Update app.py
Browse files"🚀 Primera versión de Mori Contextual 🤖🐈⬛"
app.py
CHANGED
|
@@ -1,208 +1,362 @@
|
|
| 1 |
-
|
|
|
|
|
|
|
|
|
|
| 2 |
os.environ["TF_ENABLE_ONEDNN_OPTS"] = "0"
|
| 3 |
import streamlit as st
|
| 4 |
from pathlib import Path
|
| 5 |
-
import torch, json, csv, warnings
|
| 6 |
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
|
| 7 |
-
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
| 8 |
-
from pathlib import Path
|
| 9 |
from unidecode import unidecode
|
| 10 |
from datetime import datetime
|
| 11 |
-
import
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 12 |
|
| 13 |
-
# 🆔 Asigna un ID de sesión si no existe
|
| 14 |
-
if "user_id" not in st.session_state:
|
| 15 |
-
st.session_state["user_id"] = str(uuid.uuid4())[:8] # Ej: "f6a9b3e2"
|
| 16 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 17 |
def limpiar_input():
|
| 18 |
st.session_state["entrada"] = ""
|
| 19 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 20 |
|
| 21 |
-
|
| 22 |
timestamp = datetime.now().isoformat()
|
| 23 |
|
| 24 |
-
#
|
| 25 |
stats_dir = Path("Statistics")
|
| 26 |
stats_dir.mkdir(parents=True, exist_ok=True)
|
| 27 |
|
| 28 |
-
#
|
| 29 |
archivo_csv = stats_dir / "conversaciones_log.csv"
|
| 30 |
existe_csv = archivo_csv.exists()
|
| 31 |
|
|
|
|
| 32 |
with open(archivo_csv, mode="a", encoding="utf-8", newline="") as f_csv:
|
| 33 |
writer = csv.writer(f_csv)
|
| 34 |
if not existe_csv:
|
| 35 |
-
writer.writerow(["timestamp", "user_id", "
|
| 36 |
-
writer.writerow([timestamp, user_id,
|
| 37 |
|
| 38 |
-
#
|
| 39 |
archivo_jsonl = stats_dir / "conversaciones_log.jsonl"
|
| 40 |
with open(archivo_jsonl, mode="a", encoding="utf-8") as f_jsonl:
|
| 41 |
registro = {
|
| 42 |
"timestamp": timestamp,
|
| 43 |
"user_id": user_id,
|
| 44 |
-
"
|
| 45 |
-
"pregunta":
|
| 46 |
-
"respuesta":
|
| 47 |
-
}
|
| 48 |
f_jsonl.write(json.dumps(registro, ensure_ascii=False) + "\n")
|
| 49 |
|
| 50 |
|
| 51 |
-
#
|
| 52 |
@st.cache_resource
|
| 53 |
def load_model(path_str):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 54 |
path = Path(path_str).resolve()
|
| 55 |
tokenizer = AutoTokenizer.from_pretrained(path, local_files_only=True)
|
| 56 |
model = AutoModelForSeq2SeqLM.from_pretrained(path, local_files_only=True)
|
|
|
|
| 57 |
return model, tokenizer
|
| 58 |
|
| 59 |
-
def detectar_intencion(texto_usuario):
|
| 60 |
-
texto = unidecode(texto_usuario.lower())
|
| 61 |
|
| 62 |
-
|
| 63 |
-
|
| 64 |
-
|
| 65 |
-
|
| 66 |
-
|
| 67 |
-
|
| 68 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 69 |
|
| 70 |
-
|
| 71 |
-
"modelo", "entrenamiento", "algoritmo", "regresion", "clasificacion", "overfitting", "datos", "que es",
|
| 72 |
-
"define", "explicas", "pca", "cnn", "rnn", "clustering", "precision", "recall", "supervisado", "aprendizaje"
|
| 73 |
-
]
|
| 74 |
|
| 75 |
-
|
| 76 |
-
return "Social"
|
| 77 |
-
elif any(p in texto for p in tecnico_keywords):
|
| 78 |
-
return "Técnica"
|
| 79 |
-
else:
|
| 80 |
-
return "Técnica"
|
| 81 |
|
|
|
|
| 82 |
|
| 83 |
-
|
| 84 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 85 |
|
| 86 |
-
|
| 87 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 88 |
return_tensors="pt",
|
| 89 |
padding=True,
|
| 90 |
-
truncation=True,
|
| 91 |
max_length=128 # ✅ especificado para evitar warning
|
| 92 |
).to(device)
|
| 93 |
|
| 94 |
-
|
|
|
|
| 95 |
input_ids=inputs["input_ids"],
|
| 96 |
attention_mask=inputs["attention_mask"], # ✅ FIX agregado
|
| 97 |
max_length=50,
|
| 98 |
-
pad_token_id=
|
| 99 |
do_sample=True,
|
| 100 |
top_p=0.95,
|
| 101 |
-
top_k=50
|
| 102 |
-
)
|
| 103 |
-
|
| 104 |
-
respuesta = social_tokenizer.decode(output_ids[0], skip_special_tokens=True)
|
| 105 |
-
return respuesta.strip()
|
| 106 |
|
| 107 |
-
|
| 108 |
-
|
| 109 |
-
|
|
|
|
| 110 |
|
| 111 |
-
inputs = technical_tokenizer(
|
| 112 |
-
entrada,
|
| 113 |
-
return_tensors="pt",
|
| 114 |
-
padding=True,
|
| 115 |
-
truncation=True,
|
| 116 |
-
max_length=128 # ✅ truncación explícita
|
| 117 |
-
).to(device)
|
| 118 |
|
| 119 |
-
|
| 120 |
-
|
| 121 |
-
attention_mask=inputs["attention_mask"], # ✅ FIX agregado
|
| 122 |
-
max_length=64
|
| 123 |
-
)
|
| 124 |
|
| 125 |
-
|
| 126 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 127 |
|
|
|
|
| 128 |
|
| 129 |
-
|
| 130 |
-
intencion = detectar_intencion(texto_usuario)
|
| 131 |
-
#print("a ver: ", texto_usuario)
|
| 132 |
-
if intencion == "Social":
|
| 133 |
-
return responder_social(texto_usuario)
|
| 134 |
-
elif intencion == "Técnica":
|
| 135 |
-
return responder_tecnico(texto_usuario)
|
| 136 |
-
else:
|
| 137 |
-
return responder_tecnico(texto_usuario) #"🤔 No entendí bien... ¿quieres hablar de ciencia de datos o echar el chisme?"
|
| 138 |
|
|
|
|
| 139 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 140 |
|
| 141 |
-
|
| 142 |
-
|
|
|
|
|
|
|
| 143 |
|
|
|
|
| 144 |
|
| 145 |
-
|
| 146 |
-
tokenizer_tecnico = AutoTokenizer.from_pretrained("tecuhtli/mori-tecnico-model")
|
| 147 |
-
model_tecnico = AutoModelForSeq2SeqLM.from_pretrained("tecuhtli/mori-tecnico-model")
|
| 148 |
|
| 149 |
-
|
| 150 |
-
|
| 151 |
-
|
| 152 |
|
| 153 |
-
|
| 154 |
-
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
| 155 |
-
model_social = model_social.to(device)
|
| 156 |
-
model_tecnico = model_tecnico.to(device)
|
| 157 |
|
| 158 |
-
|
| 159 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 160 |
|
| 161 |
-
#
|
|
|
|
|
|
|
| 162 |
|
| 163 |
-
|
| 164 |
-
|
| 165 |
-
submitted = st.form_submit_button("Responder")
|
| 166 |
|
| 167 |
-
|
| 168 |
-
|
| 169 |
-
|
| 170 |
-
|
| 171 |
-
|
| 172 |
-
|
| 173 |
-
else:
|
| 174 |
-
respuesta = "🤝 [Mori Social] " + responder_social(entrada_usuario, model_social, tokenizer_social)
|
| 175 |
-
st.success(respuesta)
|
| 176 |
|
| 177 |
-
# Guarda en historial
|
| 178 |
-
st.session_state.historial.append(("Mori", respuesta))
|
| 179 |
-
st.session_state.historial.append(("Tú", entrada_usuario))
|
| 180 |
|
| 181 |
-
#
|
| 182 |
-
|
|
|
|
|
|
|
| 183 |
|
| 184 |
|
| 185 |
-
|
| 186 |
|
|
|
|
|
|
|
| 187 |
|
| 188 |
-
# 🔁 Muestra historial
|
| 189 |
-
if st.session_state.historial:
|
| 190 |
-
st.markdown("---")
|
| 191 |
-
for autor, texto in reversed(st.session_state.historial):
|
| 192 |
-
if autor == "Tú":
|
| 193 |
-
st.markdown(f"🧍♂️ **{autor}**: {texto}")
|
| 194 |
else:
|
| 195 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 196 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 197 |
|
| 198 |
-
if st.session_state.historial:
|
| 199 |
-
texto_chat = ""
|
| 200 |
-
for autor, texto in st.session_state.historial:
|
| 201 |
-
texto_chat += f"{autor}: {texto}\n\n"
|
| 202 |
|
| 203 |
-
|
| 204 |
-
|
| 205 |
-
|
| 206 |
-
file_name="conversacion_mori.txt",
|
| 207 |
-
mime="text/plain"
|
| 208 |
-
)
|
|
|
|
| 1 |
+
#***************************************************************************
|
| 2 |
+
#Importing Libraries
|
| 3 |
+
#***************************************************************************
|
| 4 |
+
import os, sys, torch, json, csv, warnings, joblib, uuid
|
| 5 |
os.environ["TF_ENABLE_ONEDNN_OPTS"] = "0"
|
| 6 |
import streamlit as st
|
| 7 |
from pathlib import Path
|
|
|
|
| 8 |
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
|
|
|
|
|
|
|
| 9 |
from unidecode import unidecode
|
| 10 |
from datetime import datetime
|
| 11 |
+
from huggingface_hub import hf_hub_download
|
| 12 |
+
#***************************************************************************
|
| 13 |
+
#Defining default paths for the model to work
|
| 14 |
+
#***************************************************************************
|
| 15 |
+
|
| 16 |
+
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
| 17 |
+
|
| 18 |
+
#***************************************************************************
|
| 19 |
+
#Setting up variables
|
| 20 |
+
#***************************************************************************
|
| 21 |
|
|
|
|
|
|
|
|
|
|
| 22 |
|
| 23 |
+
#***************************************************************************
|
| 24 |
+
#Functions
|
| 25 |
+
#***************************************************************************
|
| 26 |
+
|
| 27 |
+
# Function to clean the question field
|
| 28 |
def limpiar_input():
|
| 29 |
st.session_state["entrada"] = ""
|
| 30 |
|
| 31 |
+
# Function to save user interaction
|
| 32 |
+
def saving_interaction(question, response, context, user_id):
|
| 33 |
+
|
| 34 |
+
'''
|
| 35 |
+
inputs:
|
| 36 |
+
|
| 37 |
+
question --> User input question
|
| 38 |
+
response --> Mori response to the user question
|
| 39 |
+
context --> Context related to the user input, found by the trained classifier
|
| 40 |
+
user_id --> ID for the current user (Unique ID per session)
|
| 41 |
+
|
| 42 |
+
'''
|
| 43 |
|
| 44 |
+
# Getting the current time for the saving log
|
| 45 |
timestamp = datetime.now().isoformat()
|
| 46 |
|
| 47 |
+
# Defining the path to save the current interaction
|
| 48 |
stats_dir = Path("Statistics")
|
| 49 |
stats_dir.mkdir(parents=True, exist_ok=True)
|
| 50 |
|
| 51 |
+
# Setting the file to save the interactions
|
| 52 |
archivo_csv = stats_dir / "conversaciones_log.csv"
|
| 53 |
existe_csv = archivo_csv.exists()
|
| 54 |
|
| 55 |
+
# Saving statistics as a csv
|
| 56 |
with open(archivo_csv, mode="a", encoding="utf-8", newline="") as f_csv:
|
| 57 |
writer = csv.writer(f_csv)
|
| 58 |
if not existe_csv:
|
| 59 |
+
writer.writerow(["timestamp", "user_id", "contexto", "pregunta", "respuesta"])
|
| 60 |
+
writer.writerow([timestamp, user_id, context, question, response])
|
| 61 |
|
| 62 |
+
# Saving statiistics as a json file
|
| 63 |
archivo_jsonl = stats_dir / "conversaciones_log.jsonl"
|
| 64 |
with open(archivo_jsonl, mode="a", encoding="utf-8") as f_jsonl:
|
| 65 |
registro = {
|
| 66 |
"timestamp": timestamp,
|
| 67 |
"user_id": user_id,
|
| 68 |
+
"context": context,
|
| 69 |
+
"pregunta": question,
|
| 70 |
+
"respuesta": response}
|
|
|
|
| 71 |
f_jsonl.write(json.dumps(registro, ensure_ascii=False) + "\n")
|
| 72 |
|
| 73 |
|
| 74 |
+
# Function to load models within the huggingface respositories space
|
| 75 |
@st.cache_resource
|
| 76 |
def load_model(path_str):
|
| 77 |
+
|
| 78 |
+
'''
|
| 79 |
+
inputs:
|
| 80 |
+
|
| 81 |
+
path_str --> Path for loading models and tokenizers
|
| 82 |
+
|
| 83 |
+
outsputs:
|
| 84 |
+
|
| 85 |
+
model --> Loaded Model
|
| 86 |
+
tokenizer --> Loaded tokenizer
|
| 87 |
+
|
| 88 |
+
'''
|
| 89 |
+
|
| 90 |
path = Path(path_str).resolve()
|
| 91 |
tokenizer = AutoTokenizer.from_pretrained(path, local_files_only=True)
|
| 92 |
model = AutoModelForSeq2SeqLM.from_pretrained(path, local_files_only=True)
|
| 93 |
+
|
| 94 |
return model, tokenizer
|
| 95 |
|
|
|
|
|
|
|
| 96 |
|
| 97 |
+
# Funcion para clasificar las preguntas del usuario definiendo el contexto de las mismas
|
| 98 |
+
def classify_context(question, label_classes, model, tokenizer, device):
|
| 99 |
+
|
| 100 |
+
'''
|
| 101 |
+
inputs:
|
| 102 |
+
|
| 103 |
+
question --> Pregunta formulada por el usuario
|
| 104 |
+
label_classes --> Clases del label encoder para decodificar inferencias
|
| 105 |
+
model --> Clasificador para determinar el contexto de las pregutnas
|
| 106 |
+
tokenizer --> Tokenizer usada para clasificar contextos
|
| 107 |
+
device --> Usar el GPU o el CPU dependiendo de su disponibilidad
|
| 108 |
|
| 109 |
+
outsputs:
|
|
|
|
|
|
|
|
|
|
| 110 |
|
| 111 |
+
predicted_label --> Clasificacion de la pregunta en diversos contextos (clases)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 112 |
|
| 113 |
+
'''
|
| 114 |
|
| 115 |
+
# Moviendo el modelo al device disponible
|
| 116 |
+
model = model.to(device)
|
| 117 |
+
|
| 118 |
+
# Procesando la entrada del usuario
|
| 119 |
+
inputs = tokenizer(question, return_tensors="pt", padding=True, truncation=True, max_length=128)
|
| 120 |
+
inputs = {key: val.to(device) for key, val in inputs.items()}
|
| 121 |
+
|
| 122 |
+
# Clasificacion de la pregunta del usuario en contextos
|
| 123 |
+
with torch.no_grad():
|
| 124 |
+
outputs = model(**inputs)
|
| 125 |
+
logits = outputs.logits
|
| 126 |
+
|
| 127 |
+
# Inferencia del clasificador
|
| 128 |
+
pred_intent = torch.argmax(logits, dim=1).item()
|
| 129 |
+
predicted_label = label_classes[pred_intent]
|
| 130 |
+
|
| 131 |
+
return predicted_label
|
| 132 |
|
| 133 |
+
|
| 134 |
+
|
| 135 |
+
# Funcion para generar respuestas tecnicas de Mori
|
| 136 |
+
def technical_asnwer(question, context, model, tokenizer, device):
|
| 137 |
+
|
| 138 |
+
'''
|
| 139 |
+
inputs:
|
| 140 |
+
|
| 141 |
+
question --> Pregunta formulada por el usuario
|
| 142 |
+
context --> Contexto de la preguntadel usario definido por el clasificador
|
| 143 |
+
model --> Modelo de Mori para responder preguntas tecnicas
|
| 144 |
+
tokenizer --> Tokenizer usado para procesar entradas y decoodificar respuestas
|
| 145 |
+
device --> Usar el GPU o el CPU dependiendo de su disponibilidad
|
| 146 |
+
|
| 147 |
+
outsputs:
|
| 148 |
+
|
| 149 |
+
response --> Respues de Mori tecnico (Modelo tecnico)
|
| 150 |
+
|
| 151 |
+
'''
|
| 152 |
+
|
| 153 |
+
# Moviendo el modelo al device disponible
|
| 154 |
+
model = model.to(device)
|
| 155 |
+
|
| 156 |
+
# Promp Engineering para ayudar a Mori a encontrar la mejor respuesta
|
| 157 |
+
input_text = f"Context: {context} [SEP] Question: {question}"
|
| 158 |
+
|
| 159 |
+
# Tokenizando el texto de entrada
|
| 160 |
+
inputs = tokenizer(input_text, return_tensors="pt").to(device)
|
| 161 |
+
|
| 162 |
+
# Generando la respuesta
|
| 163 |
+
summary_ids = model.generate(inputs['input_ids'], max_length=150, num_beams=5, early_stopping=True)
|
| 164 |
+
|
| 165 |
+
# Decodificando la respuesta
|
| 166 |
+
response = tokenizer.decode(summary_ids[0], skip_special_tokens=True)
|
| 167 |
+
|
| 168 |
+
return "🧠 [Mori Técnico] " + response.strip()
|
| 169 |
+
|
| 170 |
+
|
| 171 |
+
# Funcion para generar respuestas sociales de Mori
|
| 172 |
+
def social_asnwer(question, model, tokenizer, device):
|
| 173 |
+
|
| 174 |
+
'''
|
| 175 |
+
inputs:
|
| 176 |
+
|
| 177 |
+
question --> Pregunta formulada por el usuario
|
| 178 |
+
model --> Modelo de Mori para responder preguntas sociales
|
| 179 |
+
tokenizer --> Tokenizer usado para procesar entradas y decoodificar respuestas
|
| 180 |
+
device --> Usar el GPU o el CPU dependiendo de su disponibilidad
|
| 181 |
+
|
| 182 |
+
outsputs:
|
| 183 |
+
|
| 184 |
+
response --> Respues de Mori social (Modelo social)
|
| 185 |
+
|
| 186 |
+
'''
|
| 187 |
+
|
| 188 |
+
# Moviendo el modelo al device disponible
|
| 189 |
+
model = model.to(device)
|
| 190 |
+
|
| 191 |
+
# Tokenizando la entrada del usuario sin agregar <eos> explícitamente
|
| 192 |
+
inputs = tokenizer(
|
| 193 |
+
question, # ✅ sin agregar eos_token
|
| 194 |
return_tensors="pt",
|
| 195 |
padding=True,
|
| 196 |
+
truncation=True,
|
| 197 |
max_length=128 # ✅ especificado para evitar warning
|
| 198 |
).to(device)
|
| 199 |
|
| 200 |
+
# Generando respuesta usando muestreo
|
| 201 |
+
output_ids = model.generate(
|
| 202 |
input_ids=inputs["input_ids"],
|
| 203 |
attention_mask=inputs["attention_mask"], # ✅ FIX agregado
|
| 204 |
max_length=50,
|
| 205 |
+
pad_token_id= tokenizer.eos_token_id,
|
| 206 |
do_sample=True,
|
| 207 |
top_p=0.95,
|
| 208 |
+
top_k=50)
|
|
|
|
|
|
|
|
|
|
|
|
|
| 209 |
|
| 210 |
+
# Decodificando y limpiando la respuesta
|
| 211 |
+
response = tokenizer.decode(output_ids[0], skip_special_tokens=True)
|
| 212 |
+
|
| 213 |
+
return "🤝 [Mori Social] " + response.strip()
|
| 214 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 215 |
|
| 216 |
+
# Funcion para generar respuesta general de Mori
|
| 217 |
+
def contextual_asnwer(question, label_classes, context_model, cont_tok, tec_model, tec_tok, soc_model, soc_tok, device):
|
|
|
|
|
|
|
|
|
|
| 218 |
|
| 219 |
+
'''
|
| 220 |
+
inputs:
|
| 221 |
+
|
| 222 |
+
question --> Pregunta formulada por el usuario
|
| 223 |
+
label_classes --> Clases del label encoder para decodificar inferencias
|
| 224 |
+
context_model --> Clasificador para determinar el contexto de las pregutnas
|
| 225 |
+
cont_tok --> Tokenizer usada para clasificar contextos
|
| 226 |
+
tec_model --> Modelo de Mori para responder preguntas tecnicas
|
| 227 |
+
tec_tok --> Tokenizer usado por Mori Tenico
|
| 228 |
+
soc_model --> Modelo de Mori para responder preguntas sociales
|
| 229 |
+
soc_tok --> Tokenizer usado por Mori Social
|
| 230 |
+
device --> Usar el GPU o el CPU dependiendo de su disponibilidad
|
| 231 |
|
| 232 |
+
outsputs:
|
| 233 |
|
| 234 |
+
response --> Respues de Mori General (Respues con Prompt Engineering)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 235 |
|
| 236 |
+
'''
|
| 237 |
|
| 238 |
+
# Detectar contexto usando el clasificador
|
| 239 |
+
context = classify_context(question, label_classes, context_model, cont_tok, device)
|
| 240 |
+
|
| 241 |
+
context_icons = {"social": "💬",
|
| 242 |
+
"modelos": "🔧",
|
| 243 |
+
"evaluación": "📏",
|
| 244 |
+
"optimización": "⚙️",
|
| 245 |
+
"visualización": "📈",
|
| 246 |
+
"aprendizaje": "🧠",
|
| 247 |
+
"vida digital" : "🧑💻",
|
| 248 |
+
"estadística": "📊",
|
| 249 |
+
"infraestructura": "🖥",
|
| 250 |
+
"datos": "📂",
|
| 251 |
+
"transformación digital": "🌀"}
|
| 252 |
+
|
| 253 |
+
icon = context_icons.get(context, "🧠")
|
| 254 |
+
#print(f"{icon} Contexto detectado: {context}") # (opcional para debug)
|
| 255 |
+
st.markdown(f"**{icon} Contexto detectado:** `{context}`")
|
| 256 |
+
|
| 257 |
+
if context == 'social':
|
| 258 |
+
|
| 259 |
+
# Generar respuesta contextual usando el modelo social
|
| 260 |
+
response = social_asnwer(question, soc_model,soc_tok, device)
|
| 261 |
|
| 262 |
+
else:
|
| 263 |
+
|
| 264 |
+
# Generar respuesta contextual usando el modelo tecnico
|
| 265 |
+
response = technical_asnwer(question, context, tec_model, tec_tok, device)
|
| 266 |
|
| 267 |
+
return response, context
|
| 268 |
|
| 269 |
+
|
|
|
|
|
|
|
| 270 |
|
| 271 |
+
#***************************************************************************
|
| 272 |
+
#MAIN
|
| 273 |
+
#***************************************************************************
|
| 274 |
|
| 275 |
+
if __name__ == '__main__':
|
|
|
|
|
|
|
|
|
|
| 276 |
|
| 277 |
+
# Setting historial for the current user
|
| 278 |
+
if "historial" not in st.session_state:
|
| 279 |
+
st.session_state.historial = []
|
| 280 |
+
|
| 281 |
+
# Addigning a new ID to the current user
|
| 282 |
+
if "user_id" not in st.session_state:
|
| 283 |
+
st.session_state["user_id"] = str(uuid.uuid4())[:8] # Ej: "f6a9b3e2"
|
| 284 |
+
|
| 285 |
+
# Loading classifier encoder classes:
|
| 286 |
+
labels_path = hf_hub_download(repo_id="tecuhtli/mori-context-model", filename="context_labels.pkl")
|
| 287 |
+
label_classes = joblib.load(labels_path)
|
| 288 |
+
|
| 289 |
+
# Loading Saved Models
|
| 290 |
+
# Modelo Contexto
|
| 291 |
+
cont_tok = AutoTokenizer.from_pretrained("tecuhtli/mori-context-model")
|
| 292 |
+
context_model = AutoModelForSeq2SeqLM.from_pretrained("tecuhtli/mori-context-model")
|
| 293 |
+
|
| 294 |
+
# Modelo Técnico
|
| 295 |
+
tec_tok = AutoTokenizer.from_pretrained("tecuhtli/mori-tecnico-model")
|
| 296 |
+
tec_model = AutoModelForSeq2SeqLM.from_pretrained("tecuhtli/mori-tecnico-model")
|
| 297 |
|
| 298 |
+
# Modelo Social
|
| 299 |
+
soc_tok = AutoTokenizer.from_pretrained("tecuhtli/mori-social-model")
|
| 300 |
+
soc_model = AutoModelForSeq2SeqLM.from_pretrained("tecuhtli/mori-social-model")
|
| 301 |
|
| 302 |
+
# Available Device
|
| 303 |
+
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
|
|
|
| 304 |
|
| 305 |
+
# Defining Moris Presetation
|
| 306 |
+
st.title("🤖 Mori - Tu Asistente Personal 🐈")
|
| 307 |
+
st.caption("💬 Puedes preguntarme conceptos técnicos como visualización, limpieza, BI, etc.")
|
| 308 |
+
st.caption("😅 Por el momento, solo puedo contestar preguntas como: ")
|
| 309 |
+
st.caption("🤓 ¿Como estas? ¿Que son?, Explícame algo, Define algo, ¿Para que sirven?")
|
| 310 |
+
st.caption("✏️ Escribe 'salir' para terminar.\n")
|
|
|
|
|
|
|
|
|
|
| 311 |
|
|
|
|
|
|
|
|
|
|
| 312 |
|
| 313 |
+
#entrada_usuario = st.text_area("📝 Escribe tu pregunta aquí")
|
| 314 |
+
with st.form("formulario_mori"):
|
| 315 |
+
user_question = st.text_area("📝 Escribe tu pregunta aquí", key="entrada", height=100)
|
| 316 |
+
submitted = st.form_submit_button("Responder")
|
| 317 |
|
| 318 |
|
| 319 |
+
if submitted:
|
| 320 |
|
| 321 |
+
if not user_question:
|
| 322 |
+
print("Mori: ¿Podrías repetir eso? No entendí bien 😅")
|
| 323 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 324 |
else:
|
| 325 |
+
|
| 326 |
+
#if st.button("Responder") and entrada_usuario:
|
| 327 |
+
response, context = contextual_asnwer(user_question, label_classes, context_model, cont_tok, tec_model, tec_tok, soc_model, soc_tok, device)
|
| 328 |
+
st.success(response)
|
| 329 |
+
|
| 330 |
+
# Guarda en historial
|
| 331 |
+
st.session_state.historial.append(("Mori", response))
|
| 332 |
+
st.session_state.historial.append(("Tú", user_question))
|
| 333 |
+
|
| 334 |
+
# 💾 Guarda en archivo para stats/dataset
|
| 335 |
+
saving_interaction(user_question, response, context, st.session_state["user_id"])
|
| 336 |
+
|
| 337 |
+
|
| 338 |
+
# 🔁 Muestra historial
|
| 339 |
+
if st.session_state.historial:
|
| 340 |
+
st.markdown("---")
|
| 341 |
+
for autor, texto in reversed(st.session_state.historial):
|
| 342 |
+
if autor == "Tú":
|
| 343 |
+
st.markdown(f"🧍♂️ **{autor}**: {texto}")
|
| 344 |
+
else:
|
| 345 |
+
st.markdown(f"🤖 **{autor}**: {texto}")
|
| 346 |
+
|
| 347 |
+
|
| 348 |
+
if st.session_state.historial:
|
| 349 |
+
texto_chat = ""
|
| 350 |
+
for autor, texto in st.session_state.historial:
|
| 351 |
+
texto_chat += f"{autor}: {texto}\n\n"
|
| 352 |
|
| 353 |
+
st.download_button(
|
| 354 |
+
label="💾 Descargar conversación como .txt",
|
| 355 |
+
data=texto_chat,
|
| 356 |
+
file_name="conversacion_mori.txt",
|
| 357 |
+
mime="text/plain")
|
| 358 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 359 |
|
| 360 |
+
#***************************************************************************
|
| 361 |
+
#FIN
|
| 362 |
+
#***************************************************************************
|
|
|
|
|
|
|
|
|