Upload project files

src/app/api/consciousness/asr/route.ts

@@ -0,0 +1,81 @@
+import { NextRequest, NextResponse } from "next/server";
+
+export async function POST(req: NextRequest) {
+  try {
+    const body = await req.json();
+    const { audio } = body;
+
+    if (!audio) {
+      return NextResponse.json(
+        { error: "Audio data is required" },
+        { status: 400 }
+      );
+    }
+
+    // Try to use z-ai-web-dev-sdk for ASR
+    try {
+      const ZAI = (await import("z-ai-web-dev-sdk")).default;
+      const zai = await ZAI.create();
+
+      // Convert base64 to a proper File/Blob for the Whisper API
+      // The SDK expects an actual file-like object, not a raw base64 string
+      let audioBuffer: Buffer;
+
+      if (typeof audio === "string") {
+        // Remove data URL prefix if present (e.g., "data:audio/webm;base64,")
+        const base64Data = audio.includes(",") ? audio.split(",")[1] : audio;
+        audioBuffer = Buffer.from(base64Data, "base64");
+      } else {
+        return NextResponse.json(
+          { error: "Audio data must be a base64 string" },
+          { status: 400 }
+        );
+      }
+
+      // Create a File-like object with proper MIME type
+      const audioBlob = new Blob([audioBuffer], { type: "audio/webm" });
+      const audioFile = new File([audioBlob], "recording.webm", {
+        type: "audio/webm",
+        lastModified: Date.now(),
+      });
+
+      const result = await zai.audio.transcriptions.create({
+        model: "whisper-1",
+        file: audioFile,
+      });
+
+      const transcribedText = (result as Record<string, unknown>).text || (result as { text?: string }).text || "";
+
+      if (transcribedText && typeof transcribedText === "string" && transcribedText.trim()) {
+        return NextResponse.json({
+          text: transcribedText.trim(),
+          success: true,
+        });
+      }
+
+      // If transcription returned empty, still return gracefully
+      return NextResponse.json({
+        text: "",
+        success: true,
+        note: "Transcription returned empty — please try again",
+      });
+    } catch (apiError) {
+      console.error("ASR API error:", apiError);
+
+      // Enhanced fallback: use Web Speech API hint in the response
+      // The client-side will handle browser-based recognition as backup
+      return NextResponse.json({
+        text: "",
+        success: false,
+        fallback: true,
+        note: "Server ASR unavailable — use browser SpeechRecognition API",
+      });
+    }
+  } catch (error) {
+    console.error("ASR route error:", error);
+    return NextResponse.json(
+      { error: "Speech recognition failed" },
+      { status: 500 }
+    );
+  }
+}