How jackdh/rasatalk works
Project type and purpose
• This is a full-stack, single-page React/Redux application (“app” folder) paired with an Express/MongoDB backend (“server” folder).
• Purpose: Rasa Talk is an on-premise dialog-management tool that sits on top of Rasa NLU. It provides a UI for creating agents, intents, entities, dialog nodes, and training Rasa models. It also supports hooking chatbots into third-party channels (Facebook, Telegram, etc.).
File structure
• Top-level folders:
• app – the React/Redux front end
• server – the Node/Express REST API & MongoDB integration
• internals – build, generator, testing, and CI scripts
• app/ contains: components (presentational), containers (Redux + sagas), configureStore, reducers, sagas, utils, translations, global styles, index.html
• server/ contains: Express app setup, authentication routes, mongo controllers & schemas, third-party adapters, middleware, API tests
• Docker and CI config in docker-compose.yml, .travis.yml, appveyor.yml
README for additional insights
• Confirms tech stack: React, Redux, Redux-Saga, Material-UI, Express, MongoDB, PassportJS, Docker, Rasa NLU
• Describes features: training-data generator, dialog management, webhook integrations, permission system, model training, analytics
• External dependencies: Rasa NLU service, MongoDB, optional Botkit for chat channels
Essential architectural elements
• Separate Presentation vs. Business logic: React components vs. Redux/Saga
• API segmentation: Auth, Agents, Intents, Dialog, Training, Chatbot testing, Third-Party integrations
• Service boundaries: MongoDB, Rasa NLU, Chat channels