLM processing failed: {llm_error}")
                    error_response = "I apologize, but I encountered an issue processing your question. Could you please try rephrasing it?"
                    
                    await websocket.send_json(
                        {"type": "llm_response", "text": error_response}
                    )
                    
                    # Try to generate error audio
                    try:
                        error_audio = await conversational_audio_service.synthesize_speech(error_response)
                        await websocket.send_json({"type": "tts_start"})
                        if error_audio:
                            await websocket.send_bytes(error_audio)
                        await websocket.send_json({"type": "tts_end"})
                    except:
                        pass

            except WebSocketDisconnect:
                logger.info("πŸ”Œ WebSocket disconnected.")
                break
            except Exception as e:
                logger.exception(f"❌ Error during conversation: {e}")
                try:
                    await websocket.send_json({"error": str(e)})
                except:
                    pass
                break

    finally:
        # Session summary generation (like friend's bot)
        if flag and len(conversation_history) > 2:
            try:
                logger.info("πŸ’Ύ Generating session summary...")
                
                # Generate session summary
                summary = await session_service.generate_session_summary(messages, thread_id)
                
                # Store the session summary
                await session_service.store_session_summary(
                    thread_id,
                    summary,
                    conversation_history
                )
                
                logger.info(f"βœ… Session summary stored for user {thread_id}")

            except Exception as e:
                logger.exception(f"❌ Error storing session summary: {e}")
        else:
            logger.info("πŸ”„ Session ended - no summary needed")

        logger.info(f"πŸ”„ Conversational session {thread_id} ended")