first pass of .var preview

app.py

@@ -50,6 +50,9 @@ PARQUET_EMB_COL = "tx1-70m"
 OBS_NONE_OPTION = "(none)"
 MAX_CATEGORIES = 50
 
+# constants for .var preview
+VAR_PREVIEW_MAX = 5
+
 # helper to read AnnData header
 def read_anndata_header(fileobj):
     adata = sc.read_h5ad(fileobj.name, backed="r")
@@ -68,6 +71,41 @@ def _pick_layer(adata, layer_name):
         X = X.toarray()
     return X
 
+# helper to summarize .var columns
+def _summarize_var_columns(adata_var_df, preview_max=VAR_PREVIEW_MAX):
+    summaries = {}
+    choices = []
+    for col in adata_var_df.columns:
+        s = adata_var_df[col]
+        dtype = str(s.dtype)
+        ex = pd.Series(s.astype(object)).dropna().astype(str).head(preview_max).tolist()
+        preview = ", ".join(ex) if ex else "—"
+        if len(preview) > 50:
+            preview = preview[:47] + "..."
+        lbl = f"{col} · {dtype} · e.g. {preview}"
+        summaries[col] = {
+            "dtype": dtype,
+            "nunique": int(pd.unique(s.astype(object)).size),
+            "examples": ex,
+            "example_text": ", ".join(ex) if ex else "—",
+        }
+        choices.append((lbl, col))
+    return choices, summaries
+
+# helper to format .var column preview in markdown
+def _format_var_preview_md(col, summaries):
+    if not col or summaries is None or col not in summaries:
+        return "Select a .var column to see a preview."
+    info = summaries[col]
+    ex = info["examples"]
+    rows = "\n".join([f"| {i+1} | {v} |" for i, v in enumerate(ex)]) if ex else "| — | — |"
+    return (
+        f"**Column:** `{col}`  \n"
+        f"**dtype:** `{info['dtype']}` · **unique:** {info['nunique']}\n\n"
+        f"**Examples (up to {len(ex)}):**\n\n"
+        f"| # | value |\n|---:|:------|\n{rows}"
+    )
+
 # helper to compute UMAP from given embeddings
 def _compute_umap_from_emb(emb):
     ad_umap = ad.AnnData(X=emb)
@@ -110,17 +148,30 @@ def ensure_dropdowns(fileobj):
         return (
             gr.Dropdown(choices=["<use .X>"], value="<use .X>"),
             gr.Dropdown(choices=[], value=None),
+            gr.State({}),
+            gr.Markdown.update(value="")
         )
     try:
-        layers, var_cols, _ = read_anndata_header(fileobj)
+        adata = sc.read_h5ad(fileobj.name, backed="r")
+        adata.var = adata.var.reset_index(drop=False, names="index")
+        layers = list(adata.layers.keys())
+        var_choices, var_summaries = _summarize_var_columns(adata.var)
+        del adata
+        gc.collect()
+        default_var = var_choices[0][1] if var_choices else None
+        preview_md = _format_var_preview_md(default_var, var_summaries) if default_var else ""
         return (
             gr.Dropdown(choices=["<use .X>"] + layers, value="<use .X>"),
-            gr.Dropdown(choices=var_cols, value=(var_cols[0] if var_cols else None)),
+            gr.Dropdown(choices=var_choices, value=default_var),
+            gr.State(var_summaries)
+            gr.Markdown.update(value=preview_md)
         )
     except Exception:
         return (
             gr.Dropdown(choices=["<use .X>"], value="<use .X>"),
             gr.Dropdown(choices=[], value=None),
+            gr.State({}),
+            gr.Markdown.update(value="")
         )
     
 # draw an uncolored UMAP
@@ -562,6 +613,7 @@ with gr.Blocks(title=APP_TITLE, css=css) as demo:
         layer_dd = gr.Dropdown(choices=["<use .X>"], value="<use .X>", label="Layer to use (default: .X)", scale=1)
         with gr.Column(scale=1):
             var_dd = gr.Dropdown(choices=[], value=None, label="Name of .var column with Ensembl gene IDs (or gene symbols)")
+            var_preview_md = gr.Markdown("")
             use_symbols_chk = gr.Checkbox(label="Selected .var column contains gene symbols (attempt conversion to Ensembl IDs)", value=False)
 
     # run button
@@ -580,10 +632,40 @@ with gr.Blocks(title=APP_TITLE, css=css) as demo:
         emb_parquet = gr.DownloadButton(label="Download embeddings (.parquet)")
         adata_with_emb = gr.DownloadButton(label="Download AnnData with embeddings in .obsm (.h5ad)")
 
-    # wire embedding pipeline
-    f_in.change(ensure_dropdowns, inputs=[f_in], outputs=[layer_dd, var_dd], queue=False)
-    evt = run_btn.click(run_pipeline, inputs=[f_in, layer_dd, var_dd, use_symbols_chk], outputs=[umap_img, emb_parquet, adata_with_emb, layer_dd, var_dd, obs_dd, coords_state, h5ad_state], queue=True)
-    evt.then(ensure_dropdowns, inputs=[f_in], outputs=[layer_dd, var_dd], queue=False)
+    # when file changes, refresh dropdowns
+    var_summaries_state = gr.State({})
+    f_in.change(
+        ensure_dropdowns,
+        inputs=[f_in],
+        outputs=[layer_dd, var_dd, var_summaries_state, var_preview_md],
+        queue=False
+    )
+
+    # when var selection changes, update the preview markdown
+    def _update_var_preview(selected, summaries):
+        return _format_var_preview_md(selected, summaries or {})
+    var_dd.change(
+        _update_var_preview,
+        inputs=[var_dd, var_summaries_state],
+        outputs=[var_preview_md],
+        queue=False
+    )
+
+    # run pipeline on button click
+    evt = run_btn.click(
+        run_pipeline,
+        inputs=[f_in, layer_dd, var_dd, use_symbols_chk],
+        outputs=[umap_img, emb_parquet, adata_with_emb, layer_dd, var_dd, obs_dd, coords_state, h5ad_state],
+        queue=True
+    )
+
+    # refresh dropdowns after run
+    evt.then(
+        ensure_dropdowns,
+        inputs=[f_in],
+        outputs=[layer_dd, var_dd],
+        queue=False
+    )
 
     # wire UMAP recoloring
     obs_dd.change(recolor_umap, inputs=[obs_dd, coords_state, h5ad_state], outputs=[umap_img], queue=False)