fix(chat): show "tools disabled" info message only once per session

When a local model rejects function calling (BadRequestError), the flag
is set in a session-scoped dict so subsequent messages skip the tool-use
path entirely — no repeated info message on every turn.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

src/pyra/chat/session.py

@@ -176,6 +176,8 @@ def start_chat() -> None:
         "[dim]Type /help for commands, /quit to exit.[/dim]"
     )
 
+    _flags: dict = {"use_tools": True}
+
     while True:
         try:
             user_input = session.prompt("› ").strip()
@@ -226,7 +228,7 @@ def start_chat() -> None:
         history.add_user(user_input)
 
         try:
-            response_text = _call_ai(cfg, history, registry, executor)
+            response_text = _call_ai(cfg, history, registry, executor, _flags)
         except Exception as exc:
             render_error(f"AI error: {exc}")
             history._messages.pop()
@@ -244,6 +246,7 @@ def _call_ai(
     history: ConversationHistory,
     registry: PluginRegistry,
     executor: ToolExecutor,
+    flags: dict | None = None,
 ) -> str:
     from pyra.vault.reader import get_key
 
@@ -272,8 +275,9 @@ def _call_ai(
         for t in tools
     ]
 
-    # No plugins active — use streaming (original behavior)
+    # No tools active, or provider known not to support function calling
-    if not tools_spec:
+    use_tools = flags is None or flags.get("use_tools", True)
+    if not tools_spec or not use_tools:
         stream = litellm.completion(
             **base_kwargs,
             messages=history.build_for_api(),
@@ -304,6 +308,8 @@ def _call_ai(
         return render_text_response("Error: tool-use loop exceeded maximum iterations.")
 
     except litellm.BadRequestError:
+        if flags is not None:
+            flags["use_tools"] = False
         render_info("This model does not support function calling — tools disabled.")
         stream = litellm.completion(
             **base_kwargs,