fix(chat): fall back to streaming when provider rejects function calling

Local models (e.g. Gemma on LM Studio) return HTTP 400 when sent a
tools-spec request. Catch litellm.BadRequestError in the tool-use loop,
inform the user once that tools are disabled, and retry as a plain
streaming call so the conversation continues normally.

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

src/pyra/chat/session.py

@@ -282,6 +282,7 @@ def _call_ai(
         return render_streaming_response(stream)
 
     # Plugin tool-use loop (non-streaming for tool calls, renders final response)
+    try:
         for _iteration in range(10):
             response = litellm.completion(
                 **base_kwargs,
@@ -302,6 +303,15 @@ def _call_ai(
 
         return render_text_response("Error: tool-use loop exceeded maximum iterations.")
 
+    except litellm.BadRequestError:
+        render_info("This model does not support function calling — tools disabled.")
+        stream = litellm.completion(
+            **base_kwargs,
+            messages=history.build_for_api(),
+            stream=True,
+        )
+        return render_streaming_response(stream)
+
 
 def _show_help(plugin_slash: dict) -> None:
     lines = ["[bold]Slash commands:[/bold]"]