fix Inference API slownless

src/app/api/providers/huggingface/predictWithHuggingFace.ts

@@ -9,6 +9,7 @@ export async function predict({
   systemPrompt,
   userPrompt,
   nbMaxNewTokens,
+  turbo,
   prefix,
 }: LLMPredictionFunctionParams): Promise<string> {
 
@@ -18,7 +19,10 @@ export async function predict({
   try {
     for await (const output of hf.textGenerationStream({
       // model: "mistralai/Mixtral-8x7B-v0.1",
-      model: "mistralai/Mixtral-8x7B-Instruct-v0.1",
+      model: 
+        turbo
+          ? "HuggingFaceH4/zephyr-7b-beta"
+          : "mistralai/Mixtral-8x7B-Instruct-v0.1",
       inputs: createZephyrPrompt([
         { role: "system", content: systemPrompt },
         { role: "user", content: userPrompt }

src/app/api/providers/types.ts

@@ -16,6 +16,7 @@ export type LLMPredictionFunctionParams = {
   systemPrompt: string
   userPrompt: string
   nbMaxNewTokens: number
+  turbo?: boolean
   prefix?: string
   // llmVendorConfig: LLMVendorConfig
 }

src/app/api/v1/create/index.ts

@@ -15,10 +15,12 @@ export async function create(request: {
   prompt?: string
   width?: number
   height?: number
+  turbo?: boolean
 }= {
   prompt: "",
   width: 1024,
   height: 576,
+  turbo: false,
 }): Promise<ClapProject> {
   const prompt = `${request?.prompt || ""}`.trim()
 
@@ -28,7 +30,8 @@ export async function create(request: {
 
   const width = getValidNumber(request?.width, 256, 8192, 1024)
   const height = getValidNumber(request?.height, 256, 8192, 576)
-
+  const turbo = request?.turbo ? true : false
+  
   const userPrompt = `Movie story to generate: ${prompt}
 
 Output: `
@@ -46,6 +49,7 @@ Output: `
     userPrompt,
     nbMaxNewTokens,
     prefix,
+    turbo,
   })
 
   console.log("api/v1/create(): rawString: ", rawString)
@@ -64,6 +68,7 @@ Output: `
       userPrompt: userPrompt + ".", // we trick the Hugging Face cache
       nbMaxNewTokens,
       prefix,
+      turbo,
     })
   
     console.log("api/v1/create(): rawString: ", rawString)

src/app/api/v1/create/route.ts

@@ -14,6 +14,7 @@ export async function POST(req: NextRequest) {
     prompt: string
     width: number
     height: number
+    turbo: boolean
     // can add more stuff for the V2 of Stories Factory
   }
   
@@ -22,7 +23,8 @@ export async function POST(req: NextRequest) {
   const clap = await create({
     prompt: `${request?.prompt || ""}`.trim(),
     width: getValidNumber(request?.width, 256, 8192, 1024),
-    height: getValidNumber(request?.height, 256, 8192, 576)
+    height: getValidNumber(request?.height, 256, 8192, 576),
+    turbo: request?.turbo ? true : false,
   })
 
   // TODO replace by Clap file streaming