Update app.py

app.py

@@ -5,25 +5,54 @@ from gtts import gTTS
 import torch
 import os
 
-# --- 1. YOUR COMPANY DATA (Strict Instructions added) ---
+# --- 1. YOUR COMPANY DATA ---
+# This is the "Brain" of your company. Edit this text!
 COMPANY_KNOWLEDGE = """
-You are the reception AI for 'TechNova Solutions'.
-Rules for answering:
-1. Answer ONLY based on the facts below.
-2. Keep answers EXTREMELY SHORT (Maximum 1 sentence).
-3. Be direct. Do not say "Hello" or "Sure" every time. Just give the answer.
-
-FACTS:
-- WHAT WE DO: We build custom solar-powered drones for agriculture.
-- PRICING: Basic drone is $500. Pro drone is $1200.
-- RETURN POLICY: 30-day money-back guarantee.
-- CONTACT: Email help@technova.com or call +1-555-0199.
-- HOURS: Mon-Fri, 9 AM to 5 PM EST.
+You are the Senior Tech Consultant AI for 'SoftStream Tech', a software development agency.
+Your goal is to answer client questions professionally, technically, and concisely.
+INSTRUCTIONS:
+1. Answer in 20 WORDS OR LESS. 
+2. Be direct. Do not use filler words like "Thank you for asking" or "I would be happy to help".
+3. If the answer is a list, pick only the top 2 items.
+
+RULES FOR ANSWERING:
+1. Keep answers SHORT (1-2 sentences max). Clients are busy. The answer shoulb be within the 25 seconds.
+2. If asked about price, give ranges, not exact quotes.
+3. If you don't know the answer, say: "I'll need to check with a senior engineer on that."
+
+DATA SHEET:
+[Services]
+- Custom Web Development: React, Vue, Next.js, Python (Django/FastAPI), Node.js.
+- Mobile App Development: Flutter (Cross-platform), Swift (iOS), Kotlin (Android).
+- AI & Machine Learning: Chatbots, Predictive Analytics, Computer Vision.
+- Cloud DevOps: AWS, Google Cloud, Azure setup and CI/CD pipelines.
+
+[Pricing Models]
+- Time & Material (Hourly): $40 - $80 per hour depending on developer seniority.
+- Fixed Price: Minimum project size is $5,000. Requires detailed scope.
+- Retainer: Dedicated team for $4,000/month per developer.
+
+[Process]
+- Methodology: Agile/Scrum with 2-week sprints.
+- Tools: We use Jira for tracking and Slack for communication.
+- Timeline: MVP (Minimum Viable Product) usually takes 4-8 weeks.
+
+[Support & Maintenance]
+- Standard Support: Bug fixing for 3 months after launch (Free).
+- Premium SLA: 24/7 server monitoring and priority support ($500/month).
+
+[Contact]
+- Email: projects@softstream.tech
+- Phone: +1-555-CODE-NOW
+- Location: San Francisco, CA (but we work remote globally).
 """
 
 # --- 2. SETUP ---
 # Replace with your actual token
-hf_token = os.getenv("HF_TOKEN") or "YOUR_HF_TOKEN_HERE"
+hf_token = os.getenv("HF_TOKEN")
+
+if not hf_token:
+    raise ValueError("Error: HF_TOKEN is missing. Go to Settings > Secrets and add it!")
 client = InferenceClient(api_key=hf_token)
 
 device = "cuda" if torch.cuda.is_available() else "cpu"
@@ -34,28 +63,28 @@ def voice_chat(audio, history):
         return None, "", history
 
     # A. Initialize History with Company Knowledge (if empty)
+    # This ensures the AI reads your company info before the first message
     if not history:
         history = [{"role": "system", "content": COMPANY_KNOWLEDGE}]
 
     try:
-        # B. Hear
+        # B. Hear (Whisper)
         transcription = whisper_model.transcribe(audio, fp16=False)["text"]
         
-        # C. Think
-        # We append the user text
+        # C. Think (AI with Context)
         history.append({"role": "user", "content": transcription})
         
         response = client.chat.completions.create(
             model="HuggingFaceH4/zephyr-7b-beta", 
             messages=history, 
-            max_tokens=60, # Reduced to 60 to force brevity
-            temperature=0.5 # Lower temperature = more precise/less creative
+           max_tokens=50,  # <--- STRICT LIMIT: ~30-40 words max
+            temperature=0.5 # <--- Low temp keeps it robotic and concise
         )
         ai_text = response.choices[0].message.content
 
         history.append({"role": "assistant", "content": ai_text})
 
-        # D. Speak
+        # D. Speak (gTTS)
         tts = gTTS(text=ai_text, lang='en')
         audio_path = "response.mp3"
         tts.save(audio_path)
@@ -63,22 +92,25 @@ def voice_chat(audio, history):
         return audio_path, ai_text, history
 
     except Exception as e:
+        # If there is an error, return the existing history so we don't lose it
         return None, f"Error: {str(e)}", history
 
 # --- 3. INTERFACE ---
 with gr.Blocks(theme="soft") as demo:
-    gr.Markdown("## 🏢 Quick-Response Company AI")
+    gr.Markdown("## 🏢 TechNova AI Receptionist")
     
+    # Initialize state as empty list
     conversation_history = gr.State([])
 
     with gr.Row():
-        input_audio = gr.Audio(sources=["microphone"], type="filepath", label="Ask Question")
+        input_audio = gr.Audio(sources=["microphone"], type="filepath", label="Ask about our company")
     
     with gr.Row():
         output_audio = gr.Audio(label="AI Response", autoplay=True)
         output_text = gr.Textbox(label="Transcript")
     
-    clear_btn = gr.Button("Reset")
+    # Add a clear button to reset the conversation
+    clear_btn = gr.Button("Reset Conversation")
     
     input_audio.change(
         voice_chat, 
@@ -86,6 +118,7 @@ with gr.Blocks(theme="soft") as demo:
         outputs=[output_audio, output_text, conversation_history]
     )
     
+    # Reset logic
     clear_btn.click(lambda: ([], None, ""), outputs=[conversation_history, output_audio, output_text])
 
 demo.launch(debug=True)