Update app.py

app.py

@@ -34,30 +34,32 @@ Este análisis es una herramienta informativa de IA y **NO es un diagnóstico m
 **Si tus síntomas son severos, busca ayuda médica urgente.**
 """
 
-# --- CONFIGURACIÓN DE LA API GEMINI ---
-try:
-    if 'APIkeydetective' in st.secrets:
-        APIkeydetective = st.secrets['APIkeydetective']
-    else:
-        APIkeydetective = os.getenv("APIkeydetective")
-    if not APIkeydetective:
-        st.error("No se encontró la APIkeydetective.")
-        st.stop()
-    
-    client = genai.Client(
-        api_key=APIkeydetective,
-        http_options=types.HttpOptions(timeout=30000)
-    )
-
-except Exception as e:
-    st.error(f"❌ Error al configurar Gemini API: {e}")
-    st.stop()
-
+# --- CONFIGURACIÓN DE LA API GEMINI (google-genai V3.0) ---
 @st.cache_resource
 def get_gemini_client():
-    return client
+    try:
+        if 'APIkeydetective' in st.secrets:
+            APIkeydetective = st.secrets['APIkeydetective']
+        else:
+            APIkeydetective = os.getenv("APIkeydetective")
+            
+        if not APIkeydetective:
+            st.error("No se encontró la APIkeydetective.")
+            return None
+        
+        return genai.Client(
+            api_key=APIkeydetective,
+            http_options=types.HttpOptions(timeout=30000) # 30 segundos
+        )
+    except Exception as e:
+        st.error(f"❌ Error al configurar Gemini API: {e}")
+        return None
 
-# --- CARGA ULTRA-RÁPIDA DE DATOS SIN COPIA EN MEMORIA ---
+client = get_gemini_client()
+if not client:
+    st.stop()
+
+# --- CARGA ULTRA-RÁPIDA DE DATOS ---
 @st.cache_resource
 def load_data():
     try:
@@ -110,7 +112,7 @@ FOOD_TO_COMPOUND_MAP = {
     "cholado": ["fructosa", "lactosa", "histamina", "azúcar refinado"]
 }
 
-# --- SEGURO LÓGICO: REGLAS ESTRICTAS DE EXCLUSIÓN BIOQUÍMICA ---
+# --- SEGURO LÓGICO DE INCOMPATIBILIDAD BIOQUÍMICA ---
 STRICT_COMPOUND_TRIGGERS = {
     "cafeína": ["cafeína", "caffeine", "café"],
     "gluten": ["gluten", "trigo", "cebada", "centeno"],
@@ -122,12 +124,6 @@ STRICT_COMPOUND_TRIGGERS = {
 }
 
 def validar_seguro_logico(entry, user_foods, food_compound_map):
-    """
-    SEGURO LÓGICO: Si una condición atribuida en la BD o hipótesis requiere un
-    compuesto exclusivo (ej: cafeína, gluten, lactosa, tiramina), verifica que AL MENOS
-    UNO de los alimentos ingresados por el usuario contenga efectivamente dicho compuesto.
-    Evita falsos positivos como asociar cafeína al jugo de naranja.
-    """
     if not user_foods:
         return True
 
@@ -275,7 +271,7 @@ INTEGRATED_NEURO_FOOD_MAP = {
     }
 }
 
-# --- FUNCIONES DE AUXILIO ---
+# --- FUNCIONES DE APOYO Y EXTRACCIÓN ---
 @st.cache_data(show_spinner=False)
 def generate_word_report_cached(report_text):
     try:
@@ -321,7 +317,9 @@ def extract_keywords(text):
 
 @retry(wait=wait_fixed(2), stop=stop_after_attempt(2))
 def extract_entities_with_gemini(query):
-    if not client: return None
+    gemini_client = get_gemini_client()
+    if not gemini_client: return {"alimentos": [], "sintomas": []}
+    
     system_prompt = f"""
     Rol: Extractor Médico JSON experto en nutrición y síntomas.
     Tarea: Analiza el texto del usuario y extrae un JSON estricto con dos claves:
@@ -332,7 +330,7 @@ def extract_entities_with_gemini(query):
     Output (JSON estricto):
     """
     try:
-        response = client.models.generate_content(
+        response = gemini_client.models.generate_content(
             model='gemini-2.5-flash-lite',
             contents=system_prompt
         )
@@ -395,7 +393,7 @@ def find_best_matches_hybrid(entities, data, food_compound_map=None):
 
     results = []
     for entry in data:
-        # APLICACIÓN DEL SEGURO LÓGICO
+        # SEGURO LÓGICO
         if food_compound_map and not validar_seguro_logico(entry, user_foods_list, food_compound_map):
             continue
 
@@ -497,7 +495,8 @@ def find_best_foodb_matches(user_foods_es, _foodb_index, _foodb_lower_map, food_
 
 @retry(wait=wait_fixed(2), stop=stop_after_attempt(2))
 def generate_detailed_analysis(query, match, entities=None, food_compound_map=None, neuro_map=None):
-    if not client: return "Error: IA no disponible."
+    gemini_client = get_gemini_client()
+    if not gemini_client: return "Error: IA no disponible."
 
     compound_context = ""
     if entities and food_compound_map and neuro_map:
@@ -528,7 +527,7 @@ def generate_detailed_analysis(query, match, entities=None, food_compound_map=No
     NO incluyas descargos de responsabilidad.
     """
     try:
-        response = client.models.generate_content(
+        response = gemini_client.models.generate_content(
             model='gemini-2.5-flash-lite',
             contents=prompt
         )
@@ -604,7 +603,14 @@ def create_relevance_chart(results):
         x='Relevancia', y=alt.Y('Condición', sort='-x'), tooltip=['Condición', 'Relevancia']
     ).properties(title='Top Coincidencias')
 
-# --- INTERFAZ STREAMLIT ---
+# --- INICIALIZACIÓN DE ESTADO PERSISTENTE ---
+if 'search_results' not in st.session_state: st.session_state.search_results = None
+if 'user_query' not in st.session_state: st.session_state.user_query = ""
+if 'entities' not in st.session_state: st.session_state.entities = None
+if 'analysis_cache' not in st.session_state: st.session_state.analysis_cache = {}
+if 'has_searched' not in st.session_state: st.session_state.has_searched = False
+
+# --- INTERFAZ GRAFICA STREAMLIT ---
 col_img1, col_text, col_img2 = st.columns([1, 4, 1])
 with col_img1:
     if os.path.exists("imagen.png"): st.image("imagen.png", width=150)
@@ -614,77 +620,86 @@ with col_text:
 with col_img2:
     if os.path.exists("buho.png"): st.image("buho.png", width=120)
 
-if 'search_results' not in st.session_state: st.session_state.search_results = None
-if 'user_query' not in st.session_state: st.session_state.user_query = ""
-if 'entities' not in st.session_state: st.session_state.entities = None
-if 'analysis_cache' not in st.session_state: st.session_state.analysis_cache = {}
-
-with st.form(key="search_form"):
-    query = st.text_area("Tu Caso:", height=150, placeholder="Ej: Cuando tomo mucho jugo de naranja no puedo dormir.")
+with st.form(key="search_form", clear_on_submit=False):
+    query_input = st.text_area(
+        "Tu Caso:", 
+        height=150, 
+        placeholder="Ej: Cuando tomo mucho jugo de naranja no puedo dormir.",
+        key="input_query_text"
+    )
     submitted = st.form_submit_button("Analizar Caso", type="primary")
 
-if submitted and query:
-    st.session_state.user_query = query
-    st.session_state.search_results = None
-    st.session_state.analysis_cache = {}
-    
-    with st.spinner("🔍 Analizando pistas..."):
-        raw_entities = extract_entities_with_gemini(query)
-        reinforced = reinforce_entities_with_keywords(raw_entities, query, FOOD_TO_COMPOUND_MAP, MASTER_SYMPTOM_MAP)
-        final_symptoms = translate_symptoms_local(reinforced.get("sintomas", []), MASTER_SYMPTOM_MAP)
-        final_entities = {"alimentos": reinforced.get("alimentos", []), "sintomas": final_symptoms}
-        st.session_state.entities = final_entities
+if submitted:
+    if query_input and query_input.strip():
+        current_query = query_input.strip()
+        st.session_state.user_query = current_query
+        st.session_state.has_searched = True
+        st.session_state.analysis_cache = {} # Limpiar análisis previo
+        st.session_state.search_results = None
+        st.session_state.entities = None
         
-        results = find_best_matches_hybrid(final_entities, alimentos_data, FOOD_TO_COMPOUND_MAP)
-
-        if not results:
-            local_hypothesis = build_local_hypothesis(final_entities, FOOD_TO_COMPOUND_MAP, INTEGRATED_NEURO_FOOD_MAP)
-            if local_hypothesis:
-                results = [{'entry': local_hypothesis, 'score': {'total': 50}, 'matched_keywords': []}]
-
-        st.session_state.search_results = results
+        with st.spinner("🔍 Analizando pistas..."):
+            raw_entities = extract_entities_with_gemini(current_query)
+            reinforced = reinforce_entities_with_keywords(raw_entities, current_query, FOOD_TO_COMPOUND_MAP, MASTER_SYMPTOM_MAP)
+            final_symptoms = translate_symptoms_local(reinforced.get("sintomas", []), MASTER_SYMPTOM_MAP)
+            final_entities = {"alimentos": reinforced.get("alimentos", []), "sintomas": final_symptoms}
+            st.session_state.entities = final_entities
+            
+            results = find_best_matches_hybrid(final_entities, alimentos_data, FOOD_TO_COMPOUND_MAP)
+
+            if not results:
+                local_hypothesis = build_local_hypothesis(final_entities, FOOD_TO_COMPOUND_MAP, INTEGRATED_NEURO_FOOD_MAP)
+                if local_hypothesis:
+                    results = [{'entry': local_hypothesis, 'score': {'total': 50}, 'matched_keywords': []}]
+
+            st.session_state.search_results = results if results else []
+    else:
+        st.warning("⚠️ Por favor describe tus síntomas o alimentos consumidos antes de analizar.")
 
-if st.session_state.search_results:
+# --- DISPLAY DE RESULTADOS INDEPENDIENTE Y REUTILIZABLE ---
+if st.session_state.has_searched:
     results = st.session_state.search_results
-    best_match = results[0]['entry']
-    
-    st.success(f"🔎 Coincidencia Principal: **{best_match.get('condicion_asociada')}**")
     
-    cache_key = f"analysis_{best_match.get('condicion_asociada')}"
-    if cache_key not in st.session_state.analysis_cache:
-        with st.spinner("✍️ Redactando informe..."):
-            analysis = generate_detailed_analysis(
-                st.session_state.user_query, best_match,
-                st.session_state.entities, FOOD_TO_COMPOUND_MAP, INTEGRATED_NEURO_FOOD_MAP
+    if results and len(results) > 0:
+        best_match = results[0]['entry']
+        st.success(f"🔎 Coincidencia Principal: **{best_match.get('condicion_asociada')}**")
+        
+        # Clave de caché única por consulta y condición
+        cache_key = f"analysis_{hash(st.session_state.user_query)}_{best_match.get('condicion_asociada')}"
+        if cache_key not in st.session_state.analysis_cache:
+            with st.spinner("✍️ Redactando informe..."):
+                analysis = generate_detailed_analysis(
+                    st.session_state.user_query, best_match,
+                    st.session_state.entities, FOOD_TO_COMPOUND_MAP, INTEGRATED_NEURO_FOOD_MAP
+                )
+                st.session_state.analysis_cache[cache_key] = analysis
+
+        tab_main, tab_neuro, tab_mol = st.tabs(["💡 Interpretación Clínica", "🧠 Efectos Neuropsicológicos", "🔬 Análisis Molecular"])
+        
+        with tab_main:
+            st.markdown(st.session_state.analysis_cache[cache_key])
+            if len(results) > 1:
+                st.altair_chart(create_relevance_chart(results), use_container_width=True)
+
+        with tab_neuro:
+            st.markdown(generate_neuro_report_text(st.session_state.entities, FOOD_TO_COMPOUND_MAP, INTEGRATED_NEURO_FOOD_MAP))
+
+        with tab_mol:
+            st.markdown(generate_molecular_report_text(
+                best_match, st.session_state.entities, foodb_index, foodb_lower_map,
+                FOOD_NAME_TO_FOODB_KEY, COMPOUND_SYNONYM_MAP, FOOD_TO_COMPOUND_MAP, INTEGRATED_NEURO_FOOD_MAP
+            ))
+
+        full_report = f"REPORTE\n\n{st.session_state.analysis_cache[cache_key]}"
+        word_bytes = generate_word_report_cached(full_report)
+        if word_bytes:
+            st.download_button(
+                label="📄 Descargar Informe",
+                data=word_bytes,
+                file_name="Reporte.docx",
+                mime="application/vnd.openxmlformats-officedocument.wordprocessingml.document",
+                key=f"dl_btn_{hash(st.session_state.user_query)}"
             )
-            st.session_state.analysis_cache[cache_key] = analysis
-
-    tab_main, tab_neuro, tab_mol = st.tabs(["💡 Interpretación Clínica", "🧠 Efectos Neuropsicológicos", "🔬 Análisis Molecular"])
-    
-    with tab_main:
-        st.markdown(st.session_state.analysis_cache[cache_key])
-        if len(results) > 1:
-            st.altair_chart(create_relevance_chart(results), use_container_width=True)
-
-    with tab_neuro:
-        st.markdown(generate_neuro_report_text(st.session_state.entities, FOOD_TO_COMPOUND_MAP, INTEGRATED_NEURO_FOOD_MAP))
-
-    with tab_mol:
-        st.markdown(generate_molecular_report_text(
-            best_match, st.session_state.entities, foodb_index, foodb_lower_map,
-            FOOD_NAME_TO_FOODB_KEY, COMPOUND_SYNONYM_MAP, FOOD_TO_COMPOUND_MAP, INTEGRATED_NEURO_FOOD_MAP
-        ))
-
-    full_report = f"REPORTE\n\n{st.session_state.analysis_cache[cache_key]}"
-    word_bytes = generate_word_report_cached(full_report)
-    if word_bytes:
-        st.download_button(
-            label="📄 Descargar Informe",
-            data=word_bytes,
-            file_name="Reporte.docx",
-            mime="application/vnd.openxmlformats-officedocument.wordprocessingml.document"
-        )
-
-elif submitted:
-    st.warning("⚠️ No pudimos identificar una causa clara.")
-    st.info("Intenta reformular: 'Siento [SÍNTOMA] cuando como [ALIMENTO]'.")
+    else:
+        st.warning("⚠️ No pudimos identificar una causa clara para este caso específico.")
+        st.info("Intenta reformular especificando el alimento y el síntoma. Ej: 'Siento [SÍNTOMA] cuando como [ALIMENTO]'.")