error handling

src/components/game-component.tsx

@@ -65,7 +65,7 @@ const Switch = ({
 };
 
 type Message = {
-  role: "user" | "assistant" | "game" | "result";
+  role: "user" | "assistant" | "game" | "result" | "error";
   content: string;
   metadata?: {
     page?: string;
@@ -145,7 +145,7 @@ export default function GameComponent({
   const {
     status: modelStatus,
     partialText,
-    inference,
+    inference
   } = useInference({
     apiKey:
       window.localStorage.getItem("huggingface_access_token") || undefined,
@@ -228,12 +228,22 @@ export default function GameComponent({
       content: prompt,
     });
 
-    const modelResponse = await inference({
+    const {status, result: modelResponse} = await inference({
       model: model,
       prompt,
       maxTokens: maxTokens,
     });
 
+    if (status === "error") {
+      pushConvo({
+        role: "error",
+        content: "Error during inference: " + modelResponse,
+      });
+
+      setAutoRunning(false);
+      return;
+    }
+
     pushConvo({
       role: "assistant",
       content: modelResponse,
@@ -703,6 +713,12 @@ export default function GameComponent({
                     </div>
                   </div>
                 );
+              } else if (message.role === "error") {
+                return (
+                  <div className="p-2 rounded-lg bg-red-50 border border-red-100 text-xs">
+                    <p>{message.content}</p>
+                  </div>
+                );
               }
               
               return null;

src/lib/inference.tsx

@@ -98,19 +98,26 @@ export function useInference({ apiKey }) {
     prompt,
     model,
     maxTokens,
+    provider = "fireworks-ai",
   }: {
     prompt: string;
     model: string;
     maxTokens: number;
+    provider: string;
   }) => {
     setIsLoading(true);
     setPartialText("");
 
     const client = new InferenceClient(apiKey);
 
+    // fireworks doesnt support max tokens
+    if (provider === "fireworks-ai") {
+      maxTokens = undefined;
+    }
+
     try {
       const stream = client.chatCompletionStream({
-        provider: "hyperbolic",
+        // provider,
         model,
         maxTokens,
         messages: [
@@ -132,12 +139,12 @@ export function useInference({ apiKey }) {
 
       setInferenceResult(result);
 
-      return result;
+      return {status: "success", result};
     } catch (error) {
       console.error("Error in inference", error);
       setError(error.message);
       setIsLoading(false);
-      return null;
+      return {status: "error", result: error.message};
     }
   };