Spaces:
Sleeping
Sleeping
File size: 9,458 Bytes
c82da72 |
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 |
from gtts import gTTS
import io
import gradio as gr
from datetime import datetime, timedelta
import json
class MemoryHandler:
def __init__(self):
self.conversation_history = []
self.max_history = 5 # Keep last 5 interactions
self.context_timeout = timedelta(minutes=2) # Context expires after 2 minutes
self.last_interaction_time = None
self.partial_info = {
'project_number': None,
'project_name': None,
'amount': None,
'reason': None,
'timestamp': None
}
self.confidence_scores = {
'project_number': 0.0,
'project_name': 0.0,
'amount': 0.0,
'reason': 0.0
}
def add_interaction(self, text: str, extracted_info: dict = None) -> None:
"""
Add a new interaction to the conversation history and update partial information
Args:
text: The text from the voice/text input
extracted_info: Dictionary containing extracted request details
"""
current_time = datetime.now()
# Check if we should clear context due to timeout
if self.last_interaction_time and \
(current_time - self.last_interaction_time) > self.context_timeout:
self.clear_partial_info()
# Update conversation history
if text:
# Add timestamp to conversation history
self.conversation_history.append({
'text': text,
'timestamp': current_time.isoformat(),
'extracted_info': extracted_info
})
if len(self.conversation_history) > self.max_history:
self.conversation_history.pop(0)
# Update partial information if provided
if extracted_info:
self._update_partial_info(extracted_info, current_time)
self.last_interaction_time = current_time
def _update_partial_info(self, extracted_info: dict, current_time: datetime) -> None:
"""
Update partial information with confidence scoring
"""
for key in self.partial_info:
if key in extracted_info and extracted_info[key]:
new_value = extracted_info[key]
current_value = self.partial_info[key]
# Update if empty or higher confidence
if (current_value is None or
extracted_info.get(f'{key}_confidence', 0.5) >
self.confidence_scores.get(key, 0)):
self.partial_info[key] = new_value
self.confidence_scores[key] = extracted_info.get(f'{key}_confidence', 0.5)
self.partial_info['timestamp'] = current_time
def get_context(self) -> str:
"""
Get the current conversation context including partial information
"""
# Start with the most recent conversation history
context_parts = []
# Add conversation history with timestamps
for entry in self.conversation_history:
timestamp = datetime.fromisoformat(entry['timestamp']).strftime('%H:%M:%S')
context_parts.append(f"[{timestamp}] {entry['text']}")
context = " ".join(context_parts)
# Add partial information to context if available
partial_context = []
for key, value in self.partial_info.items():
if value and key != 'timestamp':
confidence = self.confidence_scores.get(key, 0)
partial_context.append(f"{key}: {value} (confidence: {confidence:.2f})")
if partial_context:
context += "\nPartial information: " + ", ".join(partial_context)
return context
def get_partial_info(self) -> dict:
"""Get current partial information with confidence scores"""
info = {k: v for k, v in self.partial_info.items()
if k != 'timestamp' and v is not None}
info['confidence_scores'] = self.confidence_scores
return info
def merge_partial_info(self, new_info: dict) -> None:
"""
Merge new information with existing partial info based on confidence scores
"""
for key in self.partial_info:
if key in new_info and new_info[key] is not None:
new_confidence = new_info.get(f'{key}_confidence', 0.5)
if (self.partial_info[key] is None or
new_confidence > self.confidence_scores.get(key, 0)):
self.partial_info[key] = new_info[key]
self.confidence_scores[key] = new_confidence
def clear_partial_info(self) -> None:
"""Clear partial information and confidence scores"""
self.partial_info = {
'project_number': None,
'project_name': None,
'amount': None,
'reason': None,
'timestamp': None
}
self.confidence_scores = {
'project_number': 0.0,
'project_name': 0.0,
'amount': 0.0,
'reason': 0.0
}
def clear_memory(self) -> None:
"""Clear all conversation history and partial information"""
self.conversation_history = []
self.clear_partial_info()
self.last_interaction_time = None
return "Memory cleared!"
def get_missing_fields(self) -> list:
"""Get list of missing required fields with confidence thresholds"""
missing = []
confidence_threshold = 0.5 # Minimum confidence required
for field in ['project_number', 'project_name', 'amount', 'reason']:
if (self.partial_info.get(field) is None or
self.confidence_scores.get(field, 0) < confidence_threshold):
missing.append(field)
return missing
def text_to_speech(self, text: str) -> tuple[str, str]:
"""Convert text to speech and return audio path"""
try:
tts = gTTS(text=text, lang='en')
audio_path = "temp_audio.mp3"
tts.save(audio_path)
return audio_path, None
except Exception as e:
return None, f"Error generating audio: {str(e)}"
def create_confirmation_audio(self, project_number: str, project_name: str,
amount: float, reason: str) -> tuple[str, str]:
"""Create confirmation message audio with confidence information"""
confidence_info = "\n".join([
f"{field}: {self.confidence_scores.get(field, 0):.2f} confidence"
for field in ['project_number', 'project_name', 'amount', 'reason']
])
confirmation_text = (
f"You are going to add request money for project ID: {project_number}, "
f"Project name: {project_name}, request amount: {amount}, "
f"reason: {reason}.\n\nConfidence scores:\n{confidence_info}\n"
f"Are you sure you want to proceed?"
)
return self.text_to_speech(confirmation_text)
def get_prompt_for_missing_info(self) -> str:
"""Generate a prompt for missing information with confidence scores"""
missing = self.get_missing_fields()
if not missing:
return "All required information has been provided with sufficient confidence."
current_info = self.get_partial_info()
prompt = "Current information:\n"
# Show current information with confidence scores
for key, value in current_info.items():
if key != 'confidence_scores' and value is not None:
confidence = self.confidence_scores.get(key, 0)
prompt += f"- {key}: {value} (confidence: {confidence:.2f})\n"
prompt += "\nPlease provide or clarify the following information:\n"
for field in missing:
current_confidence = self.confidence_scores.get(field, 0)
if current_confidence > 0:
prompt += f"- {field} (current confidence: {current_confidence:.2f}, needs improvement)\n"
else:
prompt += f"- {field} (missing)\n"
return prompt
def to_json(self) -> str:
"""Serialize the memory state to JSON"""
return json.dumps({
'conversation_history': self.conversation_history,
'partial_info': self.partial_info,
'confidence_scores': self.confidence_scores,
'last_interaction_time': self.last_interaction_time.isoformat() if self.last_interaction_time else None
})
def from_json(self, json_str: str) -> None:
"""Restore memory state from JSON"""
data = json.loads(json_str)
self.conversation_history = data['conversation_history']
self.partial_info = data['partial_info']
self.confidence_scores = data['confidence_scores']
self.last_interaction_time = (datetime.fromisoformat(data['last_interaction_time'])
if data['last_interaction_time'] else None) |