Add web demo for AI Real-Time Translation Bridge

[harpreetTelnyx]

What this adds

A browser-based demo of the translation engine from the AI Real-Time Translation Bridge example. Same Telnyx AI Inference API, same model, same prompt — just a web UI instead of telephony.

Why

The phone-call sample requires a purchased number, a Call Control application, webhook tunneling (ngrok), and two test phones to see it work. That's a lot of setup before you can verify the translation engine actually works. This web demo lets you see the same translation live in a browser with just an API key — useful for demos, videos, and quick verification.

What's in the PR

File	Purpose
demo/app.py	Flask backend, /translate calls Telnyx AI Inference (zai-org/GLM-5.1-FP8)
demo/templates/index.html	Web UI — dual chat panels (Caller A / Caller B) + live pipeline visualization (Speech → STT → AI Translate → TTS). Vanilla JS, no build step
demo/requirements.txt	Flask, requests, python-dotenv
demo/.env.example	Needs only TELNYX_API_KEY
demo/README.md	Run instructions + mapping table showing how the web demo maps to the phone-call sample

How it maps to the phone-call sample

Phone call sample	Web demo
Caller speaks into phone	Type in Caller A/B input
Telnyx Call Control webhook	/translate endpoint
STT via Telnyx Call Control	Simulated (text input)
translate() → AI Inference	Same — real API call
TTS played back to caller	Translated text appears in panel

The translation engine is identical — same endpoint, same model, same system prompt. The demo just replaces telephony I/O with a web UI.

Languages supported

English, Spanish, French, German, Italian, Portuguese, Hindi, Arabic, Chinese, Japanese, Korean, Russian.

Run locally

cd ai-real-time-translation-bridge-python/demo
cp .env.example .env          # add TELNYX_API_KEY
pip install -r requirements.txt
python app.py                 # open http://127.0.0.1:5001

Get an API key at https://portal.telnyx.com/#/app/ai/keys.

Verification

Tested all three demo turns locally via Playwright:

• EN → ES: "Hello, how are you today?" → "Hola, ¿cómo estás hoy?" ✓
• ES → EN: "Estoy muy bien, gracias. ¿Y tú?" → "I'm very well, thank you. And you?" ✓
• EN → JA: "I would like to book a hotel room for three nights" → "3泊のホテルの部屋を予約したいのですが。" ✓

Notes

• The phone-call sample (app.py at the example root) is unchanged. This PR only adds a demo/ subdirectory.
• Default model is zai-org/GLM-5.1-FP8 (Telnyx-hosted, multilingual). Configurable via AI_MODEL in .env.
• The model is a reasoning model, so translations take ~2-3s per turn. This is expected — the reasoning step happens before the translation is returned.
• .env is gitignored (confirmed: git check-ignore covers demo/.env).