Fix: Explicitly use Featherless AI provider for abliterated models via HF Router and add fallback

server/_core/llm.ts

@@ -287,6 +287,8 @@ export async function invokeLLM(params: InvokeParams): Promise<InvokeResult> {
     response_format,
   } = params;
 
+  // Use a more widely supported model if the abliterated one fails, 
+  // or ensure the provider is correctly specified.
   const model = params.model || "huihui-ai/Qwen2.5-72B-Instruct-abliterated";
 
   const payload: Record<string, unknown> = {
@@ -322,21 +324,50 @@ export async function invokeLLM(params: InvokeParams): Promise<InvokeResult> {
   const apiUrl = resolveApiUrl(model);
   console.log(`[LLM] Invoking ${model} at ${apiUrl}`);
 
+  // For abliterated models, we MUST specify the provider in the payload or headers correctly
+  // Some providers on HF Router require specific headers
+  const headers: Record<string, string> = {
+    "content-type": "application/json",
+    authorization: `Bearer ${apiKey}`,
+    "x-use-cache": "false",
+    "x-wait-for-model": "true",
+  };
+
+  // If using HF Router, we can try to force a provider that supports this model
+  if (apiUrl.includes("router.huggingface.co")) {
+    headers["x-inference-provider"] = "featherless";
+  }
+
   const response = await fetch(apiUrl, {
     method: "POST",
-    headers: {
-      "content-type": "application/json",
-      authorization: `Bearer ${apiKey}`,
-      "x-use-cache": "false",
-      "x-wait-for-model": "true",
-      "x-inference-provider": "featherless",
-    },
+    headers: headers,
     body: JSON.stringify(payload),
   });
 
   if (!response.ok) {
     const errorText = await response.text();
     console.error(`[LLM Error] Status: ${response.status}, Body: ${errorText}`);
+    
+    // Fallback mechanism: If the abliterated model fails, try the standard Qwen model
+    if (model.includes("abliterated") && (response.status === 400 || response.status === 404)) {
+      console.log("[LLM] Falling back to standard Qwen2.5-72B-Instruct");
+      const fallbackModel = "Qwen/Qwen2.5-72B-Instruct";
+      payload.model = fallbackModel;
+      
+      const fallbackResponse = await fetch(apiUrl, {
+        method: "POST",
+        headers: {
+          "content-type": "application/json",
+          authorization: `Bearer ${apiKey}`,
+        },
+        body: JSON.stringify(payload),
+      });
+      
+      if (fallbackResponse.ok) {
+        return (await fallbackResponse.json()) as InvokeResult;
+      }
+    }
+
     throw new Error(
       `LLM invoke failed: ${response.status} ${response.statusText} – ${errorText}`
     );