Dialogflow CX uses a state machine approach with visual flows to simplify the management of complex conversational states.
Structure complex assistants using flows for high-level features and pages for individual states to create reusable conversational components.
Natural Language Understanding (NLU) works by matching user utterances to defined intents and extracting key parameters as entities.
Follow the iterative lifecycle of designing, building, training, and evaluating the NLU model to improve your assistant's accuracy.
The system architecture involves a client application calling the Dialogflow CX API, which processes the state and can trigger external webhooks.
A recommended backend uses Google Cloud Functions for serverless webhooks, Cloud Storage for assets, and Firebase for a persistent database.
Use the available client SDKs in languages like Python and JavaScript to programmatically manage agents as part of a structured development workflow.
Validate your assistant using the built-in test agent for manual checks and save interactions to create automated test cases for CI/CD.
Connect your bot using plug-and-play integrations like Facebook Messenger or build a man-in-the-middle service for custom channels like Telegram.
This walkthrough of the user interface shows how to navigate flows, pages, routes, intents, and entity types to build a bot.
A community-built command-line interface allows developers to perform CRUD operations and test Dialogflow CX agents from the terminal.
The CLI's NLU profiler uses YAML files to create and run automated test suites for validating intent and entity matching.
Use webhooks to call large language models like Google's PaLM or Vertex AI to generate more natural and varied bot responses.
Dialogflow CX offers an easier development experience than competitors while leveraging the security and compliance of the Google Cloud Platform.
Create separate agents for languages with very different interaction patterns and use official tools to help migrate from Dialogflow ES.
Connect to external systems like CRMs via the API and justify bot development to leadership by presenting data on time saved.
msg
Ismaning, Germany
58:06Anna Fritsch-Weninger
59:39Julian Joseph
58:00Vijay Krishan Gupta & Gauravdeep Singh Lotey
56:55Ekaterina Sirazitdinova
52:37Kateryna Hrytsaienko
30:24Markus Walker
30:59Lee Boonstra
48:27Felix Augenstein
Jobs that call for the skills explored in this talk.
ONTRAC SOLUTIONS LLC
Landbot
Municipality of Madrid, Spain
