fix

pages/api/chat/index.ts

@@ -2,6 +2,7 @@ import { OpenAIStream, StreamingTextResponse } from "ai";
 import { Configuration, OpenAIApi } from "openai-edge";
 import { HfInference } from '@huggingface/inference';
 import { HuggingFaceStream } from 'ai';
+import { experimental_buildLlama2Prompt } from 'ai/prompts';
 import { LLMStream } from "../../../utils/llm";
 
 export const runtime = 'edge';
@@ -38,7 +39,7 @@ export default async function(req: Request) {
         const response = Hf.textGenerationStream({
             // @ts-ignore
             model: aiProvider,
-            inputs: messages,
+            inputs: experimental_buildLlama2Prompt(messages),
             parameters: {
                 repetition_penalty: 1,
                 return_full_text: false