Spaces:
Running
Running
File size: 9,841 Bytes
75033ed ffb5f88 75033ed ffb5f88 75033ed ffb5f88 75033ed ffb5f88 75033ed ffb5f88 75033ed ffb5f88 75033ed ffb5f88 75033ed ffb5f88 75033ed ffb5f88 75033ed ffb5f88 75033ed ffb5f88 75033ed ffb5f88 75033ed ffb5f88 75033ed ffb5f88 75033ed ffb5f88 75033ed ffb5f88 75033ed ffb5f88 75033ed ffb5f88 75033ed ffb5f88 75033ed 2ecdea6 75033ed 2ecdea6 75033ed 2ecdea6 75033ed 2ecdea6 75033ed 2ecdea6 75033ed |
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 |
"""
Hybrid Chat Endpoint: RAG + Scenario FSM
Routes between scripted scenarios and knowledge retrieval
"""
from fastapi import HTTPException
from datetime import datetime
from typing import Dict, Any
# Import scenario handlers
from scenario_handlers.price_inquiry import PriceInquiryHandler
from scenario_handlers.event_recommendation import EventRecommendationHandler
from scenario_handlers.post_event_feedback import PostEventFeedbackHandler
from scenario_handlers.exit_intent_rescue import ExitIntentRescueHandler
async def hybrid_chat_endpoint(
request, # ChatRequest
conversation_service,
intent_classifier,
embedding_service, # For handlers
qdrant_service, # For handlers
tools_service,
advanced_rag,
chat_history_collection,
hf_token,
lead_storage
):
"""
Hybrid conversational chatbot: Scenario FSM + RAG
Flow:
1. Load session & scenario state
2. Classify intent (scenario vs RAG)
3. Route:
- Scenario: Execute FSM flow with dedicated handlers
- RAG: Knowledge retrieval
- RAG+Resume: Answer question then resume scenario
4. Save state & history
"""
try:
# ===== SESSION MANAGEMENT =====
session_id = request.session_id
if not session_id:
session_id = conversation_service.create_session(
metadata={"user_agent": "api", "created_via": "hybrid_chat"},
user_id=request.user_id
)
print(f"✓ Created session: {session_id} (user: {request.user_id or 'anon'})")
else:
if not conversation_service.session_exists(session_id):
raise HTTPException(404, detail=f"Session {session_id} not found")
# ===== LOAD SCENARIO STATE =====
scenario_state = conversation_service.get_scenario_state(session_id) or {}
# ===== INTENT CLASSIFICATION =====
intent = intent_classifier.classify(request.message, scenario_state)
print(f"🎯 Intent: {intent}")
# ===== ROUTING =====
if intent.startswith("scenario:"):
# Route to dedicated scenario handler
response_data = await handle_scenario(
intent,
request.message,
session_id,
scenario_state,
embedding_service,
qdrant_service,
conversation_service,
lead_storage
)
elif intent == "rag:with_resume":
# Answer question but keep scenario active
response_data = await handle_rag_with_resume(
request,
session_id,
scenario_state,
embedding_service,
qdrant_service,
conversation_service
)
else: # rag:general
# Pure RAG query
response_data = await handle_pure_rag(
request,
session_id,
advanced_rag,
embedding_service,
qdrant_service,
tools_service,
chat_history_collection,
hf_token,
conversation_service
)
# ===== SAVE HISTORY =====
conversation_service.add_message(
session_id,
"user",
request.message,
metadata={"intent": intent}
)
conversation_service.add_message(
session_id,
"assistant",
response_data["response"],
metadata={
"mode": response_data.get("mode", "unknown"),
"context_used": response_data.get("context_used", [])[:3]
}
)
return {
"response": response_data["response"],
"session_id": session_id,
"mode": response_data.get("mode"),
"scenario_active": response_data.get("scenario_active", False),
"timestamp": datetime.utcnow().isoformat()
}
except Exception as e:
print(f"❌ Error in hybrid_chat: {str(e)}")
raise HTTPException(500, detail=f"Chat error: {str(e)}")
async def handle_scenario(
intent,
user_message,
session_id,
scenario_state,
embedding_service,
qdrant_service,
conversation_service,
lead_storage
):
"""
Handle scenario-based conversation using dedicated handlers
Replaces old scenario_engine with per-scenario handlers
"""
# Initialize all scenario handlers
handlers = {
'price_inquiry': PriceInquiryHandler(embedding_service, qdrant_service, lead_storage),
'event_recommendation': EventRecommendationHandler(embedding_service, qdrant_service, lead_storage),
'post_event_feedback': PostEventFeedbackHandler(embedding_service, qdrant_service, lead_storage),
'exit_intent_rescue': ExitIntentRescueHandler(embedding_service, qdrant_service, lead_storage)
}
if intent == "scenario:continue":
# Continue existing scenario
scenario_id = scenario_state.get("active_scenario")
if scenario_id not in handlers:
return {
"response": f"Xin lỗi, scenario '{scenario_id}' không tồn tại.",
"mode": "error",
"scenario_active": False
}
handler = handlers[scenario_id]
result = handler.next_step(
current_step=scenario_state.get("scenario_step", 1),
user_input=user_message,
scenario_data=scenario_state.get("scenario_data", {})
)
else:
# Start new scenario
scenario_type = intent.split(":", 1)[1]
if scenario_type not in handlers:
return {
"response": f"Xin lỗi, scenario '{scenario_type}' không tồn tại.",
"mode": "error",
"scenario_active": False
}
handler = handlers[scenario_type]
# Get initial_data from scenario_state (if any)
initial_data = scenario_state.get("scenario_data", {})
result = handler.start(initial_data=initial_data)
# Update scenario state
if result.get("end_scenario") or not result.get("scenario_active", True):
conversation_service.clear_scenario(session_id)
scenario_active = False
elif result.get("new_state"):
conversation_service.set_scenario_state(session_id, result["new_state"])
scenario_active = True
else:
# new_state is None → stay at same step (e.g., validation failed)
scenario_active = True
return {
"response": result.get("message", ""),
"mode": "scenario",
"scenario_active": scenario_active,
"loading_message": result.get("loading_message") # For UI
}
async def handle_rag_with_resume(
request,
session_id,
scenario_state,
embedding_service,
qdrant_service,
conversation_service
):
"""
Handle RAG query mid-scenario
Answer question properly, then remind user to continue scenario
"""
# Query RAG with proper search
context_used = []
if request.use_rag:
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,
ef=256
)
context_used = results
# Build REAL RAG response (not placeholder)
if context_used and len(context_used) > 0:
# Format top results nicely
top_result = context_used[0]
text = top_result['metadata'].get('text', '')
# Extract most relevant snippet (first 300 chars)
if text:
rag_response = text[:300].strip()
if len(text) > 300:
rag_response += "..."
else:
rag_response = "Tôi tìm thấy thông tin nhưng không thể hiển thị chi tiết."
# If multiple results, add count
if len(context_used) > 1:
rag_response += f"\n\n(Tìm thấy {len(context_used)} kết quả liên quan)"
else:
rag_response = "Xin lỗi, tôi không tìm thấy thông tin về câu hỏi này trong tài liệu."
# Add resume hint
resume_hint = "\n\n---\n💬 Vậy nha! Quay lại câu hỏi trước, bạn đã quyết định chưa?"
return {
"response": rag_response + resume_hint,
"mode": "rag_with_resume",
"scenario_active": True,
"context_used": context_used
}
async def handle_pure_rag(
request,
session_id,
advanced_rag,
embedding_service,
qdrant_service,
tools_service,
chat_history_collection,
hf_token,
conversation_service
):
"""
Handle pure RAG query (fallback to existing logic)
"""
# Import existing chat_endpoint logic
from chat_endpoint import chat_endpoint
# Call existing endpoint
result = await chat_endpoint(
request,
conversation_service,
tools_service,
advanced_rag,
embedding_service,
qdrant_service,
chat_history_collection,
hf_token
)
return {
"response": result["response"],
"mode": "rag",
"context_used": result.get("context_used", [])
}
|