File size: 6,907 Bytes
75033ed
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2ecdea6
 
 
 
 
75033ed
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2ecdea6
75033ed
 
2ecdea6
 
 
 
75033ed
 
 
2ecdea6
 
75033ed
2ecdea6
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
75033ed
 
2ecdea6
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
"""

Intent Classifier for Hybrid RAG + FSM Chatbot

Detects user intent to route between scenario flows and RAG queries

"""
from typing import Dict, Optional, List
import re


class IntentClassifier:
    """

    Classify user intent using keyword matching

    Routes to either:

    - Scenario flows (scripted conversations)

    - RAG queries (knowledge retrieval)

    """
    
    def __init__(self, scenarios_dir: str = "scenarios"):
        """

        Initialize with auto-loading triggers from scenario JSON files

        

        Args:

            scenarios_dir: Directory containing scenario JSON files

        """
        # Auto-load scenario patterns from JSON files
        self.scenario_patterns = self._load_scenario_patterns(scenarios_dir)
        
        # General question patterns (RAG)
        self.general_patterns = [
            # Location
            "ở đâu", "địa điểm", "location", "where",
            "chỗ nào", "tổ chức tại",
            
            # Time
            "mấy giờ", "khi nào", "when", "time",
            "bao giờ", "thời gian", "ngày nào",
            
            # Info
            "thông tin", "info", "information",
            "chi tiết", "details", "về",
            
            # Parking
            "đậu xe", "parking", "gửi xe",
            
            # Contact
            "liên hệ", "contact", "số điện thoại",
            
            # Events/content - NEW (Bug fix #3)
            "sự kiện", "event", "đâu", "show nào",
            "line-up", "lineup", "performer"
        ]
    
    def _load_scenario_patterns(self, scenarios_dir: str) -> dict:
        """

        Auto-load triggers from all scenario JSON files

        

        Returns:

            {"scenario_id": ["trigger1", "trigger2", ...]}

        """
        import json
        import os
        
        patterns = {}
        
        if not os.path.exists(scenarios_dir):
            print(f"⚠ Scenarios directory not found: {scenarios_dir}")
            return patterns
        
        for filename in os.listdir(scenarios_dir):
            if filename.endswith('.json'):
                filepath = os.path.join(scenarios_dir, filename)
                try:
                    with open(filepath, 'r', encoding='utf-8') as f:
                        scenario = json.load(f)
                        scenario_id = scenario.get('scenario_id')
                        triggers = scenario.get('triggers', [])
                        
                        if scenario_id and triggers:
                            patterns[scenario_id] = triggers
                            print(f"✓ Loaded triggers for: {scenario_id} ({len(triggers)} patterns)")
                except Exception as e:
                    print(f"⚠ Error loading {filename}: {e}")
        
        return patterns
    
    def classify(

        self, 

        message: str, 

        conversation_state: Optional[Dict] = None

    ) -> str:
        """

        Classify user intent with IMPROVED mid-scenario detection (Bug fix #3)

        

        Returns:

            - "scenario:{scenario_id}" - Trigger new scenario

            - "scenario:continue" - Continue active scenario

            - "rag:general" - General RAG query (no active scenario)

            - "rag:with_resume" - RAG query mid-scenario (then resume)

        """
        message_lower = message.lower().strip()
        
        # Check if user is in active scenario
        active_scenario = conversation_state.get('active_scenario') if conversation_state else None
        
        if active_scenario:
            # User is in a scenario - check if this is off-topic or continuation
            
            # IMPROVED: Detect off-topic questions better
            # Check for question words + patterns
            question_indicators = ["?", "đâu", "gì", "sao", "where", "what", "how", "when"]
            has_question = any(q in message_lower for q in question_indicators)
            
            # Check if matches general patterns
            matches_general = self._matches_any_pattern(message_lower, self.general_patterns)
            
            # Short messages with questions are likely off-topic
            word_count = len(message_lower.split())
            is_short_question = word_count <= 4 and has_question
            
            # Decision logic
            if matches_general or is_short_question:
                # User asking off-topic question → RAG with resume
                print(f"🔀 Off-topic detected: '{message}' → rag:with_resume")
                return "rag:with_resume"
            else:
                # Normal scenario continuation
                return "scenario:continue"
        
        # Not in scenario - check for scenario triggers
        for scenario_id, patterns in self.scenario_patterns.items():
            for pattern in patterns:
                if pattern.lower() in message_lower:
                    return f"scenario:{scenario_id}"
        
        # No scenario match - general RAG query
        return "rag:general"
    
    def _matches_any_pattern(self, message: str, patterns: List[str]) -> bool:
        """

        Check if message matches any pattern in list

        """
        for pattern in patterns:
            # Simple substring match
            if pattern in message:
                return True
            
            # Word boundary check
            if re.search(rf'\b{re.escape(pattern)}\b', message, re.IGNORECASE):
                return True
        
        return False
    
    def get_scenario_type(self, intent: str) -> Optional[str]:
        """

        Extract scenario type from intent string

        

        Args:

            intent: "scenario:price_inquiry" or "scenario:continue"

        

        Returns:

            "price_inquiry" or None

        """
        if not intent.startswith("scenario:"):
            return None
        
        parts = intent.split(":", 1)
        if len(parts) < 2:
            return None
        
        scenario_type = parts[1]
        if scenario_type == "continue":
            return None
        
        return scenario_type
    
    def add_scenario_pattern(self, scenario_id: str, patterns: List[str]):
        """

        Dynamically add new scenario patterns

        """
        if scenario_id in self.scenario_patterns:
            self.scenario_patterns[scenario_id].extend(patterns)
        else:
            self.scenario_patterns[scenario_id] = patterns
    
    def add_general_pattern(self, patterns: List[str]):
        """

        Dynamically add new general question patterns

        """
        self.general_patterns.extend(patterns)