feat: use llama

utils/aiGatewayClient.ts

@@ -1,54 +1,90 @@
-import { generateText } from 'ai';
+import OpenAI from 'openai';
 
-interface AiGatewayResponse {
+let ovhAI: OpenAI | null = null;
+
+function getClient(): OpenAI {
+  if (!ovhAI) {
+    ovhAI = new OpenAI({
+      apiKey: process.env.OVHCLOUD_API_KEY,
+      baseURL: 'https://oai.endpoints.kepler.ai.cloud.ovh.net/v1'
+    });
+  }
+  return ovhAI;
+}
+
+interface AiResponse {
   text: string;
   tokensUsed: number;
 }
 
-export async function queryAiGateway(
-  text: string,
-  model: string
-): Promise<AiGatewayResponse> {
+const MAX_RETRIES = 3;
+
+export async function queryAi(text: string): Promise<AiResponse> {
   const requestId = Math.random().toString(36).substring(7);
   const startTime = Date.now();
 
-  console.info(`[AI-${requestId}] Starting Vercel Gateway AI request`, {
+  console.info(`[AI-${requestId}] Starting OVH AI request`, {
     promptLength: text.length,
-    model,
     timestamp: new Date().toISOString()
   });
 
-  try {
-    const response = await generateText({
-      model,
-      prompt: text
-    });
+  let lastError: Error | null = null;
 
-    const duration = Date.now() - startTime;
-    const tokensUsed = response.usage?.totalTokens || 0;
+  for (let attempt = 1; attempt <= MAX_RETRIES; attempt++) {
+    try {
+      console.info(
+        `[AI-${requestId}] OVH AI request attempt ${attempt}/${MAX_RETRIES}`
+      );
 
-    console.info(
-      `[AI-${requestId}] Vercel Gateway AI response received in ${duration}ms`,
-      {
-        responseLength: response.text.length,
-        tokensUsed,
-        usage: response.usage
+      const completion = await getClient().chat.completions.create({
+        model: 'Meta-Llama-3_3-70B-Instruct',
+        temperature: 0.7,
+        max_tokens: 4096,
+        messages: [
+          {
+            role: 'user',
+            content: text
+          }
+        ]
+      });
+
+      const responseText = completion.choices[0]?.message?.content || '';
+      const tokensUsed =
+        (completion.usage?.prompt_tokens || 0) +
+        (completion.usage?.completion_tokens || 0);
+
+      const duration = Date.now() - startTime;
+      console.info(
+        `[AI-${requestId}] OVH AI response received in ${duration}ms`,
+        {
+          responseLength: responseText.length,
+          tokensUsed,
+          usage: completion.usage
+        }
+      );
+
+      return {
+        text: responseText,
+        tokensUsed
+      };
+    } catch (error) {
+      const duration = Date.now() - startTime;
+      console.error(
+        `[AI-${requestId}] OVH AI attempt ${attempt} failed after ${duration}ms:`,
+        {
+          error: error instanceof Error ? error.message : String(error),
+          promptLength: text.length
+        }
+      );
+      lastError = error as Error;
+
+      if (attempt < MAX_RETRIES) {
+        const delay = 1000 * attempt;
+        console.info(`[AI-${requestId}] Retrying in ${delay}ms...`);
+        await new Promise(resolve => setTimeout(resolve, delay));
       }
-    );
-
-    return {
-      text: response.text,
-      tokensUsed
-    };
-  } catch (error) {
-    const duration = Date.now() - startTime;
-    console.error(
-      `[AI-${requestId}] Vercel Gateway AI error after ${duration}ms:`,
-      {
-        error: error instanceof Error ? error.message : String(error),
-        promptLength: text.length
-      }
-    );
-    throw new Error(`Vercel Gateway AI error: ${JSON.stringify(error)}.`);
+    }
   }
+
+  throw lastError || new Error('OVH AI error: all retry attempts failed.');
 }