TensorFlow.js Full-Stack Starter Kit

Coming to work on an AI project there are lots of things to consider, and getting some proof-of-concept going with a bunch of tools is the way to go. But once this phase of the project is over you want to develop an application or even just a testbed to work on. You need an applicative environment to develop on. Recently, when I tried setting up an application stack for an AI/TensoFlow based project, I found that the MEAN-based templates I used in the past are way behind and do not fully integrate the latest technologies. So I decided to put together the libraries I found working for me, as a new full-stack template.

Google's TensorFlow is run here inside a NodeJS environment - V8 single-threaded Javascript engine that is HW-accelerated with eighter WebGL or CUDA binaries. It seems that training a large-complex model on this environment is a no-go, even on a strong machine. But (1) using it to explore a model, visualize and adjust it's parameters is a good candidate. A more reasonable direction is to (2) take a large model built elsewhere, convert it, and use this backend to serve clients' requests. It's easy to run such an app on the cloud/K8S and scale it horizontally.
Regardless of the AI engine here, this is a (3) nicely integrated set of libraries and a modern developement environment to jumpstart any web/mobile/native project.

A well integrated, full-stack working sample:

• TensorFlow.js backend engine preloaded with a simple model to see it in action.
• Visualization library integrated into a responsive web/mobile application: a dashboard to exercise the above model and visualize its predictions.
• One coding language across the stack - TypeScript/Javascript fully ES6 with async/await across all code.
• Latest APIs based on a GraphQL schema.
• Modern dev tools with hot-reload on changes, linting, unit testing, and code coverage reports. Well integrated into VSCode as an IDE and debugger.
• Keep it simple philosophy across! I've saved a ton of learning-curve time by not including a bunch of libs that I found old-fashioned or cumbersome: Redux, Apollo server/client, Enzym, REST, yarn, Babel, Winston/Bunyan, Relay...
• As a bonus, you get the trivial Todo application integrated here as well. So this can work for you "even" if this is not an AI-based project.

• TensorFlow.js engine
• GraphQL based API to feed the - Node.JS/Typescript environment and tooling with mondemon/hot-reloading TFJS engine
• KoaJS web-server, middleware and health-checks
• typegql - Simplified GraphQL schema based on TypeScript annotations
• GraphiQL in-browser IDE for exploring GraphQL schema
• Dotenv-expand for environment-based config (see .env file)
• Jest - unit tests with good coverage
• VSCode launch settings

• Vega-lite - for charting and visualizations
• Ant Design/Typescript React components - responsive dashboard
• LESS/LCSS - better styling
• ts-fsm - simplified state managenent for React
• graphql-request - simplified GraphQL client
• Jest - React-based unit tests
• Node.JS/Typescript tooling, including webpack with babel and hot-module-replacement (HMR)
• Client is based on react-create-app with antd-demo scripts and is not ejected.

The repo is divided into a server-side (backend) project and a client-side (frontend) project. The Backend runs the TensorFlow model and an API to call it. The Client is a single-page responsive-app that calls the model, thru the API, and presents a graph of the results.

1. TensorFlow model is implemented in src/tensorFlowProvider folder as a 'Model Provider'. It follows a simple init >> train >> compile >> predict workflow. It is easy to implement a similar provider based on some other AI engine if a different one is needed.
2. The client side of the model is implemented in client/src/components/server-predict-card. Where you can find an input-output form, a visualizations panel and a small model to drive them. The Visualization panel specification is found in VegaLiteSpec file. You can find a bunch of examples for charting specs on the Vega site.
3. The Backend API implements on a GraphQL Schema in src/graphqlApi and served with a Schema Browser (dev builds only).
4. The client side of the API is implemented as a general-purpose GraphQL client in client/src/utils/graphqlQuery and a bunch of interfaces in client/src/components/graphql-types.
5. Performance: As noted above, TensorFlow.js on the backend a very different animal than the "standard" TF-Python based environment. I could not find good benchmarks. Happy if you refer me or test yourself.

• NodeJS 8+ installed locally
• Git, VSCode or just a terminal
• npm install -g create-react-app react-app-rewired

> git clone https://github.com/eram/tensorflow-stack-ts.git stack
> cd stack
> npm install 
> npm build:live
> cd client
> npm install
> npm start

• Hoveron and zoom in chart
• Add model training page next to dashboard
• TFJS to use tfjs-node (canvas/WebGL based) and/or CUDA libraries
• Load client input from csv/file upload
• Join client and server into one repo.
• TLS cert and serve https
• Add DB model based on for backend input/output stores
• Add DB interafce (SQLite for local-dev, PostgreQLs for prod)
• Dockerize and integrate into Google devops stack
• PM2 config script and PMX integration: cluster manager, central logging, central counters, hot environment changes and health-checks.
• User management and JWT
• Implement error-chain across API data-flow instead of using throws.
• Auto patch level versioning on build
• Replace nodemon with ts-node-dev (don't restart the server on file change).

!!! In general, I highly advise you not to mess around with the json conf files. it's fragile... better UNDO!

> npm run clean
> npx jest --clearCache
> rimraf node_modules
> rimraf package-lock.json
> Slide back to a known-to-be-working time and ``npm install``
> npm cache clean --force

Happy to get remarks, assistance with getting the todos done, issues and pull-requests. Also happy to share admin and get more devs on board. If you need some consultation with your project - talk to me.

Cheers!

eram at weblegions dot com