responsiveness: show model sizes, warm model on select, skip tools for non-tool models

metatron.py

@@ -80,13 +80,18 @@ def post_stream(path, payload):
 
 def list_models():
     d = get("/api/tags")
-    return [m["name"] for m in d.get("models", [])]
+    out = []
+    for m in d.get("models", []):
+        ps = (m.get("details") or {}).get("parameter_size", "") or ""
+        out.append((m["name"], ps))
+    return out
 
 
 def pick_model(models):
-    print(f"\n{C['b']}{C['m']}  METATRON{C['r']} — select a model\n")
-    for i, name in enumerate(models, 1):
-        print(f"  {C['c']}[{i}]{C['r']} {name}")
+    print(f"\n{C['b']}{C['m']}  METATRON{C['r']} — select a model "
+          f"{C['y']}(smaller = faster){C['r']}\n")
+    for i, (name, ps) in enumerate(models, 1):
+        print(f"  {C['c']}[{i}]{C['r']} {name}  {C['g']}{ps}{C['r']}")
     print(f"  {C['c']}[0]{C['r']} {C['y']}quit{C['r']}\n")
     while True:
         try:
@@ -95,12 +100,41 @@ def pick_model(models):
                 sys.exit(0)
             idx = int(sel) - 1
             if 0 <= idx < len(models):
-                return models[idx]
+                return models[idx][0]
         except ValueError:
             pass
         print(f"{C['y']}  pick a number 1-{len(models)}{C['r']}")
 
 
+def warm_model(model):
+    """Load the model into VRAM so the first real message is fast (no cold start)."""
+    sys.stdout.write(f"{C['y']}  warming {model} ...{C['r']}")
+    sys.stdout.flush()
+    payload = {"model": model, "messages": [{"role": "user", "content": "hi"}],
+               "stream": False, "think": False,
+               "options": {"num_predict": 1}}
+    try:
+        data = json.dumps(payload).encode()
+        req = urllib.request.Request(OLLAMA + "/api/chat", data=data,
+                                     headers={"Content-Type": "application/json"})
+        urllib.request.urlopen(req, timeout=180)
+        sys.stdout.write(f"\r{C['g']}  ready.{C['r']}      \n")
+    except Exception as e:
+        sys.stdout.write(f"\r{C['y']}  warm failed ({e}){C['r']}      \n")
+
+
+def supports_tools(model):
+    """True if this model has native function-calling (else omit `tools`)."""
+    try:
+        data = json.dumps({"model": model}).encode()
+        req = urllib.request.Request(OLLAMA + "/api/show", data=data,
+                                     headers={"Content-Type": "application/json"})
+        d = json.load(urllib.request.urlopen(req, timeout=15))
+        return "tools" in (d.get("capabilities") or [])
+    except Exception:
+        return False
+
+
 def run_command(cmd):
     try:
         r = subprocess.run(cmd, shell=True, capture_output=True, text=True,
@@ -129,6 +163,7 @@ def web_search(query):
 
 def chat(model):
     messages = [{"role": "system", "content": SYSTEM_PROMPT}]
+    use_tools = supports_tools(model)
     print(f"\n{C['b']}  METATRON {C['c']}:: {model}{C['r']}  "
           f"{C['y']}(!cmd = run, ?query = search, exit = quit){C['r']}\n")
     while True:
@@ -154,7 +189,9 @@ def chat(model):
         # agentic loop: let the model call run_command until it's satisfied
         for _ in range(6):
             payload = {"model": model, "messages": messages, "stream": True,
-                       "think": False, "tools": TOOLS}
+                       "think": False}
+            if use_tools:
+                payload["tools"] = TOOLS
             buf = ""
             tool_calls = []
             try:
@@ -210,6 +247,7 @@ def main():
         print(f"{C['y']}  no models on nightmare{C['r']}")
         sys.exit(1)
     model = pick_model(models)
+    warm_model(model)
     chat(model)