Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
@@ -11,10 +11,6 @@ import torch
|
|
11 |
import numpy as np
|
12 |
import networkx as nx
|
13 |
from collections import Counter
|
14 |
-
import asyncio
|
15 |
-
import edge_tts
|
16 |
-
import speech_recognition as sr
|
17 |
-
import random
|
18 |
|
19 |
@dataclass
|
20 |
class ChatMessage:
|
@@ -25,14 +21,14 @@ class ChatMessage:
|
|
25 |
return {"role": self.role, "content": self.content}
|
26 |
|
27 |
class XylariaChat:
|
28 |
-
def
|
29 |
self.hf_token = os.getenv("HF_TOKEN")
|
30 |
if not self.hf_token:
|
31 |
raise ValueError("HuggingFace token not found in environment variables")
|
32 |
|
33 |
self.client = InferenceClient(
|
34 |
-
model="Qwen/
|
35 |
-
|
36 |
)
|
37 |
|
38 |
self.image_api_url = "https://api-inference.huggingface.co/models/Salesforce/blip-image-captioning-large"
|
@@ -53,7 +49,7 @@ class XylariaChat:
|
|
53 |
"bias_detection": 0.0,
|
54 |
"strategy_adjustment": ""
|
55 |
}
|
56 |
-
|
57 |
self.internal_state = {
|
58 |
"emotions": {
|
59 |
"valence": 0.5,
|
@@ -82,7 +78,7 @@ class XylariaChat:
|
|
82 |
]
|
83 |
|
84 |
self.system_prompt = """You are a helpful and harmless assistant. You are Xylaria developed by Sk Md Saad Amin. You should think step-by-step """
|
85 |
-
|
86 |
self.causal_rules_db = {
|
87 |
"rain": ["wet roads", "flooding"],
|
88 |
"fire": ["heat", "smoke"],
|
@@ -96,11 +92,8 @@ class XylariaChat:
|
|
96 |
"democracy": "government by the people",
|
97 |
"photosynthesis": "process used by plants to convert light to energy"
|
98 |
}
|
99 |
-
|
100 |
-
|
101 |
-
self.voice_mode_active = False
|
102 |
-
self.selected_voice = "en-US-JennyNeural" # Default voice
|
103 |
-
# === Voice Mode Initialization (End) ===
|
104 |
|
105 |
def update_internal_state(self, emotion_deltas, cognitive_load_deltas, introspection_delta, engagement_delta):
|
106 |
for emotion, delta in emotion_deltas.items():
|
@@ -128,7 +121,7 @@ class XylariaChat:
|
|
128 |
|
129 |
def update_belief_system(self, statement, belief_score):
|
130 |
self.belief_system[statement] = belief_score
|
131 |
-
|
132 |
def dynamic_belief_update(self, user_message):
|
133 |
sentences = [s.strip() for s in user_message.split('.') if s.strip()]
|
134 |
sentence_counts = Counter(sentences)
|
@@ -234,7 +227,7 @@ class XylariaChat:
|
|
234 |
return "Current strategy is effective. Continue with the current approach."
|
235 |
else:
|
236 |
return " ".join(adjustments)
|
237 |
-
|
238 |
def introspect(self):
|
239 |
introspection_report = "Introspection Report:\n"
|
240 |
introspection_report += f" Current Emotional State:\n"
|
@@ -284,7 +277,7 @@ class XylariaChat:
|
|
284 |
response = "I'm feeling quite energized and ready to assist! " + response
|
285 |
else:
|
286 |
response = "I'm in a good mood and happy to help. " + response
|
287 |
-
|
288 |
if curiosity > 0.7:
|
289 |
response += " I'm very curious about this topic, could you tell me more?"
|
290 |
if frustration > 0.5:
|
@@ -310,7 +303,7 @@ class XylariaChat:
|
|
310 |
if goal["goal"] == "Provide helpful, informative, and contextually relevant responses":
|
311 |
goal["priority"] = max(goal["priority"] - 0.1, 0.0)
|
312 |
goal["progress"] = max(goal["progress"] - 0.2, 0.0)
|
313 |
-
|
314 |
if "learn more" in feedback_lower:
|
315 |
for goal in self.goals:
|
316 |
if goal["goal"] == "Actively learn and adapt from interactions to improve conversational abilities":
|
@@ -321,7 +314,7 @@ class XylariaChat:
|
|
321 |
if goal["goal"] == "Maintain a coherent, engaging, and empathetic conversation flow":
|
322 |
goal["priority"] = max(goal["priority"] - 0.1, 0.0)
|
323 |
goal["progress"] = max(goal["progress"] - 0.2, 0.0)
|
324 |
-
|
325 |
if self.internal_state["emotions"]["curiosity"] > 0.8:
|
326 |
for goal in self.goals:
|
327 |
if goal["goal"] == "Identify and fill knowledge gaps by seeking external information":
|
@@ -398,8 +391,8 @@ class XylariaChat:
|
|
398 |
|
399 |
try:
|
400 |
self.client = InferenceClient(
|
401 |
-
model="Qwen/
|
402 |
-
|
403 |
)
|
404 |
except Exception as e:
|
405 |
print(f"Error resetting API client: {e}")
|
@@ -432,7 +425,7 @@ class XylariaChat:
|
|
432 |
|
433 |
except Exception as e:
|
434 |
return f"Error processing image: {str(e)}"
|
435 |
-
|
436 |
def generate_image(self, prompt):
|
437 |
try:
|
438 |
image = self.image_gen_client.text_to_image(prompt)
|
@@ -447,58 +440,9 @@ class XylariaChat:
|
|
447 |
return text.strip()
|
448 |
except Exception as e:
|
449 |
return f"Error during Math OCR: {e}"
|
450 |
-
|
451 |
-
# === Voice Mode Methods (Start) ===
|
452 |
-
async def speak_text(self, text):
|
453 |
-
if not text:
|
454 |
-
return None, None
|
455 |
-
|
456 |
-
temp_file = "temp_audio.mp3"
|
457 |
-
try:
|
458 |
-
communicator = edge_tts.Communicate(text, self.selected_voice)
|
459 |
-
await communicator.save(temp_file)
|
460 |
-
return temp_file
|
461 |
-
except Exception as e:
|
462 |
-
print(f"Error during text-to-speech: {e}")
|
463 |
-
return None, None
|
464 |
-
|
465 |
-
def recognize_speech(self, timeout=10, phrase_time_limit=10):
|
466 |
-
recognizer = sr.Recognizer()
|
467 |
-
recognizer.energy_threshold = 4000
|
468 |
-
recognizer.dynamic_energy_threshold = True
|
469 |
-
|
470 |
-
with sr.Microphone() as source:
|
471 |
-
print("Listening...")
|
472 |
-
try:
|
473 |
-
audio_data = recognizer.listen(source, timeout=timeout, phrase_time_limit=phrase_time_limit)
|
474 |
-
print("Processing speech...")
|
475 |
-
text = recognizer.recognize_whisper_api(audio_data, api_key=self.hf_token)
|
476 |
-
print(f"Recognized: {text}")
|
477 |
-
return text
|
478 |
-
except sr.WaitTimeoutError:
|
479 |
-
print("No speech detected within the timeout period.")
|
480 |
-
return ""
|
481 |
-
except sr.UnknownValueError:
|
482 |
-
print("Speech recognition could not understand audio")
|
483 |
-
return ""
|
484 |
-
except sr.RequestError as e:
|
485 |
-
print(f"Could not request results from Whisper API; {e}")
|
486 |
-
return ""
|
487 |
-
except Exception as e:
|
488 |
-
print(f"An error occurred during speech recognition: {e}")
|
489 |
-
return ""
|
490 |
-
# === Voice Mode Methods (End) ===
|
491 |
-
|
492 |
def get_response(self, user_input, image=None):
|
493 |
try:
|
494 |
-
# === Voice Mode Adaptation (Start) ===
|
495 |
-
if self.voice_mode_active:
|
496 |
-
print("Voice mode is active, using speech recognition.")
|
497 |
-
user_input = self.recognize_speech() # Get input from speech
|
498 |
-
if not user_input:
|
499 |
-
return "I didn't hear anything." , None
|
500 |
-
# === Voice Mode Adaptation (End) ===
|
501 |
-
|
502 |
messages = []
|
503 |
|
504 |
messages.append(ChatMessage(
|
@@ -525,7 +469,7 @@ class XylariaChat:
|
|
525 |
role="user",
|
526 |
content=user_input
|
527 |
).to_dict())
|
528 |
-
|
529 |
entities = []
|
530 |
relationships = []
|
531 |
|
@@ -535,19 +479,19 @@ class XylariaChat:
|
|
535 |
extracted_relationships = self.extract_relationships(message['content'])
|
536 |
entities.extend(extracted_entities)
|
537 |
relationships.extend(extracted_relationships)
|
538 |
-
|
539 |
self.update_knowledge_graph(entities, relationships)
|
540 |
self.run_metacognitive_layer()
|
541 |
-
|
542 |
for message in messages:
|
543 |
if message['role'] == 'user':
|
544 |
self.dynamic_belief_update(message['content'])
|
545 |
-
|
546 |
for cause, effects in self.causal_rules_db.items():
|
547 |
if any(cause in msg['content'].lower() for msg in messages if msg['role'] == 'user') and any(
|
548 |
effect in msg['content'].lower() for msg in messages for effect in effects):
|
549 |
self.store_information("Causal Inference", f"It seems {cause} might be related to {', '.join(effects)}.")
|
550 |
-
|
551 |
for concept, generalization in self.concept_generalizations.items():
|
552 |
if any(concept in msg['content'].lower() for msg in messages if msg['role'] == 'user'):
|
553 |
self.store_information("Inferred Knowledge", f"This reminds me of a general principle: {generalization}.")
|
@@ -555,54 +499,28 @@ class XylariaChat:
|
|
555 |
if self.internal_state["emotions"]["curiosity"] > 0.8 and any("?" in msg['content'] for msg in messages if msg['role'] == 'user'):
|
556 |
print("Simulating external knowledge seeking...")
|
557 |
self.store_information("External Knowledge", "This is a placeholder for external information I would have found")
|
558 |
-
|
559 |
self.store_information("User Input", user_input)
|
560 |
|
561 |
input_tokens = sum(len(msg['content'].split()) for msg in messages)
|
562 |
max_new_tokens = 16384 - input_tokens - 50
|
563 |
|
564 |
max_new_tokens = min(max_new_tokens, 10020)
|
565 |
-
|
566 |
-
# === Voice Mode Output (Start) ===
|
567 |
-
if self.voice_mode_active:
|
568 |
-
stream = self.client.chat_completion(
|
569 |
-
messages=messages,
|
570 |
-
model="Qwen/Qwen-32B-Preview",
|
571 |
-
temperature=0.7,
|
572 |
-
max_tokens=max_new_tokens,
|
573 |
-
top_p=0.9,
|
574 |
-
stream=True
|
575 |
-
)
|
576 |
-
|
577 |
-
full_response = ""
|
578 |
-
for chunk in stream:
|
579 |
-
if chunk.choices and chunk.choices[0].delta and chunk.choices[0].delta.content:
|
580 |
-
full_response += chunk.choices[0].delta.content
|
581 |
-
|
582 |
-
full_response = self.adjust_response_based_on_state(full_response)
|
583 |
-
audio_file = asyncio.run(self.speak_text(full_response))
|
584 |
|
585 |
-
|
586 |
-
|
587 |
-
|
588 |
-
|
589 |
-
|
590 |
-
|
591 |
-
|
592 |
-
|
593 |
-
|
594 |
-
|
595 |
-
|
596 |
-
temperature=0.7,
|
597 |
-
max_tokens=max_new_tokens,
|
598 |
-
top_p=0.9,
|
599 |
-
stream=True
|
600 |
-
)
|
601 |
-
|
602 |
-
return stream
|
603 |
except Exception as e:
|
604 |
print(f"Detailed error in get_response: {e}")
|
605 |
-
return f"Error generating response: {str(e)}"
|
606 |
|
607 |
def extract_entities(self, text):
|
608 |
words = text.split()
|
@@ -619,7 +537,7 @@ class XylariaChat:
|
|
619 |
if words[i].istitle() and words[i+2].istitle():
|
620 |
relationships.append((words[i], words[i+1], words[i+2]))
|
621 |
return relationships
|
622 |
-
|
623 |
def messages_to_prompt(self, messages):
|
624 |
prompt = ""
|
625 |
for msg in messages:
|
@@ -633,149 +551,30 @@ class XylariaChat:
|
|
633 |
return prompt
|
634 |
|
635 |
def create_interface(self):
|
636 |
-
|
637 |
-
|
638 |
-
self.voice_mode_active = active_state
|
639 |
-
if self.voice_mode_active:
|
640 |
-
# Get the list of available voices
|
641 |
-
voices = asyncio.run(edge_tts.list_voices())
|
642 |
-
voice_names = [voice['ShortName'] for voice in voices]
|
643 |
-
|
644 |
-
# Select a random voice from the list
|
645 |
-
random_voice = random.choice(voice_names)
|
646 |
-
self.selected_voice = random_voice
|
647 |
-
|
648 |
-
return gr.Button.update(value="Stop Voice Mode"), gr.Dropdown.update(value=random_voice)
|
649 |
-
else:
|
650 |
-
return gr.Button.update(value="Start Voice Mode"), gr.Dropdown.update(value=self.selected_voice)
|
651 |
-
|
652 |
-
def update_selected_voice(voice_name):
|
653 |
-
self.selected_voice = voice_name
|
654 |
-
return voice_name
|
655 |
-
|
656 |
-
# === Voice-Specific UI Elements (End) ===
|
657 |
-
|
658 |
-
def streaming_response(message, chat_history, image_filepath, math_ocr_image_path, voice_mode_state, selected_voice):
|
659 |
-
if self.voice_mode_active:
|
660 |
-
response_text, audio_output = self.get_response(message)
|
661 |
-
|
662 |
-
if isinstance(response_text, str):
|
663 |
-
updated_history = chat_history + [[message, response_text]]
|
664 |
-
if audio_output:
|
665 |
-
yield updated_history, audio_output, None, None, ""
|
666 |
-
else:
|
667 |
-
yield updated_history, None, None, None, ""
|
668 |
-
else:
|
669 |
-
full_response = ""
|
670 |
-
updated_history = chat_history + [[message, ""]]
|
671 |
-
try:
|
672 |
-
for chunk in response_text:
|
673 |
-
if chunk.choices and chunk.choices[0].delta and chunk.choices[0].delta.content:
|
674 |
-
chunk_content = chunk.choices[0].delta.content
|
675 |
-
full_response += chunk_content
|
676 |
-
updated_history[-1][1] = full_response
|
677 |
-
if audio_output:
|
678 |
-
yield updated_history, audio_output, None, None, ""
|
679 |
-
else:
|
680 |
-
yield updated_history, None, None, None, ""
|
681 |
-
except Exception as e:
|
682 |
-
print(f"Streaming error: {e}")
|
683 |
-
updated_history[-1][1] = f"Error during response: {e}"
|
684 |
-
if audio_output:
|
685 |
-
yield updated_history, audio_output, None, None, ""
|
686 |
-
else:
|
687 |
-
yield updated_history, None, None, None, ""
|
688 |
-
return
|
689 |
-
|
690 |
-
full_response = self.adjust_response_based_on_state(full_response)
|
691 |
-
|
692 |
-
audio_file = asyncio.run(self.speak_text(full_response))
|
693 |
-
|
694 |
-
self.update_goals(message)
|
695 |
-
|
696 |
-
emotion_deltas = {}
|
697 |
-
cognitive_load_deltas = {}
|
698 |
-
engagement_delta = 0
|
699 |
-
|
700 |
-
if any(word in message.lower() for word in ["sad", "unhappy", "depressed", "down"]):
|
701 |
-
emotion_deltas.update({"valence": -0.2, "arousal": 0.1, "confidence": -0.1, "sadness": 0.3, "joy": -0.2})
|
702 |
-
engagement_delta = -0.1
|
703 |
-
elif any(word in message.lower() for word in ["happy", "good", "great", "excited", "amazing"]):
|
704 |
-
emotion_deltas.update({"valence": 0.2, "arousal": 0.2, "confidence": 0.1, "sadness": -0.2, "joy": 0.3})
|
705 |
-
engagement_delta = 0.2
|
706 |
-
elif any(word in message.lower() for word in ["angry", "mad", "furious", "frustrated"]):
|
707 |
-
emotion_deltas.update({"valence": -0.3, "arousal": 0.3, "dominance": -0.2, "frustration": 0.2, "sadness": 0.1, "joy": -0.1})
|
708 |
-
engagement_delta = -0.2
|
709 |
-
elif any(word in message.lower() for word in ["scared", "afraid", "fearful", "anxious"]):
|
710 |
-
emotion_deltas.update({"valence": -0.2, "arousal": 0.4, "dominance": -0.3, "confidence": -0.2, "sadness": 0.2})
|
711 |
-
engagement_delta = -0.1
|
712 |
-
elif any(word in message.lower() for word in ["surprise", "amazed", "astonished"]):
|
713 |
-
emotion_deltas.update({"valence": 0.1, "arousal": 0.5, "dominance": 0.1, "curiosity": 0.3, "sadness": -0.1, "joy": 0.1})
|
714 |
-
engagement_delta = 0.3
|
715 |
-
elif any(word in message.lower() for word in ["confused", "uncertain", "unsure"]):
|
716 |
-
cognitive_load_deltas.update({"processing_intensity": 0.2})
|
717 |
-
emotion_deltas.update({"curiosity": 0.2, "confidence": -0.1, "sadness": 0.1})
|
718 |
-
engagement_delta = 0.1
|
719 |
-
else:
|
720 |
-
emotion_deltas.update({"valence": 0.05, "arousal": 0.05})
|
721 |
-
engagement_delta = 0.05
|
722 |
-
|
723 |
-
if "learn" in message.lower() or "explain" in message.lower() or "know more" in message.lower():
|
724 |
-
emotion_deltas.update({"curiosity": 0.3})
|
725 |
-
cognitive_load_deltas.update({"processing_intensity": 0.1})
|
726 |
-
engagement_delta = 0.2
|
727 |
-
|
728 |
-
self.update_internal_state(emotion_deltas, cognitive_load_deltas, 0.1, engagement_delta)
|
729 |
-
|
730 |
-
self.conversation_history.append(ChatMessage(role="user", content=message).to_dict())
|
731 |
-
self.conversation_history.append(ChatMessage(role="assistant", content=full_response).to_dict())
|
732 |
-
|
733 |
-
if len(self.conversation_history) > 10:
|
734 |
-
self.conversation_history = self.conversation_history[-10:]
|
735 |
-
|
736 |
-
if audio_file:
|
737 |
-
yield updated_history, audio_file, None, None, ""
|
738 |
-
else:
|
739 |
-
yield updated_history, None, None, None, ""
|
740 |
-
|
741 |
-
# Handling /image command for image generation
|
742 |
if "/image" in message:
|
743 |
image_prompt = message.replace("/image", "").strip()
|
744 |
|
745 |
-
#
|
746 |
placeholder_image = "data:image/svg+xml," + requests.utils.quote(f'''
|
747 |
<svg width="256" height="256" viewBox="0 0 256 256" xmlns="http://www.w3.org/2000/svg">
|
748 |
-
|
749 |
-
|
750 |
-
|
751 |
-
|
752 |
-
|
753 |
-
0% {{ fill: #626262; }}
|
754 |
-
50% {{ fill: #111111; }}
|
755 |
-
100% {{ fill: #626262; }}
|
756 |
-
}}
|
757 |
-
text {{
|
758 |
-
font-family: 'Helvetica Neue', Arial, sans-serif; /* Choose a good font */
|
759 |
-
font-weight: 300; /* Slightly lighter font weight */
|
760 |
-
text-shadow: 0px 2px 4px rgba(0, 0, 0, 0.4); /* Subtle shadow */
|
761 |
-
}}
|
762 |
-
</style>
|
763 |
-
<rect width="256" height="256" rx="20" fill="#888888" />
|
764 |
-
<text x="50%" y="50%" dominant-baseline="middle" text-anchor="middle" font-size="24" fill="white" opacity="0.8">
|
765 |
-
<tspan>creating your image</tspan>
|
766 |
-
<tspan x="50%" dy="1.2em">with xylaria iris</tspan>
|
767 |
-
</text>
|
768 |
</svg>
|
769 |
''')
|
770 |
|
771 |
updated_history = chat_history + [[message, gr.Image(value=placeholder_image, type="pil", visible=True)]]
|
772 |
-
yield
|
773 |
|
774 |
try:
|
775 |
generated_image = self.generate_image(image_prompt)
|
776 |
|
777 |
updated_history[-1][1] = gr.Image(value=generated_image, type="pil", visible=True)
|
778 |
-
yield
|
779 |
|
780 |
self.conversation_history.append(ChatMessage(role="user", content=message).to_dict())
|
781 |
self.conversation_history.append(ChatMessage(role="assistant", content="Image generated").to_dict())
|
@@ -783,15 +582,15 @@ class XylariaChat:
|
|
783 |
return
|
784 |
except Exception as e:
|
785 |
updated_history[-1][1] = f"Error generating image: {e}"
|
786 |
-
yield
|
787 |
return
|
788 |
-
|
789 |
ocr_text = ""
|
790 |
if math_ocr_image_path:
|
791 |
ocr_text = self.perform_math_ocr(math_ocr_image_path)
|
792 |
if ocr_text.startswith("Error"):
|
793 |
updated_history = chat_history + [[message, ocr_text]]
|
794 |
-
yield
|
795 |
return
|
796 |
else:
|
797 |
message = f"Math OCR Result: {ocr_text}\n\nUser's message: {message}"
|
@@ -800,10 +599,10 @@ class XylariaChat:
|
|
800 |
response_stream = self.get_response(message, image_filepath)
|
801 |
else:
|
802 |
response_stream = self.get_response(message)
|
803 |
-
|
804 |
if isinstance(response_stream, str):
|
805 |
updated_history = chat_history + [[message, response_stream]]
|
806 |
-
yield
|
807 |
return
|
808 |
|
809 |
full_response = ""
|
@@ -814,13 +613,13 @@ class XylariaChat:
|
|
814 |
if chunk.choices and chunk.choices[0].delta and chunk.choices[0].delta.content:
|
815 |
chunk_content = chunk.choices[0].delta.content
|
816 |
full_response += chunk_content
|
817 |
-
|
818 |
updated_history[-1][1] = full_response
|
819 |
-
yield
|
820 |
except Exception as e:
|
821 |
print(f"Streaming error: {e}")
|
822 |
updated_history[-1][1] = f"Error during response: {e}"
|
823 |
-
yield
|
824 |
return
|
825 |
|
826 |
full_response = self.adjust_response_based_on_state(full_response)
|
@@ -853,14 +652,14 @@ class XylariaChat:
|
|
853 |
else:
|
854 |
emotion_deltas.update({"valence": 0.05, "arousal": 0.05})
|
855 |
engagement_delta = 0.05
|
856 |
-
|
857 |
if "learn" in message.lower() or "explain" in message.lower() or "know more" in message.lower():
|
858 |
emotion_deltas.update({"curiosity": 0.3})
|
859 |
cognitive_load_deltas.update({"processing_intensity": 0.1})
|
860 |
engagement_delta = 0.2
|
861 |
-
|
862 |
self.update_internal_state(emotion_deltas, cognitive_load_deltas, 0.1, engagement_delta)
|
863 |
-
|
864 |
self.conversation_history.append(ChatMessage(role="user", content=message).to_dict())
|
865 |
self.conversation_history.append(ChatMessage(role="assistant", content=full_response).to_dict())
|
866 |
|
@@ -874,74 +673,6 @@ class XylariaChat:
|
|
874 |
background-color: #f5f5f5;
|
875 |
font-family: 'Source Sans Pro', sans-serif;
|
876 |
}
|
877 |
-
|
878 |
-
.voice-mode-button {
|
879 |
-
background-color: #4CAF50; /* Green */
|
880 |
-
border: none;
|
881 |
-
color: white;
|
882 |
-
padding: 15px 32px;
|
883 |
-
text-align: center;
|
884 |
-
text-decoration: none;
|
885 |
-
display: inline-block;
|
886 |
-
font-size: 16px;
|
887 |
-
margin: 4px 2px;
|
888 |
-
cursor: pointer;
|
889 |
-
border-radius: 10px; /* Rounded corners */
|
890 |
-
transition: all 0.3s ease; /* Smooth transition for hover effect */
|
891 |
-
}
|
892 |
-
|
893 |
-
/* Style when voice mode is active */
|
894 |
-
.voice-mode-button.active {
|
895 |
-
background-color: #f44336; /* Red */
|
896 |
-
}
|
897 |
-
|
898 |
-
/* Hover effect */
|
899 |
-
.voice-mode-button:hover {
|
900 |
-
opacity: 0.8;
|
901 |
-
}
|
902 |
-
|
903 |
-
/* Style for the voice mode overlay */
|
904 |
-
.voice-mode-overlay {
|
905 |
-
position: fixed; /* Stay in place */
|
906 |
-
left: 0;
|
907 |
-
top: 0;
|
908 |
-
width: 100%; /* Full width */
|
909 |
-
height: 100%; /* Full height */
|
910 |
-
background-color: rgba(0, 0, 0, 0.7); /* Black w/ opacity */
|
911 |
-
z-index: 10; /* Sit on top */
|
912 |
-
display: flex;
|
913 |
-
justify-content: center;
|
914 |
-
align-items: center;
|
915 |
-
border-radius: 10px;
|
916 |
-
}
|
917 |
-
|
918 |
-
/* Style for the growing circle */
|
919 |
-
.voice-mode-circle {
|
920 |
-
width: 100px;
|
921 |
-
height: 100px;
|
922 |
-
background-color: #4CAF50;
|
923 |
-
border-radius: 50%;
|
924 |
-
display: flex;
|
925 |
-
justify-content: center;
|
926 |
-
align-items: center;
|
927 |
-
animation: grow 2s infinite;
|
928 |
-
}
|
929 |
-
|
930 |
-
/* Keyframes for the growing animation */
|
931 |
-
@keyframes grow {
|
932 |
-
0% {
|
933 |
-
transform: scale(1);
|
934 |
-
opacity: 0.8;
|
935 |
-
}
|
936 |
-
50% {
|
937 |
-
transform: scale(1.5);
|
938 |
-
opacity: 0.5;
|
939 |
-
}
|
940 |
-
100% {
|
941 |
-
transform: scale(1);
|
942 |
-
opacity: 0.8;
|
943 |
-
}
|
944 |
-
}
|
945 |
|
946 |
.gradio-container {
|
947 |
max-width: 900px;
|
@@ -1102,23 +833,6 @@ class XylariaChat:
|
|
1102 |
display: flex;
|
1103 |
align-items: center;
|
1104 |
}
|
1105 |
-
|
1106 |
-
.audio-container {
|
1107 |
-
display: flex;
|
1108 |
-
align-items: center;
|
1109 |
-
margin-top: 10px;
|
1110 |
-
}
|
1111 |
-
|
1112 |
-
.audio-player {
|
1113 |
-
width: 100%;
|
1114 |
-
border-radius: 15px;
|
1115 |
-
}
|
1116 |
-
|
1117 |
-
.audio-icon {
|
1118 |
-
width: 30px;
|
1119 |
-
height: 30px;
|
1120 |
-
margin-right: 10px;
|
1121 |
-
}
|
1122 |
"""
|
1123 |
|
1124 |
with gr.Blocks(theme=gr.themes.Soft(
|
@@ -1139,30 +853,6 @@ class XylariaChat:
|
|
1139 |
)
|
1140 |
)
|
1141 |
|
1142 |
-
# === Voice Mode UI (Start) ===
|
1143 |
-
voice_mode_btn = gr.Button("Start Voice Mode", elem_classes="voice-mode-button")
|
1144 |
-
|
1145 |
-
voices = asyncio.run(edge_tts.list_voices())
|
1146 |
-
voice_names = [voice['ShortName'] for voice in voices]
|
1147 |
-
|
1148 |
-
voice_dropdown = gr.Dropdown(
|
1149 |
-
label="Select Voice",
|
1150 |
-
choices=voice_names,
|
1151 |
-
value=self.selected_voice,
|
1152 |
-
interactive=True
|
1153 |
-
)
|
1154 |
-
voice_dropdown.input(
|
1155 |
-
fn=update_selected_voice,
|
1156 |
-
inputs=voice_dropdown,
|
1157 |
-
outputs=voice_dropdown
|
1158 |
-
)
|
1159 |
-
voice_mode_btn.click(
|
1160 |
-
fn=toggle_voice_mode,
|
1161 |
-
inputs=voice_mode_btn,
|
1162 |
-
outputs=[voice_mode_btn, voice_dropdown]
|
1163 |
-
)
|
1164 |
-
# === Voice Mode UI (End) ===
|
1165 |
-
|
1166 |
with gr.Accordion("Image Input", open=False, elem_classes="gr-accordion"):
|
1167 |
with gr.Row(elem_classes="image-container"):
|
1168 |
with gr.Column(elem_classes="image-upload"):
|
@@ -1193,16 +883,15 @@ class XylariaChat:
|
|
1193 |
clear = gr.Button("Clear Conversation", variant="stop")
|
1194 |
clear_memory = gr.Button("Clear Memory")
|
1195 |
|
1196 |
-
# Pass voice_mode_state and selected_voice to the streaming_response function
|
1197 |
btn.click(
|
1198 |
fn=streaming_response,
|
1199 |
-
inputs=[txt, chatbot, img, math_ocr_img
|
1200 |
-
outputs=[
|
1201 |
)
|
1202 |
txt.submit(
|
1203 |
fn=streaming_response,
|
1204 |
-
inputs=[txt, chatbot, img, math_ocr_img
|
1205 |
-
outputs=[
|
1206 |
)
|
1207 |
|
1208 |
clear.click(
|
@@ -1231,5 +920,5 @@ def main():
|
|
1231 |
debug=True
|
1232 |
)
|
1233 |
|
1234 |
-
if
|
1235 |
main()
|
|
|
11 |
import numpy as np
|
12 |
import networkx as nx
|
13 |
from collections import Counter
|
|
|
|
|
|
|
|
|
14 |
|
15 |
@dataclass
|
16 |
class ChatMessage:
|
|
|
21 |
return {"role": self.role, "content": self.content}
|
22 |
|
23 |
class XylariaChat:
|
24 |
+
def _init_(self):
|
25 |
self.hf_token = os.getenv("HF_TOKEN")
|
26 |
if not self.hf_token:
|
27 |
raise ValueError("HuggingFace token not found in environment variables")
|
28 |
|
29 |
self.client = InferenceClient(
|
30 |
+
model="Qwen/QwQ-32B-Preview",
|
31 |
+
api_key=self.hf_token
|
32 |
)
|
33 |
|
34 |
self.image_api_url = "https://api-inference.huggingface.co/models/Salesforce/blip-image-captioning-large"
|
|
|
49 |
"bias_detection": 0.0,
|
50 |
"strategy_adjustment": ""
|
51 |
}
|
52 |
+
|
53 |
self.internal_state = {
|
54 |
"emotions": {
|
55 |
"valence": 0.5,
|
|
|
78 |
]
|
79 |
|
80 |
self.system_prompt = """You are a helpful and harmless assistant. You are Xylaria developed by Sk Md Saad Amin. You should think step-by-step """
|
81 |
+
|
82 |
self.causal_rules_db = {
|
83 |
"rain": ["wet roads", "flooding"],
|
84 |
"fire": ["heat", "smoke"],
|
|
|
92 |
"democracy": "government by the people",
|
93 |
"photosynthesis": "process used by plants to convert light to energy"
|
94 |
}
|
95 |
+
|
96 |
+
# ... (other methods: update_internal_state, update_knowledge_graph, etc.) ...
|
|
|
|
|
|
|
97 |
|
98 |
def update_internal_state(self, emotion_deltas, cognitive_load_deltas, introspection_delta, engagement_delta):
|
99 |
for emotion, delta in emotion_deltas.items():
|
|
|
121 |
|
122 |
def update_belief_system(self, statement, belief_score):
|
123 |
self.belief_system[statement] = belief_score
|
124 |
+
|
125 |
def dynamic_belief_update(self, user_message):
|
126 |
sentences = [s.strip() for s in user_message.split('.') if s.strip()]
|
127 |
sentence_counts = Counter(sentences)
|
|
|
227 |
return "Current strategy is effective. Continue with the current approach."
|
228 |
else:
|
229 |
return " ".join(adjustments)
|
230 |
+
|
231 |
def introspect(self):
|
232 |
introspection_report = "Introspection Report:\n"
|
233 |
introspection_report += f" Current Emotional State:\n"
|
|
|
277 |
response = "I'm feeling quite energized and ready to assist! " + response
|
278 |
else:
|
279 |
response = "I'm in a good mood and happy to help. " + response
|
280 |
+
|
281 |
if curiosity > 0.7:
|
282 |
response += " I'm very curious about this topic, could you tell me more?"
|
283 |
if frustration > 0.5:
|
|
|
303 |
if goal["goal"] == "Provide helpful, informative, and contextually relevant responses":
|
304 |
goal["priority"] = max(goal["priority"] - 0.1, 0.0)
|
305 |
goal["progress"] = max(goal["progress"] - 0.2, 0.0)
|
306 |
+
|
307 |
if "learn more" in feedback_lower:
|
308 |
for goal in self.goals:
|
309 |
if goal["goal"] == "Actively learn and adapt from interactions to improve conversational abilities":
|
|
|
314 |
if goal["goal"] == "Maintain a coherent, engaging, and empathetic conversation flow":
|
315 |
goal["priority"] = max(goal["priority"] - 0.1, 0.0)
|
316 |
goal["progress"] = max(goal["progress"] - 0.2, 0.0)
|
317 |
+
|
318 |
if self.internal_state["emotions"]["curiosity"] > 0.8:
|
319 |
for goal in self.goals:
|
320 |
if goal["goal"] == "Identify and fill knowledge gaps by seeking external information":
|
|
|
391 |
|
392 |
try:
|
393 |
self.client = InferenceClient(
|
394 |
+
model="Qwen/QwQ-32B-Preview",
|
395 |
+
api_key=self.hf_token
|
396 |
)
|
397 |
except Exception as e:
|
398 |
print(f"Error resetting API client: {e}")
|
|
|
425 |
|
426 |
except Exception as e:
|
427 |
return f"Error processing image: {str(e)}"
|
428 |
+
|
429 |
def generate_image(self, prompt):
|
430 |
try:
|
431 |
image = self.image_gen_client.text_to_image(prompt)
|
|
|
440 |
return text.strip()
|
441 |
except Exception as e:
|
442 |
return f"Error during Math OCR: {e}"
|
443 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
444 |
def get_response(self, user_input, image=None):
|
445 |
try:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
446 |
messages = []
|
447 |
|
448 |
messages.append(ChatMessage(
|
|
|
469 |
role="user",
|
470 |
content=user_input
|
471 |
).to_dict())
|
472 |
+
|
473 |
entities = []
|
474 |
relationships = []
|
475 |
|
|
|
479 |
extracted_relationships = self.extract_relationships(message['content'])
|
480 |
entities.extend(extracted_entities)
|
481 |
relationships.extend(extracted_relationships)
|
482 |
+
|
483 |
self.update_knowledge_graph(entities, relationships)
|
484 |
self.run_metacognitive_layer()
|
485 |
+
|
486 |
for message in messages:
|
487 |
if message['role'] == 'user':
|
488 |
self.dynamic_belief_update(message['content'])
|
489 |
+
|
490 |
for cause, effects in self.causal_rules_db.items():
|
491 |
if any(cause in msg['content'].lower() for msg in messages if msg['role'] == 'user') and any(
|
492 |
effect in msg['content'].lower() for msg in messages for effect in effects):
|
493 |
self.store_information("Causal Inference", f"It seems {cause} might be related to {', '.join(effects)}.")
|
494 |
+
|
495 |
for concept, generalization in self.concept_generalizations.items():
|
496 |
if any(concept in msg['content'].lower() for msg in messages if msg['role'] == 'user'):
|
497 |
self.store_information("Inferred Knowledge", f"This reminds me of a general principle: {generalization}.")
|
|
|
499 |
if self.internal_state["emotions"]["curiosity"] > 0.8 and any("?" in msg['content'] for msg in messages if msg['role'] == 'user'):
|
500 |
print("Simulating external knowledge seeking...")
|
501 |
self.store_information("External Knowledge", "This is a placeholder for external information I would have found")
|
502 |
+
|
503 |
self.store_information("User Input", user_input)
|
504 |
|
505 |
input_tokens = sum(len(msg['content'].split()) for msg in messages)
|
506 |
max_new_tokens = 16384 - input_tokens - 50
|
507 |
|
508 |
max_new_tokens = min(max_new_tokens, 10020)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
509 |
|
510 |
+
stream = self.client.chat_completion(
|
511 |
+
messages=messages,
|
512 |
+
model="Qwen/QwQ-32B-Preview",
|
513 |
+
temperature=0.7,
|
514 |
+
max_tokens=max_new_tokens,
|
515 |
+
top_p=0.9,
|
516 |
+
stream=True
|
517 |
+
)
|
518 |
+
|
519 |
+
return stream
|
520 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
521 |
except Exception as e:
|
522 |
print(f"Detailed error in get_response: {e}")
|
523 |
+
return f"Error generating response: {str(e)}"
|
524 |
|
525 |
def extract_entities(self, text):
|
526 |
words = text.split()
|
|
|
537 |
if words[i].istitle() and words[i+2].istitle():
|
538 |
relationships.append((words[i], words[i+1], words[i+2]))
|
539 |
return relationships
|
540 |
+
|
541 |
def messages_to_prompt(self, messages):
|
542 |
prompt = ""
|
543 |
for msg in messages:
|
|
|
551 |
return prompt
|
552 |
|
553 |
def create_interface(self):
|
554 |
+
def streaming_response(message, chat_history, image_filepath, math_ocr_image_path):
|
555 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
556 |
if "/image" in message:
|
557 |
image_prompt = message.replace("/image", "").strip()
|
558 |
|
559 |
+
# Placeholder for image generation
|
560 |
placeholder_image = "data:image/svg+xml," + requests.utils.quote(f'''
|
561 |
<svg width="256" height="256" viewBox="0 0 256 256" xmlns="http://www.w3.org/2000/svg">
|
562 |
+
<rect width="256" height="256" rx="20" fill="#888888" />
|
563 |
+
<animate attributeName="fill" values="#888888;#000000;#888888" dur="2s" repeatCount="indefinite" />
|
564 |
+
<text x="50%" y="50%" dominant-baseline="middle" text-anchor="middle" font-size="48" fill="white">
|
565 |
+
<tspan>/</tspan>
|
566 |
+
</text>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
567 |
</svg>
|
568 |
''')
|
569 |
|
570 |
updated_history = chat_history + [[message, gr.Image(value=placeholder_image, type="pil", visible=True)]]
|
571 |
+
yield "", updated_history, None, None, ""
|
572 |
|
573 |
try:
|
574 |
generated_image = self.generate_image(image_prompt)
|
575 |
|
576 |
updated_history[-1][1] = gr.Image(value=generated_image, type="pil", visible=True)
|
577 |
+
yield "", updated_history, None, None, ""
|
578 |
|
579 |
self.conversation_history.append(ChatMessage(role="user", content=message).to_dict())
|
580 |
self.conversation_history.append(ChatMessage(role="assistant", content="Image generated").to_dict())
|
|
|
582 |
return
|
583 |
except Exception as e:
|
584 |
updated_history[-1][1] = f"Error generating image: {e}"
|
585 |
+
yield "", updated_history, None, None, ""
|
586 |
return
|
587 |
+
|
588 |
ocr_text = ""
|
589 |
if math_ocr_image_path:
|
590 |
ocr_text = self.perform_math_ocr(math_ocr_image_path)
|
591 |
if ocr_text.startswith("Error"):
|
592 |
updated_history = chat_history + [[message, ocr_text]]
|
593 |
+
yield "", updated_history, None, None, ""
|
594 |
return
|
595 |
else:
|
596 |
message = f"Math OCR Result: {ocr_text}\n\nUser's message: {message}"
|
|
|
599 |
response_stream = self.get_response(message, image_filepath)
|
600 |
else:
|
601 |
response_stream = self.get_response(message)
|
602 |
+
|
603 |
if isinstance(response_stream, str):
|
604 |
updated_history = chat_history + [[message, response_stream]]
|
605 |
+
yield "", updated_history, None, None, ""
|
606 |
return
|
607 |
|
608 |
full_response = ""
|
|
|
613 |
if chunk.choices and chunk.choices[0].delta and chunk.choices[0].delta.content:
|
614 |
chunk_content = chunk.choices[0].delta.content
|
615 |
full_response += chunk_content
|
616 |
+
|
617 |
updated_history[-1][1] = full_response
|
618 |
+
yield "", updated_history, None, None, ""
|
619 |
except Exception as e:
|
620 |
print(f"Streaming error: {e}")
|
621 |
updated_history[-1][1] = f"Error during response: {e}"
|
622 |
+
yield "", updated_history, None, None, ""
|
623 |
return
|
624 |
|
625 |
full_response = self.adjust_response_based_on_state(full_response)
|
|
|
652 |
else:
|
653 |
emotion_deltas.update({"valence": 0.05, "arousal": 0.05})
|
654 |
engagement_delta = 0.05
|
655 |
+
|
656 |
if "learn" in message.lower() or "explain" in message.lower() or "know more" in message.lower():
|
657 |
emotion_deltas.update({"curiosity": 0.3})
|
658 |
cognitive_load_deltas.update({"processing_intensity": 0.1})
|
659 |
engagement_delta = 0.2
|
660 |
+
|
661 |
self.update_internal_state(emotion_deltas, cognitive_load_deltas, 0.1, engagement_delta)
|
662 |
+
|
663 |
self.conversation_history.append(ChatMessage(role="user", content=message).to_dict())
|
664 |
self.conversation_history.append(ChatMessage(role="assistant", content=full_response).to_dict())
|
665 |
|
|
|
673 |
background-color: #f5f5f5;
|
674 |
font-family: 'Source Sans Pro', sans-serif;
|
675 |
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
676 |
|
677 |
.gradio-container {
|
678 |
max-width: 900px;
|
|
|
833 |
display: flex;
|
834 |
align-items: center;
|
835 |
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
836 |
"""
|
837 |
|
838 |
with gr.Blocks(theme=gr.themes.Soft(
|
|
|
853 |
)
|
854 |
)
|
855 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
856 |
with gr.Accordion("Image Input", open=False, elem_classes="gr-accordion"):
|
857 |
with gr.Row(elem_classes="image-container"):
|
858 |
with gr.Column(elem_classes="image-upload"):
|
|
|
883 |
clear = gr.Button("Clear Conversation", variant="stop")
|
884 |
clear_memory = gr.Button("Clear Memory")
|
885 |
|
|
|
886 |
btn.click(
|
887 |
fn=streaming_response,
|
888 |
+
inputs=[txt, chatbot, img, math_ocr_img],
|
889 |
+
outputs=[txt, chatbot, img, math_ocr_img, txt]
|
890 |
)
|
891 |
txt.submit(
|
892 |
fn=streaming_response,
|
893 |
+
inputs=[txt, chatbot, img, math_ocr_img],
|
894 |
+
outputs=[txt, chatbot, img, math_ocr_img, txt]
|
895 |
)
|
896 |
|
897 |
clear.click(
|
|
|
920 |
debug=True
|
921 |
)
|
922 |
|
923 |
+
if _name_ == "_main_":
|
924 |
main()
|