reducing the tokens for faster generation on hugging space

app.py

@@ -81,7 +81,7 @@ with col2:
             # 2) generation (reduce tokens for now)
             out = deepseek_gen(
                 f"Use these notes to answer:\n\n{context}\n\nQ: {title}\nA:",
-                max_new_tokens=20,
+                max_new_tokens=10,
                 do_sample=False
             )
 

assist/main.py

@@ -13,8 +13,8 @@ def hello(name: str = "world"):
 def heatmap():
     """Show token-similarity heatmap from tensor.pt."""
     # import here so chat() doesn’t drag in sklearn
-    from .heatmap import show_heatmap
-    show_heatmap()
+    from .heatmap import get_heatmap_figure
+    get_heatmap_figure()
 
 @app.command()
 def chat(question: str = typer.Argument(..., help="Question to ask RepoSage")):