fix(setup,chat): pass dummy api_key for local providers

litellm requires the api_key field even for local OpenAI-compatible
servers (LM Studio, llama.cpp). Use "local" as a sentinel value for
providers that don't require a real key.

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

src/pyra/setup/wizard.py

@@ -121,16 +121,16 @@ def _test_connection(provider: Provider, model: str) -> None:
     try:
         import litellm
 
-        api_key = get_key(provider.id) if provider.requires_key else "no-key"
+        # Local providers don't need a real key but litellm still requires the field
+        api_key = get_key(provider.id) if provider.requires_key else "local"
         kwargs: dict = {
             "model": f"{provider.litellm_prefix}{model}",
             "messages": [{"role": "user", "content": "Reply with exactly: OK"}],
             "max_tokens": 10,
+            "api_key": api_key,
         }
         if provider.base_url:
             kwargs["api_base"] = provider.base_url
-        if api_key and api_key != "no-key":
-            kwargs["api_key"] = api_key
 
         litellm.completion(**kwargs)
         console.print("[green]✓ Connection OK[/green]")