Deploy Space: update app.py

app.py

@@ -35,7 +35,7 @@ from src.utils import (
 
 RESULTS_PATH = os.getenv("TSFM_RESULTS_PATH", "space/results" if Path("space/results").exists() else "results")
 REFRESH_SECONDS = int(os.getenv("TSFM_REFRESH_SECONDS", "300"))
-DOMAIN_OUTPUT_OFFSET = 13
+DOMAIN_OUTPUT_OFFSET = 14
 
 
 def _dataset_tables(grouped: dict, datasets: list[str]) -> tuple[list, list]:
@@ -80,19 +80,20 @@ def refresh_leaderboard() -> tuple:
     daily_df = load_baseline_rank_history(RESULTS_PATH)
 
     outputs: list = [
-        make_domain_pie_chart(RESULTS_PATH),                # [0] domain pie chart
-        prepare_values_table(grouped["overall_values"]),    # [1]
-        prepare_ranks_table(grouped["overall_ranks"]),      # [2]
-        daily_df,                                           # [3] daily table
-        load_baseline_rank_history_weekly(RESULTS_PATH),    # [4] weekly table
-        make_rank_trend_plot(daily_df),                     # [5] rank trend plot
-        load_gift_aggregate_metadata_md(RESULTS_PATH),      # [6]
-        load_gift_aggregate_table(RESULTS_PATH, "prediction_length"),  # [7]
-        load_gift_aggregate_table(RESULTS_PATH, "domain"),             # [8]
-        load_gift_aggregate_table(RESULTS_PATH, "frequency"),          # [9]
-        load_live_aggregate_metadata_md(RESULTS_PATH),                 # [10]
-        load_live_aggregate_table(RESULTS_PATH, "overall"),           # [11]
-        load_live_aggregate_table(RESULTS_PATH, "rank"),              # [12]
+        build_leaderboard_summary_html(RESULTS_PATH),       # [0]
+        make_domain_pie_chart(RESULTS_PATH),                # [1] domain pie chart
+        prepare_values_table(grouped["overall_values"]),    # [2]
+        prepare_ranks_table(grouped["overall_ranks"]),      # [3]
+        daily_df,                                           # [4] daily table
+        load_baseline_rank_history_weekly(RESULTS_PATH),    # [5] weekly table
+        make_rank_trend_plot(daily_df),                     # [6] rank trend plot
+        load_gift_aggregate_metadata_md(RESULTS_PATH),      # [7]
+        load_gift_aggregate_table(RESULTS_PATH, "prediction_length"),  # [8]
+        load_gift_aggregate_table(RESULTS_PATH, "domain"),             # [9]
+        load_gift_aggregate_table(RESULTS_PATH, "frequency"),          # [10]
+        load_live_aggregate_metadata_md(RESULTS_PATH),                 # [11]
+        load_live_aggregate_table(RESULTS_PATH, "overall"),           # [12]
+        load_live_aggregate_table(RESULTS_PATH, "rank"),              # [13]
     ]
 
     # [10 .. 10+D-1] per-domain status lines
@@ -119,7 +120,7 @@ demo = gr.Blocks(css=custom_css)
 with demo:
     gr.HTML(TITLE)
     gr.Markdown(INTRODUCTION_TEXT, elem_classes="markdown-text")
-    gr.HTML(build_leaderboard_summary_html(RESULTS_PATH))
+    summary_html = gr.HTML(INITIAL_OUTPUTS[0])
 
     dataset_value_dfs: list[gr.Dataframe] = []
     dataset_rank_dfs: list[gr.Dataframe] = []
@@ -130,7 +131,7 @@ with demo:
         # ── Overall Tab ──────────────────────────────────────────────────────
         with gr.TabItem("Overall"):
             live_aggregate_metadata_md = gr.Markdown(
-                INITIAL_OUTPUTS[10],
+                INITIAL_OUTPUTS[11],
                 elem_classes="markdown-text",
             )
             gr.Markdown(
@@ -138,13 +139,13 @@ with demo:
                 elem_classes="markdown-text",
             )
             live_overall_df = gr.Dataframe(
-                value=INITIAL_OUTPUTS[11],
+                value=INITIAL_OUTPUTS[12],
                 interactive=False,
                 wrap=True,
                 label="Up-to-now overall results",
             )
             live_rank_df = gr.Dataframe(
-                value=INITIAL_OUTPUTS[12],
+                value=INITIAL_OUTPUTS[13],
                 interactive=False,
                 wrap=True,
                 label="Average Rank, Win Rate, and Elo",
@@ -158,7 +159,7 @@ with demo:
                 elem_classes="markdown-text",
             )
             domain_pie_plot = gr.Plot(
-                value=INITIAL_OUTPUTS[0],
+                value=INITIAL_OUTPUTS[1],
                 label="Domain distribution",
             )
             gr.Markdown(
@@ -166,13 +167,13 @@ with demo:
                 elem_classes="markdown-text",
             )
             overall_values_df = gr.Dataframe(
-                value=INITIAL_OUTPUTS[1],
+                value=INITIAL_OUTPUTS[2],
                 interactive=False,
                 wrap=True,
                 label="Metric values and RankScore",
             )
             overall_ranks_df = gr.Dataframe(
-                value=INITIAL_OUTPUTS[2],
+                value=INITIAL_OUTPUTS[3],
                 interactive=False,
                 wrap=True,
                 label="Ranks (average per-dataset rank, lower is better)",
@@ -183,7 +184,7 @@ with demo:
                 elem_classes="markdown-text",
             )
             baseline_history_df = gr.Dataframe(
-                value=INITIAL_OUTPUTS[3],
+                value=INITIAL_OUTPUTS[4],
                 interactive=False,
                 wrap=True,
                 label="Baseline rank history — by day",
@@ -193,7 +194,7 @@ with demo:
                 elem_classes="markdown-text",
             )
             baseline_weekly_df = gr.Dataframe(
-                value=INITIAL_OUTPUTS[4],
+                value=INITIAL_OUTPUTS[5],
                 interactive=False,
                 wrap=True,
                 label="Baseline rank history — by week",
@@ -203,14 +204,14 @@ with demo:
                 elem_classes="markdown-text",
             )
             rank_trend_plot = gr.Plot(
-                value=INITIAL_OUTPUTS[5],
+                value=INITIAL_OUTPUTS[6],
                 label="Rank trend over time",
             )
 
         # ── GIFT-style grouped result tables ────────────────────────────────
         with gr.TabItem("GIFT-style Aggregates"):
             aggregate_metadata_md = gr.Markdown(
-                INITIAL_OUTPUTS[6],
+                INITIAL_OUTPUTS[7],
                 elem_classes="markdown-text",
             )
             gr.Markdown(
@@ -220,21 +221,21 @@ with demo:
             with gr.Tabs():
                 with gr.TabItem("Prediction Length"):
                     prediction_length_aggregate_df = gr.Dataframe(
-                        value=INITIAL_OUTPUTS[7],
+                        value=INITIAL_OUTPUTS[8],
                         interactive=False,
                         wrap=True,
                         label="Results on TSFM_Bench aggregated by Prediction Length",
                     )
                 with gr.TabItem("Domain"):
                     domain_aggregate_df = gr.Dataframe(
-                        value=INITIAL_OUTPUTS[8],
+                        value=INITIAL_OUTPUTS[9],
                         interactive=False,
                         wrap=True,
                         label="Results on TSFM_Bench aggregated by Domain",
                     )
                 with gr.TabItem("Frequency"):
                     frequency_aggregate_df = gr.Dataframe(
-                        value=INITIAL_OUTPUTS[9],
+                        value=INITIAL_OUTPUTS[10],
                         interactive=False,
                         wrap=True,
                         label="Results on TSFM_Bench aggregated by Frequency",
@@ -343,6 +344,7 @@ with demo:
         )
 
     refresh_outputs = [
+        summary_html,
         domain_pie_plot,
         overall_values_df,
         overall_ranks_df,