fix default model to 'lightning' for InferencePort API compatibility

.env.example

@@ -8,8 +8,9 @@ INFERENCE_API_URL=https://sharktide-lightning.hf.space/v1/chat/completions
 # Secret - set this in Space secrets, not in the repo.
 INFERENCE_API_KEY=
 
-# Model id served by that endpoint.
-INFERENCE_MODEL=llama-3.3-70b-versatile
+# Model id served by that endpoint. 'lightning' is the auto-router (recommended).
+# Override models can be configured in the InferencePort console.
+INFERENCE_MODEL=lightning
 
 # Optional: name Forge uses for itself in the UI and its own system prompt.
 AGENT_NAME=Forge

src/inferenceClient.js

@@ -2,17 +2,15 @@ const API_URL = process.env.INFERENCE_API_URL;
 const API_KEY = process.env.INFERENCE_API_KEY;
 const FETCH_TIMEOUT = 30000;
 
-let currentModel = process.env.INFERENCE_MODEL || 'llama-3.3-70b-versatile';
+let currentModel = process.env.INFERENCE_MODEL || 'lightning';
 
 const AVAILABLE_MODELS = [
+  'lightning',
+  'gpt-4o-mini',
+  'claude-sonnet-4-20250514',
   'llama-3.3-70b-versatile',
   'llama-3.1-8b-instant',
-  'llama3-70b-8192',
-  'llama3-8b-8192',
-  'mixtral-8x7b-32768',
   'gemma2-9b-it',
-  'gpt-4o-mini',
-  'claude-sonnet-4-20250514',
 ];
 
 function getModel() { return currentModel; }