Initial transfer from old repo

app.py

@@ -0,0 +1,175 @@
+import torch
+import gradio as gr
+import librosa
+import os
+import base64
+import tempfile
+
+from transformers import pipeline
+from huggingface_hub import login
+from google.cloud import translate_v3
+# from livekit import api
+
+
+# ===========================
+#  INITIAL SETUP
+# ===========================
+
+# Log in to Hugging Face
+hf_token = os.getenv("HUGGINGFACE_TOKEN")
+if hf_token:
+    login(token=hf_token)
+else:
+    print("⚠️ No Hugging Face token found. You cannot access private models.")
+
+# Load and decode Google credentials
+creds_b64 = os.getenv("GOOGLE_APPLICATION_CREDENTIALS_JSON")
+if creds_b64:
+    creds_json = base64.b64decode(creds_b64).decode("utf-8")
+    temp_file = tempfile.NamedTemporaryFile(delete=False, suffix=".json")
+    temp_file.write(creds_json.encode("utf-8"))
+    temp_file.flush()
+    os.environ["GOOGLE_APPLICATION_CREDENTIALS"] = temp_file.name
+
+# LiveKit
+# def create_livekit_token(user_name="guest"):
+#     LIVEKIT_URL = os.getenv("LIVEKIT_URL")
+#     LIVEKIT_API_KEY = os.getenv("LIVEKIT_API_KEY")
+#     LIVEKIT_API_SECRET = os.getenv("LIVEKIT_API_SECRET")
+
+#     token = api.AccessToken(LIVEKIT_API_KEY, LIVEKIT_API_SECRET)
+#     token.identity = user_name
+#     token.add_grant(api.VideoGrant(room="yoruba-demo-room"))
+#     jwt = token.to_jwt()
+
+#     return {"url": LIVEKIT_URL, "token": jwt}
+
+
+# Google Cloud project ID
+PROJECT_ID = "credentials/oluko-475823"
+
+
+# ===========================
+#  LOAD ASR MODEL
+# ===========================
+asr = pipeline("automatic-speech-recognition", model="NCAIR1/Yoruba-ASR")
+print("✅ Done loading models!\n")
+
+
+# ===========================
+#  TRANSLATION FUNCTION
+# ===========================
+def translate_text(
+    text: str,
+    source_language_code: str = "yo",
+    target_language_code: str = "en-US",
+):
+    """Translate Yoruba text into English (or other languages)."""
+    client = translate_v3.TranslationServiceClient()
+    parent = f"projects/{PROJECT_ID}/locations/global"
+
+    response = client.translate_text(
+        contents=[text],
+        parent=parent,
+        mime_type="text/plain",
+        source_language_code=source_language_code,
+        target_language_code=target_language_code,
+    )
+
+    # Return the first translation result
+    return response.translations[0].translated_text
+
+
+# ===========================
+#  ASR PROCESSING FUNCTION
+# ===========================
+# def process_audio(file_path):
+#     """Convert speech → Yoruba text using ASR."""
+#     if not file_path:
+#         return "..."
+
+#     audio, sr = librosa.load(file_path, sr=16000)
+#     result = asr(audio)
+#     transcription = result.get("text", "")
+#     return transcription
+
+
+# ===========================
+#  GRADIO INTERFACE
+# ===========================
+with gr.Blocks(title="Yoruba AI Platform", theme=gr.themes.Glass()) as app:
+    gr.Markdown("# 🇳🇬 Olùkọ́")
+    gr.Markdown(
+        "Use this app to practice your Yoruba speaking."
+    )
+
+    # --- Tab 1: ASR + Translator ---
+    with gr.Tab("🎧 Yoruba Speech-to-Text + Translator"):
+        gr.Markdown("### Talk, we'll listen!")
+        audio_input = gr.Audio(type="filepath", label="🎙️ Speak Yoruba")
+        output_transcription = gr.Textbox(
+            label="✍️ Transcription (Yoruba)", interactive=True
+        )
+        output_translation = gr.Textbox(label="💬 Translation (English)")
+
+        # Link actions
+        audio_input.change(
+            process_audio,
+            inputs=audio_input,
+            outputs=output_transcription,
+        )
+        output_transcription.change(
+            translate_text,
+            inputs=output_transcription,
+            outputs=output_translation,
+        )
+
+    # --- Tab 2: LiveKit Conversational Agent (placeholder) ---
+    with gr.Tab("🗨️ Yoruba Conversationalist"):
+        gr.Markdown(
+            "### Real-time Yoruba Conversation\n"
+            "This feature connects to a LiveKit-powered conversational AI that lets you practice Yoruba speaking in real time."
+        )
+        gr.HTML(
+            """
+            <script src="https://google.com"></script>
+            <h3>🎙️ Start a Yoruba conversation</h3>
+            <button id="join-btn">Join Live Session</button>
+            <div id="video-area"></div>
+            <script>
+              async function joinSession() {
+                const res = await fetch('/create_token');
+                const data = await res.json();
+
+                const room = new LivekitClient.Room();
+                await room.connect(data.url, data.token);
+
+                const localTracks = await LivekitClient.createLocalTracks({ audio: true });
+                await room.localParticipant.publishTracks(localTracks);
+                
+                room.on('trackSubscribed', (track) => {
+                    const el = track.attach();
+                    document.getElementById('video-area').appendChild(el);
+                });
+              }
+
+              document.getElementById('join-btn').onclick = joinSession;
+            </script>
+            """
+        )
+
+
+# ===========================
+#  LIVEKIT TOKEN ROUTE
+# ===========================
+# @app.get("/create_token")
+# def serve_token(request):
+#     name = request.query_params.get("name", "guest")
+#     return create_livekit_token(name)
+
+
+# ===========================
+#  APP LAUNCH
+# ===========================
+if __name__ == "__main__":
+    app.launch(server_name="0.0.0.0", server_port=7860)

requirements.txt

@@ -0,0 +1,17 @@
+gradio
+librosa
+torch
+transformers
+huggingface-hub
+google-cloud-speech
+google-cloud-translate
+google-cloud-texttospeech
+requests
+livekit
+livekit-agents
+livekit-plugins-openai
+livekit-plugins-google
+livekit-plugins-silero
+livekit-plugins-spitch
+livekit-plugins-turn-detector
+livekit-plugins-noise-cancellation