Update app.py

app.py

@@ -1,107 +1,146 @@
-import os
-import tempfile
-from pathlib import Path
-import gradio as gr
-
-from conversation_storyline.pipeline import run_pipeline_from_text
-
-
-THEME = gr.themes.Soft(primary_hue="blue").set(
-    body_background_fill="*neutral_50",
-    block_background_fill="*neutral_100",
-)
-
-TITLE = "Conversation Storyline Visualizer – v4"
-
-
-def process_transcript(transcript: str, model_selector: str):
-    """
-    Entrada: texto pegado (transcripción)
-    Salida: storyline.png + resumen + figs plotly
-    """
-    if not transcript or not transcript.strip():
-        raise gr.Error("Pega una transcripción en el cuadro de texto.")
-
-    outdir = Path(tempfile.mkdtemp(prefix="storyline_v4_"))
-
-    outputs = run_pipeline_from_text(
-        transcript_text=transcript,
-        out_dir=outdir,
-        openai_model=model_selector,
-    )
-
-    return (
-        str(outputs["storyline_png"]),
-        outputs["summary_text"],
-        outputs["fig_sentiment"],
-        outputs["fig_bump"],
-        outputs["fig_heatmap"],
-        outputs["fig_hist_reply_dist"],
-        outputs["fig_sankey"],
-        outputs["storyline_html"],
-        outputs["metrics_csv"],
-        outputs["interactions_jsonl"],
-        outputs["graph_json"],
-    )
-
-
-with gr.Blocks(title=TITLE, theme=THEME) as demo:
-    gr.Markdown("# Visualización Narrativa Avanzada de Conversaciones (v4)")
-    gr.Markdown(
-        "- Pega una transcripción tipo `Speaker A: ...`\n"
-        "- Soporta conversaciones largas (chunking)\n"
-        "- Reply-to Top-K embeddings + topic shifts ruptures + layout OR-Tools\n"
-    )
-
-    with gr.Row():
-        model_selector = gr.Dropdown(
-            choices=[
-                "gpt-4o-2024-08-06",
-                "gpt-4o-mini-2024-07-18",
-                "none (offline)",
-            ],
-            value="none (offline)",
-            label="Modelo (opcional; si hay OPENAI_API_KEY)",
-        )
-
-    input_text = gr.Textbox(label="Transcripción", lines=20, placeholder="Pega aquí la transcripción...")
-    btn = gr.Button("Generar Visualizaciones", variant="primary")
-
-    with gr.Tabs():
-        with gr.Tab("Storyline Principal"):
-            main_img = gr.Image(label="Storyline (PNG)")
-            summary_box = gr.Textbox(label="Resumen", lines=10)
-            storyline_html = gr.HTML(label="Storyline (HTML embebido)")
-        with gr.Tab("Análisis Detallado"):
-            sentiment_plot = gr.Plot(label="Sentiment (si aplica)")
-            bump_plot = gr.Plot(label="Bump actividad por segmento")
-            heatmap_plot = gr.Plot(label="Heatmap interacciones")
-            hist_plot = gr.Plot(label="Histograma distancia reply_to")
-            sankey_plot = gr.Plot(label="Sankey Speaker → Topic")
-
-        with gr.Tab("Descargas"):
-            metrics_csv = gr.File(label="metrics.csv")
-            interactions_jsonl = gr.File(label="interactions.jsonl")
-            graph_json = gr.File(label="graph.json")
-
-    btn.click(
-        fn=process_transcript,
-        inputs=[input_text, model_selector],
-        outputs=[
-            main_img,
-            summary_box,
-            sentiment_plot,
-            bump_plot,
-            heatmap_plot,
-            hist_plot,
-            sankey_plot,
-            storyline_html,
-            metrics_csv,
-            interactions_jsonl,
-            graph_json,
-        ],
-    )
-
-if __name__ == "__main__":
-    # HF Spaces
-    demo.launch(server_name="0.0.0.0", server_port=int(os.getenv("PORT", "7860")))
+from __future__ import annotations
+
+import asyncio
+import tempfile
+from pathlib import Path
+from typing import Optional
+
+import gradio as gr
+
+from conversation_storyline.io import load_messages, load_messages_from_text
+from conversation_storyline.pipeline import run_pipeline
+from conversation_storyline.plots import (
+    load_graph_json,
+    load_interactions_df,
+    plot_reply_sankey,
+    plot_sentiment_histogram,
+    plot_sentiment_over_time,
+    plot_speaker_activity_heatmap,
+    plot_speaker_topic_heatmap,
+    plot_topic_shift_timeline,
+)
+
+
+def get_backend(name: str):
+    if name == "openai":
+        from conversation_storyline.llm_backends.openai_backend import OpenAIBackend
+        return OpenAIBackend()
+    elif name == "outlines":
+        from conversation_storyline.llm_backends.outlines_backend import OutlinesBackend
+        return OutlinesBackend()
+    else:
+        raise ValueError("backend inválido")
+
+
+async def _run(file_path: Optional[str], transcript_text: str, backend: str):
+    transcript_text = (transcript_text or "").strip()
+    if transcript_text:
+        msgs = load_messages_from_text(transcript_text)
+    else:
+        if not file_path:
+            raise ValueError("Debes pegar un transcript o subir un archivo.")
+        msgs = load_messages(file_path)
+
+    b = get_backend(backend)
+
+    outdir = Path(tempfile.mkdtemp(prefix="storyline_"))
+    await run_pipeline(msgs, b, str(outdir))
+
+    png = outdir / "storyline.png"
+    html = outdir / "storyline.html"
+    graph = outdir / "graph.json"
+    interactions = outdir / "interactions.jsonl"
+    metrics = outdir / "metrics.parquet"
+
+    html_inline = html.read_text(encoding="utf-8", errors="ignore") if html.exists() else None
+
+    figs = [None] * 5
+    try:
+        df = load_interactions_df(outdir)
+        g = load_graph_json(outdir)
+        figs = [
+            plot_sentiment_over_time(df),
+            plot_sentiment_histogram(df),
+            plot_speaker_topic_heatmap(df),
+            plot_speaker_activity_heatmap(df),
+            plot_reply_sankey(g),
+        ]
+        topic_shift_fig = plot_topic_shift_timeline(df)
+    except Exception:
+        topic_shift_fig = None
+
+    return (
+        str(png) if png.exists() else None,
+        html_inline,
+        str(html) if html.exists() else None,
+        str(graph) if graph.exists() else None,
+        str(interactions) if interactions.exists() else None,
+        str(metrics) if metrics.exists() else None,
+        figs[0],
+        figs[1],
+        figs[2],
+        figs[3],
+        figs[4],
+        topic_shift_fig,
+    )
+
+
+def run_ui(file_obj, transcript_text: str, backend: str):
+    file_path = file_obj.name if file_obj is not None else None
+    return asyncio.run(_run(file_path, transcript_text, backend))
+
+
+with gr.Blocks(title="Conversation Storyline – v4") as demo:
+    gr.Markdown("# Conversation Storyline – v4\nPega un transcript o sube TXT/CSV.")
+
+    with gr.Row():
+        f = gr.File(label="Upload (.txt o .csv)")
+        backend = gr.Dropdown(choices=["openai", "outlines"], value="openai", label="Backend LLM")
+
+    transcript_text = gr.Textbox(label="O pega aquí el transcript", lines=10)
+
+    btn = gr.Button("Run", variant="primary")
+
+    with gr.Tabs():
+        with gr.Tab("Storyline"):
+            with gr.Row():
+                out_png = gr.Image(label="Storyline (PNG)", type="filepath")
+                out_story_html = gr.HTML(label="Storyline (HTML embebido)")
+            out_html_file = gr.File(label="Storyline HTML (descarga)")
+
+        with gr.Tab("Analítica"):
+            out_sentiment = gr.Plot(label="Sentiment timeline")
+            out_hist = gr.Plot(label="Sentiment histogram")
+            out_topic_heat = gr.Plot(label="Speaker × topic heatmap")
+            out_activity_heat = gr.Plot(label="Speaker activity heatmap")
+            out_topic_shifts = gr.Plot(label="Topic shifts timeline")
+
+        with gr.Tab("Grafo"):
+            out_sankey = gr.Plot(label="Sankey replies")
+            out_graph = gr.File(label="Graph JSON")
+
+        with gr.Tab("Artifacts"):
+            out_interactions = gr.File(label="interactions.jsonl")
+            out_metrics = gr.File(label="metrics.parquet")
+
+    btn.click(
+        fn=run_ui,
+        inputs=[f, transcript_text, backend],
+        outputs=[
+            out_png,
+            out_story_html,
+            out_html_file,
+            out_graph,
+            out_interactions,
+            out_metrics,
+            out_sentiment,
+            out_hist,
+            out_topic_heat,
+            out_activity_heat,
+            out_sankey,
+            out_topic_shifts,
+        ],
+    )
+
+if __name__ == "__main__":
+    demo.launch()