Update

app.py

@@ -1,12 +1,12 @@
+
+import gradio as gr
 import os
 import requests
-import gradio as gr
 from gtts import gTTS
 import tempfile
 
-
-TOGETHER_API_KEY = os.environ.get("TOGETHER_API_KEY")
-TOGETHER_MODEL = "meta-llama/Llama-3-70B-Instruct"
+TOGETHER_API_KEY = os.getenv("TOGETHER_API_KEY")
+MODEL = "meta-llama/Llama-3-70B-Instruct"
 
 def speak_text(text):
     tts = gTTS(text)
@@ -14,43 +14,37 @@ def speak_text(text):
         tts.save(fp.name)
         return fp.name
 
-def together_generate(prompt):
-    url = "https://api.together.xyz/v1/completions"
+def respond(message, history):
+    history = history or []
+    prompt = ""
+    for user_msg, bot_msg in history:
+        prompt += f"<|user|> {user_msg}\n<|assistant|> {bot_msg}\n"
+    prompt += f"<|user|> {message}\n<|assistant|>"
+
     headers = {
         "Authorization": f"Bearer {TOGETHER_API_KEY}",
         "Content-Type": "application/json"
     }
-    data = {
-        "model": TOGETHER_MODEL,
+    payload = {
+        "model": MODEL,
         "prompt": prompt,
         "max_tokens": 512,
         "temperature": 0.7,
         "top_p": 0.95,
-        "repetition_penalty": 1.0
+        "repetition_penalty": 1.1
     }
+    response = requests.post("https://api.together.xyz/inference", headers=headers, json=payload)
+    response_json = response.json()
+    output_text = response_json.get("output", "").strip().split("<|assistant|>")[-1].strip()
 
-    response = requests.post(url, headers=headers, json=data)
-    if response.status_code == 200:
-        return response.json()['choices'][0]['text'].strip()
-    else:
-        return "⚠️ API error: " + response.text
-
-def respond(message, history):
-    history = history or []
-    prompt = ""
-    for user_msg, bot_msg in history:
-        prompt += f"<|user|> {user_msg}\n<|assistant|> {bot_msg}\n"
-    prompt += f"<|user|> {message}\n<|assistant|>"
-
-    response = together_generate(prompt)
-    history.append((message, response))
-    audio_path = speak_text(response)
+    history.append((message, output_text))
+    audio_path = speak_text(output_text)
     return history, history, audio_path
 
 with gr.Blocks() as demo:
     gr.Markdown("# Sambit AI 🤖")
     chatbot = gr.Chatbot([], label="Chat with AI", type="messages")
-    msg = gr.Textbox(label="Type your message here", placeholder="Ask me anything...")
+    msg = gr.Textbox(label="Type your message here")
     audio_output = gr.Audio(label="AI Voice", interactive=False)
     clear = gr.Button("Clear Chat")
     state = gr.State([])
@@ -58,4 +52,4 @@ with gr.Blocks() as demo:
     msg.submit(respond, [msg, state], [chatbot, state, audio_output])
     clear.click(lambda: ([], [], None), None, [chatbot, state, audio_output])
 
-demo.queue()  
+demo.launch()

requirements.txt

@@ -1,4 +1,4 @@
 
-gradio
-huggingface_hub
+gradio==5.30.0
+requests
 gTTS