fix endings etc.

.gitignore

@@ -1,6 +1,7 @@
 #macos
 .DS_store
 venv
+sandox
 
 # Byte-compiled / optimized / DLL files
 __pycache__/

run.sh

@@ -0,0 +1,2 @@
+. venv/bin/activate
+python speak_broadcast.py -c ../assets/voice.config.json

speak.py

@@ -1,11 +1,13 @@
 import argparse, json, sys, time, random
 import utterance.voice
 import utterance.utils
+
+UTTERANCE_LEN = 64				#<--------------- these should be in config
 				
 def main() -> int:
 
 	p = argparse.ArgumentParser()
-	p.add_argument("-c", "--config", type=str, default="config.json", help="configuratin file")
+	p.add_argument("-c", "--config", type=str, default="voice.config.json", help="configuratin file")
 	p.add_argument("-i", "--iterations", type=int, default=10, help="number of iterations")
 	args = p.parse_args()
 
@@ -16,25 +18,23 @@ def main() -> int:
 
 	voices = []
 	for v in conf['voices']:
-		voice = utterance.voice.Voice(name=v["name"].upper(), model=v['model_dir'], tokenizer=v['tokeniser_file'], temp=float(v["temperature"]), lenght=32)
+		model = v['model']
+		voice = utterance.voice.Voice(name=v["name"].upper(), model=model['model_dir'], tokenizer=model['tokeniser_file'], temp=float(model["temperature"]), length=UTTERANCE_LEN)
 		voices.append(voice)
 
-	nbr_voices = len(voices)
-	current_voice = ""
+	current_voice_name = ""
 	for i in range(args.iterations):
-		rindex = random.randint(0, nbr_voices - 1)
-		v = voices[rindex]
-		if v.name != current_voice:
+		v = random.sample(voices, 1)[0]
+		if v.name != current_voice_name:
+			current_voice_name = v.name
 			print("==========")
 			print(v.name + ":")
-			current_voice = v.name
-		t = v.utter_one()
-		if t != None:
+			t = v.utter_one()
 			t = utterance.utils.clean(t)
 			t = utterance.utils.format(t)
 			print(t)
 		
-		time.sleep(4)
+		# time.sleep(1)
 
 
 if __name__ == '__main__':

speak_broadcast-backup.py

@@ -0,0 +1,359 @@
+import argparse, json, sys, time, random, logging, signal, threading
+import utterance.voice
+import utterance.utils
+import utterance.osc
+import examine.metric
+
+logging.basicConfig(level=logging.INFO)
+
+UTTERANCE_LEN = 64				#<--------------- these should be in config
+NUM_METRIC_GEN = 75
+NUM_SAMPLE_VOICES = 3
+RANDOM_SEED_TIMER_MIN = 2
+STATE_TRANSITION_TIMER_MIN = 1.5
+broadcast = None
+metric = None
+exit = False
+terminal = False
+debug = False
+state = "METRIC"
+
+def format_str(text) -> str:
+	t = utterance.utils.clean(text)
+	return utterance.utils.format(t)
+
+def tokenise_str(text):
+	return utterance.utils.tokenise(text)
+
+def utter_one(v, temp=None, length=None) -> str:
+	u = v.utter_one(temp=temp, length=length)
+	return format_str(u)
+
+def prompt_one(v, pinput: str, temp=None, length=None) -> str:
+	u = v.prompt(pinput=pinput, temp=None, length=length)
+	return format_str(u)
+
+def utter_one_vectorise(v, temp=None, length=None):
+	global metric
+	uv = utter_one(v, temp, length)
+	uv_vec = metric.vector(uv)
+	return uv, uv_vec
+
+def prompt_one_vectorise(v, pinput: str, temp=None, length=None):
+	global metric
+	uv = prompt_one(v, pinput, temp, length)
+	uv_vec = metric.vector(uv)
+	return uv, uv_vec
+
+def utter_n_vectorise_distance(v, n, vec, temp=None, length=None):	
+	global metric
+	results = []
+	texts = v.utter_n(n=n, temp=temp, length=length)
+	for t in texts:
+		t = format_str(t)
+		t_vec = metric.vector(t)
+		d = examine.metric.cos_dist(vec, t_vec)
+		results.append([d, t, v])
+	return results
+
+def terminal_utterance(utterance):
+	if terminal:
+		print(utterance, end="")
+
+def broadcast_utterance(v, utterance):	
+
+	global broadcast, exit, debug
+
+	# Send all text to server to calculate bounds in advance
+	broadcast.utterance(utterance, v.calculate)	
+
+	text = ""
+	broadcast.utterance(text, v.channel)
+	terminal_utterance(text)
+	time.sleep(2)
+
+	frags = v.fragments(utterance)
+
+	for f in frags:
+		text += f
+		broadcast.utterance(text, v.channel)
+		terminal_utterance(f)
+
+		# sleep_time = 2
+		# toks = tokenise_str(f)
+		toks = f.split()
+		sleep_time = len(toks)
+		if sleep_time <= 2:
+			sleep_time += 1
+		time.sleep(sleep_time)
+		if exit:
+			return
+
+	broadcast.command('clear')
+	print("==========")
+
+	time.sleep(3)
+
+def find_candidates(v, uv_vec, voices, results):
+	logging.info(f"LOOP::finding candidates")
+
+	start = time.time()
+	candidates = random.sample(voices, NUM_SAMPLE_VOICES)
+	for c in candidates:
+		if exit:
+			break
+		if c == v:
+			continue
+		results += utter_n_vectorise_distance(c, NUM_METRIC_GEN, uv_vec)
+
+	results.sort(key=lambda t: t[0], reverse=True)
+
+	lapse = time.time() - start
+	logging.info(f"LOOP::done - {lapse} secs")
+
+def update():
+	global exit
+	while not exit:
+		try:
+			utterance.osc.update()
+		except Exception as e:
+			logging.error(e) 
+			pass
+		
+		time.sleep(0.2)
+
+def signal_terminate(signum, frame):
+	global exit
+	logging.warning("::SIGNAL TERMINATE::")
+	exit = True
+
+def main() -> int:
+
+	global broadcast, metric, terminal, debug, state
+
+	p = argparse.ArgumentParser()
+	p.add_argument("-c", "--config", type=str, default="voice.config.json", help="configuratin file")
+	p.add_argument("-i", "--iterations", type=int, default=10, help="number of iterations")
+	p.add_argument("-t", "--terminal", action='store_true', help="print to terminal")
+	args = p.parse_args()	
+
+	logging.info(f"INIT::loading config file - {args.config}")
+
+	with open(args.config) as f:
+		conf = json.load(f)
+
+	logging.info(conf)
+
+	terminal = args.terminal
+
+	#--------------------#
+	# 		VOICES
+	#--------------------#
+	logging.info(f"INIT::creating voices")
+
+	voices = []
+	for v in conf['voices']:
+		model = v['model']
+		voice = utterance.voice.Voice(name=v["name"].upper(), model=model['model_dir'], tokenizer=model['tokeniser_file'], temp=float(model["temperature"]), length=UTTERANCE_LEN)
+		voice.set_channel(v['osc_channel']['root'], v['osc_channel']['utterance'])
+		voice.set_calculate(v['osc_channel']['root'], v['osc_channel']['calculate'])
+		voice.set_temperature(v['osc_channel']['root'], v['osc_channel']['temperature'])
+		voices.append(voice)
+
+	#--------------------#
+	# 		QUESTION
+	#--------------------#
+	logging.info(f"INIT::setting up question")
+	questions = conf['questions']
+	questions_array = questions.copy()
+	random.shuffle(questions_array)
+
+
+	#--------------------#
+	# 		NET
+	#--------------------#
+	logging.info(f"INIT::setting up OSC")
+
+	utterance.osc.start_osc()
+
+	broadcast = utterance.osc.OscBroadcaster(name="osc_broadcast", host=conf['host_voicemachine'], port=conf['port_voicemachine'], command_channel=conf['command_osc_channel'])
+
+	def receiver_cb(temp, name):
+		global debug
+		if type(temp) == str and temp == "DEBUG":
+			debug = name
+			logging.info(f"DEBUG MODE: {debug}")
+		else: # temperature
+			for v in voices:
+				if v.name == name:
+					print(f'{name} - {temp}')
+					v.temp = temp
+					# broadcast.temperature(temp, v.temperature)  # <-- doesn works because deadlocks in osc_process...
+
+	receiver = utterance.osc.OscReceiver(name="osc_receiver", host=conf['host_machinespeak'], port=conf['port_machinespeak'], callback_fn=receiver_cb)
+
+
+	#--------------------#
+	# 		METRIC
+	#--------------------#
+	logging.info(f"INIT::loading doc2vec metrics")
+	metric = examine.metric.Metric(model_input='data/models/doc2vec.model')
+
+	#--------------------#
+	# 		RANDOM
+	#--------------------#
+
+	def random_seed(seconds):
+		global t_random_seed, exit
+		i = 0
+		while i < seconds:
+			i += 1
+			time.sleep(1)
+			if exit:
+				return
+		logging.info("RANDOM::SEEDING RANDOM")
+		random.seed(time.time())
+		if not exit:
+			t_random_seed = threading.Thread(target=random_seed, args=(random.randint(60, 60 * RANDOM_SEED_TIMER_MIN), ))
+			t_random_seed.start()
+
+	t_random_seed = threading.Thread(target=random_seed, args=(random.randint(60, 60 * RANDOM_SEED_TIMER_MIN), ))
+	t_random_seed.start()
+
+	#--------------------#
+	#  STATE TRANSITION
+	#--------------------#
+
+	# STATES = ["METRIC", "QUESTION", "RANDOM"]
+	
+
+	def state_transition(seconds):
+		global t_state_transition, exit, state
+		i = 0
+		while i < seconds:
+			i += 1
+			time.sleep(1)
+			if exit:
+				return
+		logging.info("STATE::STATE TRANSITION")
+
+		state = "QUESTION"
+
+		if not exit:
+			t_state_transition = threading.Thread(target=state_transition, args=(random.randint(60, 60 * STATE_TRANSITION_TIMER_MIN), ))
+			t_state_transition.start()
+
+	t_state_transition = threading.Thread(target=state_transition, args=(random.randint(60, 60 * STATE_TRANSITION_TIMER_MIN), ))
+	t_state_transition.start()
+
+	#--------------------#
+	# 		  A
+	#--------------------#
+	logging.info(f"INIT::generating first utterance")
+	
+	v = random.choice(voices)
+	uv, uv_vec = utter_one_vectorise(v)
+
+	# -- this only updates OSC -- 
+	# -- might not need this  in production 
+	t_update = threading.Thread(target=update)
+	t_update.start()
+
+	
+	while not exit:
+
+		if state == "METRIC":
+
+			logging.info(f"- state METRIC")
+
+			results = []
+			t = threading.Thread(target=find_candidates, args=[v, uv_vec, voices, results])
+			t.start()
+
+			logging.info(f"METRIC::broadcasting {v.name}")
+			broadcast_utterance(v, uv)
+			
+			t.join()
+
+			# ok here we need to randomise maybe...?!
+			# ([d, t, v])
+
+			choice = results[0]
+			# makse sure we don't say the same thing over and over again
+			for r in results:
+				v = r[2]
+				u = r[1]
+				if v.select(u):
+					choice = r
+					break
+				else:
+					logging.info(f"METRIC::reduncancy {v.name}")
+
+			v = choice[2]
+			uv = choice[1]
+			uv_vec = metric.vector(uv)
+			logging.info(f"METRIC::next {v.name}")
+
+		elif state == "QUESTION":
+
+			logging.info(f"- state QUESTION")
+
+			if len(questions_array) <= 0:
+				questions_array = questions.copy()
+				random.shuffle(questions_array)
+
+			# random question
+			q = questions_array.pop(0)
+
+			# random voice
+			v = random.choice(voices)
+
+			# random voice asks random question
+
+			logging.info(f"QUESTION::{v.name} : {q['question']}")
+
+			broadcast.utterance(q['question'], v.calculate)
+			broadcast.utterance(q['question'], v.channel)
+
+			time.sleep(15)
+
+			# answer
+			v = [e for e in voices if e.name == q['voice']]
+			if len(v) == 1:				
+				v = v[0]
+				logging.info(f"QUESTION::answer - {v.name}")
+				uv, uv_vec = prompt_one_vectorise(v, q['prompt'])
+				v.remember(uv)
+				# broadcast_utterance(v, uv) <-- this is broadcasted as part of METRIC STATE
+
+			state = "METRIC"
+
+
+		elif state == "RANDOM":
+
+			logging.info(f"- state RANDOM")
+
+			v = random.choice(voices)
+			l = random.randint(5, UTTERANCE_LEN)
+			uv, uv_vec = utter_one_vectorise(v, length=l)
+			broadcast_utterance(v, uv)
+
+
+	t_update.join()
+
+	logging.info(f"TERMINATE::terminating OSC")
+	utterance.osc.terminate_osc()
+
+	# if t_random_seed:
+	logging.info(f"TERMINATE::random seed")
+	t_random_seed.join()
+
+	logging.info(f"TERMINATE::state transition")
+	t_state_transition.join()
+
+	logging.info(f"FIN")
+
+
+if __name__ == '__main__':
+	signal.signal(signal.SIGINT, signal_terminate)
+	sys.exit(main())

speak_broadcast.py

@@ -1,34 +1,56 @@
-import argparse, json, sys, time, random, logging, signal, threading
+import argparse, json, sys, time, random, logging, signal, threading, string
 import utterance.voice
 import utterance.utils
-import utterance.osc
+import utterance.oscosc
 import examine.metric
 
-UTTERANCE_LEN = 16				#<--------------- these should be in config
-NUM_METRIC_GEN = 50
+logging.basicConfig(level=logging.INFO)
+
+UTTERANCE_LEN = 64				#<--------------- these should be in config
+NUM_METRIC_GEN = 75
 NUM_SAMPLE_VOICES = 3
+RANDOM_SEED_TIMER_MIN = 2
+STATE_TRANSITION_TIMER_MIN = 10
 broadcast = None
 metric = None
 exit = False
+terminal = False
+debug = False
+state = "METRIC"
+B_SKIP = []
+B_SWAPS = {}
 
 def format_str(text) -> str:
 	t = utterance.utils.clean(text)
 	return utterance.utils.format(t)
 
-def utter_one(v) -> str:
-	u = v.utter_one()	
+def tokenise_str(text):
+	return utterance.utils.tokenise(text)
+
+def utter_one(v, temp=None, length=None) -> str:
+	u = v.utter_one(temp=temp, length=length)
 	return format_str(u)
 
-def utter_one_vectorise(v):
+def prompt_one(v, pinput: str, temp=None, length=None) -> str:
+	u = v.prompt(pinput=pinput, temp=None, length=length)
+	return format_str(u)
+
+def utter_one_vectorise(v, temp=None, length=None):
 	global metric
-	uv = utter_one(v)
+	uv = utter_one(v, temp, length)
 	uv_vec = metric.vector(uv)
 	return uv, uv_vec
 
-def utter_n_vectorise_distance(v, n, vec):	
+def prompt_one_vectorise(v, pinput: str, temp=None, length=None):
+	global metric
+	uv = prompt_one(v, pinput, temp, length)
+	uv_vec = metric.vector(uv)
+	return uv, uv_vec
+
+def utter_n_vectorise_distance(v, n, vec, temp=None, length=None):	
 	global metric
 	results = []
-	texts = v.utter_n(n=n)
+	texts = v.utter_n(n=n, temp=temp, length=length)
 	for t in texts:
 		t = format_str(t)
 		t_vec = metric.vector(t)
@@ -36,26 +58,98 @@ def utter_n_vectorise_distance(v, n, vec):
 		results.append([d, t, v])
 	return results
 
+def terminal_utterance(utterance):
+	if terminal:
+		print(utterance, end="")
+
+def fix_ending(frags):
+
+	result = frags.copy()
+
+	end = result[-1]
+	end = end.translate(str.maketrans('', '', string.punctuation))
+
+	fix = utterance.utils.fix_sentence(end)
+
+	if fix is None or len(fix) == 0:
+		result = result[:-1]
+		# result[-1] = result[-1]
+	else:
+		result[-1] = fix
+
+	result[-1] = utterance.utils.fix_punctuation(result[-1])
+
+	print(result)
+
+	return result
+
+def fix_beginning(frags):
+
+	global B_SKIP, B_SWAPS
+
+	result = frags.copy()
+
+	beginnig = result[0]
+	toks = beginnig.split()
+
+	if len(toks) > 0:
+		f = toks[0].lower() 
+		if f[0] in string.punctuation:
+			f = f[1:]
+		if f in B_SKIP:
+			if len(toks) > 2:
+				result[0] = " ".join(toks[1:]).capitalize() + "\n"
+			else:				
+				if result[1][0] == ' ':
+					result[1] = result[1][1:] 	
+				result[1] = result[1].capitalize()	
+				return result[1:]
+		elif toks[0] in B_SWAPS:
+			result[0] = result[0].replace(toks[0], B_SWAPS[toks[0]])
+
+	return result
+
+
 def broadcast_utterance(v, utterance):	
 
-	global broadcast, exit
+	print(utterance)
 
-	text = v.name.upper() + ":\n"
+	global broadcast, exit, debug
 
+	# Send all text to server to calculate bounds in advance
+	broadcast.utterance(utterance, v.calculate)	
+
+	text = ""
 	broadcast.utterance(text, v.channel)
+	terminal_utterance(text)
 	time.sleep(2)
 
 	frags = v.fragments(utterance)
 
+	frags = fix_beginning(frags)
+	frags = fix_ending(frags)
+
 	for f in frags:
+
+		terminal_utterance(f)
+
 		text += f
 		broadcast.utterance(text, v.channel)
-		time.sleep(2)
+
+		# sleep_time = 2
+		# toks = tokenise_str(f)
+		toks = f.split()
+		sleep_time = len(toks)
+		if sleep_time <= 2:
+			sleep_time += 1
+		time.sleep(sleep_time)
 		if exit:
 			return
 
 	broadcast.command('clear')
-	time.sleep(2)
+	print("==========")
+
+	time.sleep(3)
 
 def find_candidates(v, uv_vec, voices, results):
 	logging.info(f"LOOP::finding candidates")
@@ -74,7 +168,16 @@ def find_candidates(v, uv_vec, voices, results):
 	lapse = time.time() - start
 	logging.info(f"LOOP::done - {lapse} secs")
 
-
+# def update():
+# 	global exit
+# 	while not exit:
+# 		try:
+# 			utterance.osc.update()
+# 		except Exception as e:
+# 			logging.error(e) 
+# 			pass
+		
+# 		time.sleep(0.2)
 
 def signal_terminate(signum, frame):
 	global exit
@@ -83,14 +186,13 @@ def signal_terminate(signum, frame):
 
 def main() -> int:
 
-	global broadcast, metric
+	global broadcast, metric, terminal, debug, state, UTTERANCE_LEN, NUM_METRIC_GEN, NUM_SAMPLE_VOICES, RANDOM_SEED_TIMER_MIN, STATE_TRANSITION_TIMER_MIN, B_SKIP, B_SWAPS
 
 	p = argparse.ArgumentParser()
 	p.add_argument("-c", "--config", type=str, default="voice.config.json", help="configuratin file")
 	p.add_argument("-i", "--iterations", type=int, default=10, help="number of iterations")
-	args = p.parse_args()
-
-	logging.basicConfig(level=logging.INFO)
+	p.add_argument("-t", "--terminal", action='store_true', help="print to terminal")
+	args = p.parse_args()	
 
 	logging.info(f"INIT::loading config file - {args.config}")
 
@@ -99,6 +201,22 @@ def main() -> int:
 
 	logging.info(conf)
 
+	terminal = args.terminal
+
+	#--------------------#
+	# 		CONFIGS
+	#--------------------#
+
+	u_conf = conf['utterance_configuration']
+	UTTERANCE_LEN = u_conf['UTTERANCE_LEN']
+	NUM_METRIC_GEN = u_conf['NUM_METRIC_GEN']
+	NUM_SAMPLE_VOICES = u_conf['NUM_SAMPLE_VOICES']
+	RANDOM_SEED_TIMER_MIN = u_conf['RANDOM_SEED_TIMER_MIN']
+	STATE_TRANSITION_TIMER_MIN = u_conf['STATE_TRANSITION_TIMER_MIN']
+
+	B_SKIP = u_conf['b_skip']
+	B_SWAPS = u_conf['b_swaps']
+
 	#--------------------#
 	# 		VOICES
 	#--------------------#
@@ -107,17 +225,45 @@ def main() -> int:
 	voices = []
 	for v in conf['voices']:
 		model = v['model']
-		voice = utterance.voice.Voice(name=v["name"].upper(), model=model['model_dir'], tokenizer=model['tokeniser_file'], temp=float(model["temperature"]), lenght=UTTERANCE_LEN)
+		voice = utterance.voice.Voice(name=v["name"].upper(), model=model['model_dir'], tokenizer=model['tokeniser_file'], temp=float(model["temperature"]), length=UTTERANCE_LEN)
 		voice.set_channel(v['osc_channel']['root'], v['osc_channel']['utterance'])
+		voice.set_calculate(v['osc_channel']['root'], v['osc_channel']['calculate'])
+		voice.set_temperature(v['osc_channel']['root'], v['osc_channel']['temperature'])
 		voices.append(voice)
 
+	#--------------------#
+	# 		QUESTION
+	#--------------------#
+	logging.info(f"INIT::setting up question")
+	questions = conf['questions']
+	questions_array = questions.copy()
+	random.shuffle(questions_array)
+
 
 	#--------------------#
 	# 		NET
 	#--------------------#
 	logging.info(f"INIT::setting up OSC")
-	broadcast = utterance.osc.OscBroadcaster(name="osc_broadcast", host=conf['host'], port=conf['port'], command_channel=conf['command_osc_channel'])
-	broadcast.start_osc()	
+
+	broadcast = utterance.oscosc.OscBroadcaster(name="osc_broadcast", host=conf['host_voicemachine'], port=conf['port_voicemachine'], command_channel=conf['command_osc_channel'])
+
+	# def receiver_cb_temp(unused_addr, args, temp, name):
+	# 	for v in voices:
+	# 		if v.name == name:
+	# 			print(f'{name} - {temp}')
+	# 			v.temp = temp
+	# 			# broadcast.temperature(temp, v.temperature)  # <-- doesn works because deadlocks in osc_process...
+
+	# def receiver_cb_command(unused_addr, args, cmd):
+	# 	global debug		
+	# 	debug = name
+	# 	logging.info(f"DEBUG MODE: {debug}")
+		
+
+	# receiver = utterance.oscosc.OscReceiver(name="osc_receiver", host=conf['host_machinespeak'], port=conf['port_machinespeak'], callback_fn_command=receiver_cb_command, callback_fn_temp=receiver_cb_temp)
+
+	# t_osc_receiver = threading.Thread(target=receiver.server.serve_forever)
+	# t_osc_receiver.start()
 
 	#--------------------#
 	# 		METRIC
@@ -125,36 +271,163 @@ def main() -> int:
 	logging.info(f"INIT::loading doc2vec metrics")
 	metric = examine.metric.Metric(model_input='data/models/doc2vec.model')
 
+	#--------------------#
+	# 		RANDOM
+	#--------------------#
+
+	def random_seed(seconds):
+		global t_random_seed, exit
+		i = 0
+		while i < seconds:
+			i += 1
+			time.sleep(1)
+			if exit:
+				return
+		logging.info("RANDOM::SEEDING RANDOM")
+		random.seed(time.time())
+		if not exit:
+			t_random_seed = threading.Thread(target=random_seed, args=(random.randint(60, 60 * RANDOM_SEED_TIMER_MIN), ))
+			t_random_seed.start()
+
+	t_random_seed = threading.Thread(target=random_seed, args=(random.randint(60, 60 * RANDOM_SEED_TIMER_MIN), ))
+	t_random_seed.start()
+
+	#--------------------#
+	#  STATE TRANSITION
+	#--------------------#
+
+	# STATES = ["METRIC", "QUESTION", "RANDOM"]
+	
+
+	def state_transition(seconds):
+		global t_state_transition, exit, state
+		i = 0
+		while i < seconds:
+			i += 1
+			time.sleep(1)
+			if exit:
+				return
+		logging.info("STATE::STATE TRANSITION")
+
+		state = "QUESTION"
+
+		if not exit:
+			t_state_transition = threading.Thread(target=state_transition, args=(random.randint(60, 60 * STATE_TRANSITION_TIMER_MIN), ))
+			t_state_transition.start()
+
+	t_state_transition = threading.Thread(target=state_transition, args=(random.randint(60, 60 * STATE_TRANSITION_TIMER_MIN), ))
+	t_state_transition.start()
+
 	#--------------------#
 	# 		  A
 	#--------------------#
 	logging.info(f"INIT::generating first utterance")
-	random.seed(time.time())
+	
 	v = random.choice(voices)
 	uv, uv_vec = utter_one_vectorise(v)
+
+	# -- this only updates OSC -- 
+	# -- might not need this  in production 
+	# t_update = threading.Thread(target=update)
+	# t_update.start()
+
 	
 	while not exit:
 
-		results = []
-		t = threading.Thread(target=find_candidates, args=[v, uv_vec, voices, results])
-		t.start()
+		if state == "METRIC":
 
-		logging.info(f"LOOP::broadcasting {v.name}")
-		broadcast_utterance(v, uv)
-		
-		t.join()
+			logging.info(f"- state METRIC")
 
-		# ok here we need to randomise maybe...?!
-		# ([d, t, v])
-		choice = results[0]
-		v = choice[2]
-		uv = choice[1]
-		uv_vec = metric.vector(uv)
-		logging.info(f"LOOP::next {v.name}")
+			results = []
+			t = threading.Thread(target=find_candidates, args=[v, uv_vec, voices, results])
+			t.start()
+
+			logging.info(f"METRIC::broadcasting {v.name}")
+
+			try:
+				broadcast_utterance(v, uv)
+			except Exception as e:
+				logging.error(e)
+				pass
+			
+			t.join()
+
+			# ok here we need to randomise maybe...?!
+			# ([d, t, v])
+
+			choice = results[0]
+			# makse sure we don't say the same thing over and over again
+			for r in results:
+				v = r[2]
+				u = r[1]
+				if v.select(u):
+					choice = r
+					break
+				else:
+					logging.info(f"METRIC::reduncancy {v.name}")
+
+			v = choice[2]
+			uv = choice[1]
+			uv_vec = metric.vector(uv)
+			logging.info(f"METRIC::next {v.name}")
+
+		elif state == "QUESTION":
+
+			logging.info(f"- state QUESTION")
+
+			if len(questions_array) <= 0:
+				questions_array = questions.copy()
+				random.shuffle(questions_array)
+
+			# random question
+			q = questions_array.pop(0)
+
+			# random voice
+			v = random.choice(voices)
+
+			# random voice asks random question
+
+			logging.info(f"QUESTION::{v.name} : {q['question']}")
+
+			broadcast.utterance(q['question'], v.calculate)
+			broadcast.utterance(q['question'], v.channel)
+
+			time.sleep(15)
+
+			# answer
+			v = [e for e in voices if e.name == q['voice']]
+			if len(v) == 1:				
+				v = v[0]
+				logging.info(f"QUESTION::answer - {v.name}")
+				uv, uv_vec = prompt_one_vectorise(v, q['prompt'])
+				v.remember(uv)
+				# broadcast_utterance(v, uv) <-- this is broadcasted as part of METRIC STATE
+
+			state = "METRIC"
 
 
-	logging.info(f"TERMINATE::terminating OSC")
-	broadcast.terminate_osc()
+		elif state == "RANDOM":
+
+			logging.info(f"- state RANDOM")
+
+			v = random.choice(voices)
+			l = random.randint(5, UTTERANCE_LEN)
+			uv, uv_vec = utter_one_vectorise(v, length=l)
+			broadcast_utterance(v, uv)
+
+
+	# t_update.join()
+
+	# logging.info(f"TERMINATE::terminating OSC")
+	# t_osc_receiver.stop()
+	# t_osc_receiver.join()
+
+	# if t_random_seed:
+	logging.info(f"TERMINATE::random seed")
+	t_random_seed.join()
+
+	logging.info(f"TERMINATE::state transition")
+	t_state_transition.join()
 
 	logging.info(f"FIN")
 

utterance/osc.py

@@ -1,5 +1,24 @@
 from osc4py3.as_eventloop import *
 from osc4py3 import oscbuildparse
+from threading import Lock
+
+mutex = Lock()
+temp_mutex = Lock()
+
+
+def start_osc():
+	osc_startup()
+
+def terminate_osc():
+	osc_terminate()
+
+def temperature(temp, name):
+	print(f'{temp} - {name}')
+
+def update():
+	global mutex
+	with mutex:
+		osc_process()
 
 class OscBroadcaster:
 
@@ -8,27 +27,38 @@ class OscBroadcaster:
 		self.host = host
 		self.port = port
 		self.cmd = command_channel
-		osc_udp_client(self.host, self.port, self.name)
+		osc_udp_client(self.host, self.port, self.name)		
 
 	def utterance(self, utterance: str, channel: str):
 		msg = oscbuildparse.OSCMessage(channel, None, [utterance])
 		osc_send(msg, self.name)
-		osc_process()
+		update()
 
 	def command(self, command: str):
 		msg = oscbuildparse.OSCMessage(self.cmd, None, [command])
 		osc_send(msg, self.name)
-		osc_process()
-
-	@staticmethod
-	def start_osc():
-		osc_startup()
-
-	@staticmethod
-	def terminate_osc():
-		osc_terminate()
+		update()
 
+	def temperature(self, temp: float, channel: str):
+		global temp_mutex
+		with temp_mutex:
+			msg = oscbuildparse.OSCMessage(channel, None, [temp])			
+			osc_send(msg, self.name)			
+			osc_process()
+			
 
+class OscReceiver:
+
+	def __init__(self, name: str, host: str, port: str, callback_fn=None):
+		self.name = name
+		self.host = host
+		self.port = port
+		# osc_udp_server(self.host, self.port, self.name)		
+		osc_udp_server("0.0.0.0", self.port, self.name)
+		if callback_fn:
+			osc_method('/', callback_fn)
+		else:
+			osc_method('/', temperature)
 
 
 

utterance/oscosc.py

@@ -0,0 +1,56 @@
+
+from pythonosc import udp_client
+from pythonosc import osc_server
+from pythonosc import dispatcher
+from threading import Lock
+
+temp_mutex = Lock()
+
+class OscBroadcaster:
+
+	def __init__(self, name: str, host: str, port: str, command_channel: str):
+		self.name = name
+		self.host = host
+		self.port = int(port)
+		self.cmd = command_channel
+		self.client = udp_client.SimpleUDPClient(self.host, self.port)
+
+	def utterance(self, utterance: str, channel: str):
+		self.client.send_message(channel, utterance)
+
+	def command(self, command: str):
+		self.client.send_message(self.cmd, command)
+
+	def temperature(self, temp: float, channel: str):		
+		global temp_mutex
+		with temp_mutex:
+			self.client.send_message(self.cmd, command)
+			
+
+class OscReceiver:
+
+	def __init__(self, name: str, host: str, port: str, callback_fn_command=None, callback_fn_temp=None):
+
+		self.dispatcher = dispatcher.Dispatcher()
+		if callback_fn_command != None:
+			self.dispatcher.map('/command', callback_fn_command)
+		if callback_fn_temp != None:
+			self.dispatcher.map('/temperature', callback_fn_command)
+
+		self.name = name
+		self.host = "127.0.0.1"
+		self.port = int(port)
+		self.server = osc_server.ThreadingOSCUDPServer((self.host, self.port), self.dispatcher)
+
+
+
+
+		# # osc_udp_server(self.host, self.port, self.name)		
+		# osc_udp_server("127.0.0.1", self.port, self.name)
+		# if callback_fn:
+		# 	osc_method('/', callback_fn)
+		# else:
+		# 	osc_method('/', temperature)
+
+
+

utterance/utils.py

@@ -1,4 +1,6 @@
 import string, regex
+from gensim.utils import tokenize
+import nltk
 
 def clean(text: str) -> str:
 
@@ -16,13 +18,15 @@ def clean(text: str) -> str:
 
 def format(text: str) -> str:
 
-	return text.replace('\r\n', '\n').replace('\n\n', '\n')
+	return text.replace('\r\n', '\n').replace('\n\n', '\n').replace('‘', "'").replace("’", "'")
 
 def fragments(utterance: str):
 	frags = []
 	sentences = utterance.splitlines()
 
-	PUNCT_RE = regex.compile(r'(\p{Punctuation})')  
+	PUNCT_RE = regex.compile(r'(\p{Punctuation})') 
+
+	skip_punctuation = ["'"] 
 
 	for s in sentences:
 		sf = PUNCT_RE.split(s)
@@ -36,13 +40,74 @@ def fragments(utterance: str):
 				cum += k
 			else:
 				cum += k
-				frags.append(cum)
-				cum = ""
+				if k in skip_punctuation:
+					continue
+				else:
+					# if cum[0] not in [' ', '\n']:
+					# 	cum = ' ' + cum
+					frags.append(cum)
+					cum = ""
 		cum += '\n'
 		frags.append(cum)
 
+	# get rid of newline (2x)
+	if len(frags) > 0: 
+		if frags[-1] == '\n':
+			frags = frags[:-1]
+		if len(frags) > 0:
+			if frags[-1] == '\n':
+				frags = frags[:-1]
+
 	return frags
 
 
+def tokenise(utterance: str):
+	return list(tokenize(utterance, lower=True))
+
+def fix_sentence(s: str) -> str:
+
+	if len(s.strip()) <= 1:
+		return None
+
+	text = nltk.word_tokenize(s)
+	tags = nltk.pos_tag(text)
+
+	if len(tags) == 0:
+		return None
+	elif len(tags) == 1 and tags[0][1] in ['DT', 'WDT', 'IN']:
+		return None
+	elif len(tags) == 1 and len(tags[0][0]) == 1 and tags[0][1] in ['NN']:
+		return None
+
+	if tags[-1][1] in ['DT', 'WDT', 'IN', 'CC']:
+		tags = tags[:-1]
+		return " ".join([x[0] for x in tags])
+	elif tags[-1][1] in ['NN'] and len(tags[-1][0]) == 1:
+		tags = tags[:-1]
+		return " ".join([x[0] for x in tags])
+
+	return s	
+
+def fix_punctuation(s: str) -> str:
+
+	if len(s.strip()) == 0:
+		return ""
+
+	if len(s) == 1 and s in string.punctuation:
+		if s != ',':
+			return s
+		else:
+			return ""
+
+	e = s.rstrip()
+	if e[-1] in string.punctuation:
+		if e[-1] in [',', ':']:
+			s = e[:-1] + '.'
+	else:
+		s = e + '.'
+
+	return s	
+
+
 
 

utterance/voice.py

@@ -1,22 +1,58 @@
 from aitextgen import aitextgen
 import utterance.utils
-import regex, string
+import gensim, regex, string, time
+
+UTTERANCE_MEMORY_LEN = 15
+UTTERANCE_MEMORY_MIN_DIST = 0.85
 
 class Voice:
 
-	def __init__(self, name: str, model: str, tokenizer: str, temp: int, lenght: int):
+	def __init__(self, name: str, model: str, tokenizer: str, temp: int, length: int):
 		self.name = name
 		self.temp = temp
-		self.lenght = lenght
+		self.length = length
 		self.v = aitextgen(model_folder=model, tokenizer_file=tokenizer)
+		self.utterances = []
 
-	def utter_n(self, n: int, temp: float = None, lenght: int = None):
-		t = self.temp if temp != None else temp
-		l = self.lenght if lenght != None else lenght
-		return self.v.generate(n=n, max_lenght=l, temperature=t, return_as_list=True)
+	def utter_n(self, n: int, temp: float = None, length: int = None):
+		t = self.temp if temp == None else temp
+		l = self.length if length == None else length
+		return self.v.generate(n=n, max_length=l, temperature=t, seed=int(time.time()), return_as_list=True)
+
+	def utter_one(self, temp: float = None, length: int = None) -> str:
+		t = self.temp if temp == None else temp
+		l = self.length if length == None else length		
+		return self.utter_n(n=1, temp=temp, length=length)[0]
+
+	def prompt(self, pinput: str, temp: float = None, length: int = None) -> str:
+		t = self.temp if temp == None else temp
+		l = self.length if length == None else length
+		return self.v.generate(prompt=pinput, n=1, max_length=l, temperature=t, seed=int(time.time()), return_as_list=True)[0]
+
+	def select(self, utterance: str) -> bool:
+
+		# fuction making sure to not say the same thing
+
+		toks = set(gensim.utils.tokenize(utterance))
+		
+		i = 0
+		for u in self.utterances:
+			d = gensim.matutils.jaccard_distance(toks, u)
+			print(f"{i} - d: {d}")
+			if d < UTTERANCE_MEMORY_MIN_DIST:
+				return False
+
+		self.utterances.append(toks)
+
+		if len(self.utterances) > UTTERANCE_MEMORY_LEN:
+			self.utterances.pop(0)
+
+		return True
+
+	def remember(self, utterance: str):
+		toks = set(gensim.utils.tokenize(utterance))
+		self.utterances.append(toks)
 
-	def utter_one(self, temp: int = None, lenght: float = None) -> str:
-		return self.utter_n(n=1, temp=temp, lenght=lenght)[0]
 
 	def set_channel(self, root: str, endpoint: str):
 		self.channel = root + endpoint
@@ -24,6 +60,18 @@ class Voice:
 	def channel(self):
 		return self.channel
 
+	def set_calculate(self, root: str, endpoint: str):
+		self.calculate = root + endpoint
+
+	def calculate(self):
+		return self.calculate
+
+	def set_temperature(self, root: str, endpoint: str):
+		self.temperature = root + endpoint
+
+	def calculate(self):
+		return self.temperature
+
 	def fragments(self, utt: str):
 		self.utterance = utt
 		self.utterance_fragments = utterance.utils.fragments(utt)