v3 / modules /semantic /semantic_interface_afterParty.py
AIdeaText's picture
Upload 216 files
c58df45 verified
raw
history blame
4.26 kB
import streamlit as st
import logging
from io import BytesIO
import base64
from .semantic_float_reset import semantic_float_init, float_graph, toggle_float_visibility, update_float_content
from .semantic_process import process_semantic_analysis
from ..chatbot.chatbot import initialize_chatbot, process_semantic_chat_input
from ..database.database_oldFromV2 import (
initialize_mongodb_connection,
initialize_database_connections,
create_admin_user,
create_student_user,
get_user,
get_student_data,
store_file_contents,
retrieve_file_contents,
get_user_files,
delete_file,
store_application_request,
store_user_feedback,
store_morphosyntax_result,
store_semantic_result,
store_discourse_analysis_result,
store_chat_history,
export_analysis_and_chat,
get_user_analysis_summary,
get_user_recents_chats,
get_user_analysis_details
)
from ..utils.widget_utils import generate_unique_key
from .flexible_analysis_handler import FlexibleAnalysisHandler
semantic_float_init()
logging.basicConfig(level=logging.DEBUG)
logger = logging.getLogger(__name__)
def get_translation(t, key, default):
return t.get(key, default)
def fig_to_base64(fig):
buf = BytesIO()
fig.savefig(buf, format='png')
buf.seek(0)
img_str = base64.b64encode(buf.getvalue()).decode()
return f'<img src="data:image/png;base64,{img_str}" />'
def display_semantic_interface(lang_code, nlp_models, t):
st.set_page_config(layout="wide")
if 'semantic_chatbot' not in st.session_state:
st.session_state.semantic_chatbot = initialize_chatbot('semantic')
if 'semantic_chat_history' not in st.session_state:
st.session_state.semantic_chat_history = []
if 'show_graph' not in st.session_state:
st.session_state.show_graph = False
if 'graph_id' not in st.session_state:
st.session_state.graph_id = None
st.header(t['title'])
# Opción para introducir texto
text_input = st.text_area(
t['text_input_label'],
height=150,
placeholder=t['text_input_placeholder'],
)
# Opción para cargar archivo
uploaded_file = st.file_uploader(t['file_uploader'], type=['txt'])
if st.button(t['analyze_button']):
if text_input or uploaded_file is not None:
if uploaded_file:
text_content = uploaded_file.getvalue().decode('utf-8')
else:
text_content = text_input
# Realizar el análisis
analysis_result = process_semantic_analysis(text_content, nlp_models[lang_code], lang_code)
# Guardar el resultado en el estado de la sesión
st.session_state.semantic_result = analysis_result
# Mostrar resultados
display_semantic_results(st.session_state.semantic_result, lang_code, t)
# Guardar el resultado del análisis
if store_semantic_result(st.session_state.username, text_content, analysis_result):
st.success(t['success_message'])
else:
st.error(t['error_message'])
else:
st.warning(t['warning_message'])
elif 'semantic_result' in st.session_state:
# Si hay un resultado guardado, mostrarlo
display_semantic_results(st.session_state.semantic_result, lang_code, t)
else:
st.info(t['initial_message']) # Asegúrate de que 'initial_message' esté en tus traducciones
def display_semantic_results(result, lang_code, t):
if result is None:
st.warning(t['no_results']) # Asegúrate de que 'no_results' esté en tus traducciones
return
# Mostrar conceptos clave
with st.expander(t['key_concepts'], expanded=True):
concept_text = " | ".join([f"{concept} ({frequency:.2f})" for concept, frequency in result['key_concepts']])
st.write(concept_text)
# Mostrar el gráfico de relaciones conceptuales
with st.expander(t['conceptual_relations'], expanded=True):
st.pyplot(result['relations_graph'])