Update app.py

app.py

@@ -1,64 +1,383 @@
 import gradio as gr
-from huggingface_hub import InferenceClient
+import time
+import os
+import glob
+import torch
+import gc
+import re
+import json
+from pathlib import Path
+from omegaconf import OmegaConf
 
-"""
-For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
-"""
-client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
+from rag.constants.default_prompt import PROMT_USER_INTENT_TREE
+from rag.reader.directory_reader import DirectoryReader
+from rag.pipeline.milvus_bm25_retriever import HybridSearchRetrieverPipeline
+from rag.pipeline.contextual_generator_qa import ContextualQuestionGeneratorPipeline
+from rag.synthesizer.types import StreamingResponse
+from rag.config.configuration import ConfigurationManager
+
+torch.cuda.empty_cache()
+gc.collect()
+
+# === Cargar PDFs
+pdf_files = glob.glob(os.path.abspath(os.curdir) + "/data/*.pdf")
+reader = DirectoryReader(input_files=pdf_files)
+documents = reader.load_data()
 
+# === Configuración
+config = OmegaConf.load(Path("configs/config.yaml"))
+manager = ConfigurationManager(config)
 
-def respond(
-    message,
-    history: list[tuple[str, str]],
-    system_message,
-    max_tokens,
-    temperature,
-    top_p,
-):
-    messages = [{"role": "system", "content": system_message}]
+# === Inicializar Pipelines
+pipeline = HybridSearchRetrieverPipeline(
+    documents=documents,
+    splitter_config=manager.get_splitter_config(),
+    index_retriver_config=manager.get_index_retriever_config(),
+    embed_config=manager.get_embed_config(),
+    milvus_config=manager.get_db_config(),
+    rerank_config=manager.get_rerank_config(),
+    response_config=manager.get_response_config(),
+    llm_config=manager.get_llm_config()
+)
+
+qgen_pipeline = ContextualQuestionGeneratorPipeline(
+    llm_config=manager.get_llm_config(),
+    embed_config=manager.get_embed_config(),
+    splitter_config=manager.get_splitter_config(),
+    qgen_config=manager.get_question_gen_config(),
+)
 
-    for val in history:
-        if val[0]:
-            messages.append({"role": "user", "content": val[0]})
-        if val[1]:
-            messages.append({"role": "assistant", "content": val[1]})
 
-    messages.append({"role": "user", "content": message})
+# === Intención básica
+def handle_intent(etiqueta: str, motivo: str) -> str:
+    if etiqueta.lower() == "neutro":
+        if "saludo" in motivo.lower():
+            return "👋 ¡Hola! ¿En qué puedo ayudarte hoy?"
+        elif "despedida" in motivo.lower():
+            return "👋 ¡Hasta luego!"
+        elif "agradecimiento" in motivo.lower():
+            return "😊 ¡Con gusto!"
+        else:
+            return "🙂 Gracias por tu mensaje."
+    elif etiqueta.lower() == "negativo":
+        return "⚠️ No puedo responder a eso. Por favor, intenta con algo relacionado al estatuto."
+    return ""
 
-    response = ""
 
-    for message in client.chat_completion(
-        messages,
-        max_tokens=max_tokens,
-        stream=True,
-        temperature=temperature,
-        top_p=top_p,
-    ):
-        token = message.choices[0].delta.content
+DEV_MODE = False  # Cambia a True si quieres ver rutas de archivo en las fuentes
 
-        response += token
-        yield response
+
+def extract_sources(nodes):
+    fuentes = []
+    for i, node in enumerate(nodes):
+        meta = node.metadata or {}
+        fuente = {
+            "documento": meta.get("file_name", "Desconocido"),
+            "página": meta.get("num_page", "¿?"),
+            "autor": meta.get("Author", "Autor desconocido"),
+            "fecha": meta.get("creation_date", "Sin fecha"),
+            "ruta": meta.get("file_path", "No disponible")
+        }
+
+        fuente_md = (
+            f"**[{i + 1}] {fuente['documento']}**\n"
+            f"> 📄 Página: `{fuente['página']}`\n"
+            f"> ✍️ Autor: *{fuente['autor']}*\n"
+            f"> 🗓️ Fecha de creación: {fuente['fecha']}"
+        )
+
+        if DEV_MODE:
+            fuente_md += f"\n> 🧪 Ruta (debug): `{fuente['ruta']}`"
+
+        fuentes.append(fuente_md)
+
+    return "\n\n".join(fuentes) if fuentes else "⚠️ No se encontraron fuentes relevantes."
+
+
+def detectar_intencion(query: str) -> str:
+    try:
+        result = pipeline.service_context.llm.predict(
+            prompt=PROMT_USER_INTENT_TREE,
+            query_str=query
+        )
+        cleaned = re.sub(r"```(?:json)?|```", "", result).strip()
+        parsed = json.loads(cleaned)
+        etiqueta = parsed.get("etiqueta", "").lower()
+        motivo = parsed.get("motivo", "").lower()
+        return handle_intent(etiqueta, motivo)
+    except Exception as e:
+        print("⚠️ Error al detectar intención:", e)
+        return ""
+
+
+def bot(history, use_intent, use_subq):
+    user_msg = next(
+        (m["content"] for m in reversed(history) if m["role"] == "user" and isinstance(m["content"], str)),
+        None
+    )
+    if not user_msg:
+        return history, gr.update(visible=False), *[gr.update(visible=False)] * 10
+
+    try:
+        if use_intent:
+            intent_reply = detectar_intencion(user_msg)
+            if intent_reply:
+                history.append({"role": "assistant", "content": ""})
+                for char in intent_reply:
+                    history[-1]["content"] += char
+                    time.sleep(0.05)
+                    yield history, gr.update(visible=False), *[gr.update(visible=False)] * 10
+                return
+    except Exception as e:
+        print("⚠️ Error interpretando intención:", e)
+
+    response, nodes = pipeline.main(user_msg)
+    history.append({"role": "assistant", "content": ""})
+
+    if isinstance(response, StreamingResponse):
+        for char in response.response_gen:
+            history[-1]["content"] += char
+            time.sleep(0.02)
+            yield history, gr.update(visible=False), *[gr.update(visible=False)] * 10
+
+        subquestions = qgen_pipeline.run(user_msg, nodes) if use_subq else []
+
+        fuentes_md = extract_sources(nodes)
+        history[-1]["sources"] = fuentes_md
+
+        subq_updates = [
+            gr.update(value=q, visible=True) if i < len(subquestions) else gr.update(visible=False)
+            for i, q in enumerate(subquestions + [""] * (10 - len(subquestions)))
+        ]
+
+        yield history, gr.update(value=f"📚 **Fuentes:**\n{fuentes_md}", visible=True), *subq_updates
+
+
+def add_message(history, msg_text):
+    history.append({"role": "user", "content": msg_text})
+    return history, ""
+
+
+# === Toggle panel visibility
+def toggle_config(is_visible):
+    return not is_visible, gr.update(visible=not is_visible)
+
+
+# === Estilo CSS
+css = """
+body {
+    background-color: #f8fafc;
+    font-family: 'Segoe UI', sans-serif;
+}
+
+.gr-chatbot {
+    background-color: #ffffff;
+    border: 1px solid #d1fae5;
+    border-radius: 12px;
+    padding: 10px;
+    box-shadow: 0 4px 12px rgba(0,0,0,0.03);
+}
+
+.gr-textbox {
+    border-radius: 10px !important;
+    border: 1px solid #a7f3d0;
+    padding: 10px;
+    flex-grow: 1;
+}
+
+.gr-button.enviar {
+    background-color: #2ecc71;
+    color: white;
+    font-weight: 600;
+    border-radius: 10px !important;
+    padding: 10px 18px;
+    margin-left: 8px;
+    transition: background-color 0.3s ease;
+}
+.gr-button.enviar:hover {
+    background-color: #27ae60;
+}
+
+.gr-markdown {
+    background-color: #ecfdf5;
+    border-left: 4px solid #10b981;
+    padding: 10px;
+    margin-top: 10px;
+    border-radius: 8px;
+    color: #064e3b;
+    font-size: 15px;
+}
+
+.gr-button.subq {
+    background-color: #d1fae5;
+    border: none;
+    padding: 8px 14px;
+    border-radius: 9999px;
+    cursor: pointer;
+    font-size: 14px;
+    margin: 4px 6px 0 0;
+    transition: background-color 0.3s, transform 0.1s;
+    box-shadow: 0 1px 2px rgba(0,0,0,0.05);
+}
+.gr-button.subq:hover {
+    background-color: #a7f3d0;
+    transform: scale(1.05);
+}
+
+#floating-config-wrapper {
+    position: fixed;
+    bottom: 20px;
+    right: 20px;
+    z-index: 9999;
+    display: inline-flex;
+    flex-direction: column;
+    align-items: flex-end;
+    width: auto;           /* 👈 no ocupar ancho completo */
+    max-width: 250px;      /* 👈 ancho controlado */
+    # pointer-events: none;  /* 👈 deja pasar clics fuera del botón y panel */
+}
+
+#config-btn {
+    all: unset;                      /* 🚫 Quita todos los estilos por defecto */
+    display: flex;
+    justify-content: center;
+    align-items: center;
+    width: 40px;
+    height: 40px;
+    font-size: 22px;
+    background-color: transparent;  /* 🟢 Sin fondo */
+    color: #2ecc71;                 /* 🟢 Verde icono */
+    border-radius: 50%;
+    cursor: pointer;
+    transition: background-color 0.2s ease;
+}
+
+#config-btn:hover {
+    background-color: #d1fae5;       /* 🟢 Suave fondo al pasar el mouse */
+}
+
+
+#config-panel {
+    background-color: #ecfdf5;
+    border: 1px solid #a7f3d0;
+    border-radius: 10px;
+    padding: 12px;
+    min-width: 220px;
+    box-shadow: 0 6px 16px rgba(0,0,0,0.1);
+    font-size: 14px;
+    color: #064e3b;
+}
+
+#zonia-header h1 {
+    font-family: 'Segoe UI', sans-serif;
+    font-weight: bold;
+}
+
+#zonia-header p {
+    font-family: 'Segoe UI', sans-serif;
+}
+
+.gr-chatbot {
+    background: white;
+    border-radius: 16px;
+    border: 1px solid #d1fae5;
+    box-shadow: 0 4px 14px rgba(0,0,0,0.05);
+    padding: 12px;
+    margin-bottom: 12px;
+}
+
+.gr-chatbot .message.user {
+    background-color: #e0f2f1;
+    border-radius: 12px;
+    padding: 10px;
+    margin: 6px 0;
+    font-weight: 500;
+    color: #065f46;
+}
+
+.gr-chatbot .message.assistant {
+    background-color: #ecfdf5;
+    border-radius: 12px;
+    padding: 10px;
+    margin: 6px 0;
+    color: #065f46;
+    border-left: 4px solid #10b981;
+}
+
+.gr-chatbot .message {
+    font-family: 'Segoe UI', sans-serif;
+    font-size: 15px;
+    line-height: 1.6;
+}
+
+.gr-chatbot .avatar {
+    display: none; /* Oculta los avatares si los hay */
+}
 
 
 """
-For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
-"""
-demo = gr.ChatInterface(
-    respond,
-    additional_inputs=[
-        gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
-        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
-        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
-        gr.Slider(
-            minimum=0.1,
-            maximum=1.0,
-            value=0.95,
-            step=0.05,
-            label="Top-p (nucleus sampling)",
-        ),
-    ],
-)
 
+# === Interfaz
+with gr.Blocks(css=css) as app:
+    gr.Markdown(
+        """
+        <div style='padding: 1rem 1.5rem; border-radius: 12px; background: #ecfdf5; border-left: 5px solid #10b981; margin-bottom: 1.5rem'>
+            <h1 style='margin: 0; font-size: 26px; color: #065f46;'>🤖 ZONIA</h1>
+            <p style='margin: 0.3rem 0 0; font-size: 15px; color: #065f46;'>Un asistente inteligente para consultar estatutos de posgrado y reglamentos estudiantiles. Esta interfaz es una prueba para evaluación de usuarios.</p>
+        </div>
+        """,
+        elem_id="zonia-header"
+    )
+
+    config_visible = gr.State(False)
+    toggle_intent = gr.Checkbox(visible=False, value=True, label="intent_flag")
+    toggle_subq = gr.Checkbox(visible=False, value=True, label="subq_flag")
+
+    chatbot = gr.Chatbot(label="Conversación", type="messages", show_copy_button=True)
+    state = gr.State([])
+
+    sources_md = gr.Markdown(visible=False)
+    subq_buttons = [gr.Button(visible=False, elem_classes="subq") for _ in range(10)]
+
+    # Input + enviar
+    with gr.Row():
+        msg = gr.Textbox(placeholder="Haz tu pregunta...")
+        enviar = gr.Button("Enviar", elem_classes="enviar")
+
+    # Botón de configuración y panel flotante
+    with gr.Group(elem_id="floating-config-wrapper"):
+        show_config_btn = gr.Button("⚙️", elem_id="config-btn")
+        with gr.Column(visible=False, elem_id="config-panel") as config_panel:
+            gr.Markdown("### ⚙️ Configuración")
+            intent_checkbox = gr.Checkbox(label="Activar intención", value=True)
+            subq_checkbox = gr.Checkbox(label="Activar subpreguntas", value=True)
+
+
+    # Función toggle y sync
+    def toggle_panel(is_visible): return not is_visible, gr.update(visible=not is_visible)
+
+
+    def sync_flags(i, s): return i, s
+
+
+    show_config_btn.click(toggle_panel, inputs=[config_visible], outputs=[config_visible, config_panel])
+    intent_checkbox.change(sync_flags, [intent_checkbox, subq_checkbox], [toggle_intent, toggle_subq])
+    subq_checkbox.change(sync_flags, [intent_checkbox, subq_checkbox], [toggle_intent, toggle_subq])
+
+    # Submit por enter
+    msg.submit(add_message, [state, msg], [state, msg]).then(
+        bot, [state, toggle_intent, toggle_subq], [chatbot, sources_md] + subq_buttons
+    )
+
+    # Submit por botón
+    enviar.click(add_message, [state, msg], [state, msg]).then(
+        bot, [state, toggle_intent, toggle_subq], [chatbot, sources_md] + subq_buttons
+    )
+
+    # Subpreguntas
+    for btn in subq_buttons:
+        btn.click(lambda txt, h: add_message(h, txt), [btn, state], [state, msg]).then(
+            bot, [state, toggle_intent, toggle_subq], [chatbot, sources_md] + subq_buttons
+        )
 
-if __name__ == "__main__":
-    demo.launch()
+app.launch(True)