Automating Voice Assistants and Chatbots
summary
Describes building a conference voice assistant and chatbot that uses Dialogflow with Google TTS/STT (switching from Amazon for real‑time transcription), TypeScript tests, Selenium, Docker, ngrok and a local Solenoid grid. Reports practical issues: Dialogflow’s 256‑character limit blocks long utterances, intents can return up to 30 randomized responses which breaks repeatability, real‑time STT matches ~0.9 requiring Levenshtein thresholds, and TTS/STT costs plus noisy environments make end‑to‑end voice automation possible but brittle.
AI-written summary · about
chapters (11)
- 0:00Voice assistants intro
- 3:22Choosing platform
- 7:32Cloud limits
- 11:08Local testing issues
- 18:39Chatbot testing tools
- 24:37Live demo
- 34:52Voice testing automation
- 44:08Voice flow results
- 47:32Alternative solutions
- 49:20Conclusions
- 51:08Q&A
related talks
17mBuilding Effective Voice Agents
47mTesting Conversational AI
51mBuilding an AI assistant that makes phone calls
53mBuilding Conversational Experiences with Actions on Google
40mOk Google, we need to talk
46mChatbots, shall we shift into second gear?
33mDesigning Voice Agents for Real Conversations
34mOk Google! We need to talk