I’m building a voice-driven receptionist that greets visitors or callers, listens to their questions, and replies with a natural-sounding voice. The core flow is straightforward: • Speech-to-text: Accurate recognition of English only is required for now. • NLP: Classify the intent and pull the correct answer when a caller asks about our business hours or location. • Text-to-speech: Respond with friendly, human-like audio. • Fallback: whenever the system is unsure, it should politely ask for more details rather than handing the call off or giving a canned line. I need the full stack—STT, intent handling, response generation, and audio playback—wrapped in a module that I can drop into my existing website widget today and expand to a phone line via Twilio (or a similar SIP/VoIP service) later. Keep the integration layer simple: a REST webhook or lightweight SDK is perfect. Acceptance criteria 1. Demo page or endpoint that I can test from Chrome: user speaks, system replies. 2. Correct answers for “What time do you open?”, “When do you close?”, and “Where are you located?”. 3. When asked something unrelated, system requests clarification (“Sorry, could you tell me a bit more about that?”) and logs the transcript. 4. Clear setup instructions and source code (Python, Node, or comparable mainstream language). If you’ve worked with Dialogflow, Whisper, Azure Cognitive Services, Amazon Polly, or similar toolchains, let me know. I’m ready to start as soon as I see a concise plan and timeline.