Streamline App Process: From guide to dashboard app via hero AI manager #24

New Issue

mik-tf · 2024-07-23T16:01:44Z

mik-tf commented

2024-07-23 16:01:44 +00:00

Proposition Overview

We propose to streamline the process of building an app on the dashboard
This is done in two phases:
- from manual to flist
- from flist to dashboard app

Uniform Method of Using the TFGrid

The manual now documents the essential steps and a uniform method can be built:
- Uniform method of working with the TFGrid:
  - Using a micro VM
    - can then be easily translated to flist steps
  - Set services with zinit
    - as micro VM works with zinit and not systemd
  - Write bash CLI scripts taking variables
    - in the manual, we give steps to users as CLI commands, so they do not execute a script per se, but the bash codes are acting as scripts (i.e. CLI scripts, not File scripts).
  - Give steps to user for steps done by the user and not the app
    - this info will be written in the README.md

Hero Role

Hero takes this markdown guide and turns it into heroscript to build an app
- 1st phase
  - Turn text in Docker image
    - with README.md of all non-app steps made by user
    - with Dockerfile, zinit and scripts files + directories.
  - Turn Docker into Flist
    - the Flist's README.md contains all information to deploy the flist on the tfgrid
- 2nd phase
  - Hero turns flist into dashboard app

Markdown File Structure

Markdown file structure
- the markdown file uses a set of headings ensuring optimal hero conversion
  - type of environment
    - micro VM characteristics
  - preparing the environment
    - apt update && apt install -y ...
  - running CLI commands to deploy
    - turn this into scripts and/or Dockerfile steps

Major Aspects

Major aspects:
- differentiating between what goes into Dockerfile and what goes into scripts will be crucial for effective conversion process
- setting effective yaml zinit files will be crucial for effective conversion process

UX

As writers of guides and apps, we first do some explorations (R&D phase). Then, when we have a working app/product, we write the steps in a markdown file, then hero converts the steps into an flist (as per tf-images repo structure). Finally, Hero converts the Flist into a Dashboard app (as per tfgrid-sdk-ts repo structure).

R&D Stage: test on a micro VM the steps to build the app
- e.g. dashboard suite, zos boot generator, ubuntu image, nextcloud app, casperlabs, etc.
Manual Stage: write the steps in a markdown file with distinct section
- manual section e.g.:
  - prerequisites
  - deployment
  - etc.
- hero conversion e.g.:
  - Dockerfile, zinit, scripts, README.md
- each section constitutes logical section necessary for a docker image
  - e.g. scripts go into /scripts
  - yaml files go into /sbin/zinit
Hero conversion Stage: Hero
- takes the manual markdown files of the steps and
- turns it into a Docker image
- it pushes the image to Docker Hub
- then use the TF Hub to convert from docker to flist
- then creates a dashboard app with all the necessary environment variables and Vue layout

Two Actions: App Actions, User Actions

There are two types of actions, user action, e.g. set the DNS record of a domain (can't be automated directly), and app action, e.g. the app sets the environment (written in scripts, can be automated).

We use Hero AI assistant to easily go from one step to another of the TF App process.
- Hero can differentiate when steps are done by the flist/dashboard app OR by the user (e.g. set DNS A Record is done by user, loading latest TF repos is done by the flist, etc.)

Repos

The structures should follow existing TFGrid repos to ensure flowing CI/CD, e.g.
- tf-images:
  - https://github.com/threefoldtech/tf-images
- tfgrid-sdk-ts
  - https://github.com/threefoldtech/tfgrid-sdk-ts

## Proposition Overview - We propose to streamline the process of building an app on the dashboard - This is done in two phases: - from manual to flist - from flist to dashboard app ## Uniform Method of Using the TFGrid - The manual now documents the essential steps and a uniform method can be built: - Uniform method of working with the TFGrid: - Using a micro VM - can then be easily translated to flist steps - Set services with zinit - as micro VM works with zinit and not systemd - Write bash CLI scripts taking variables - in the manual, we give steps to users as CLI commands, so they do not execute a script per se, but the bash codes are acting as scripts (i.e. CLI scripts, not File scripts). - Give steps to user for steps done by the user and not the app - this info will be written in the README.md ## Hero Role - Hero takes this markdown guide and turns it into heroscript to build an app - 1st phase - Turn text in Docker image - with README.md of all non-app steps made by user - with Dockerfile, zinit and scripts files + directories. - Turn Docker into Flist - the Flist's README.md contains all information to deploy the flist on the tfgrid - 2nd phase - Hero turns flist into dashboard app ## Markdown File Structure - Markdown file structure - the markdown file uses a set of headings ensuring optimal hero conversion - type of environment - micro VM characteristics - preparing the environment - apt update && apt install -y ... - running CLI commands to deploy - turn this into scripts and/or Dockerfile steps ## Major Aspects - Major aspects: - differentiating between what goes into Dockerfile and what goes into scripts will be crucial for effective conversion process - setting effective yaml zinit files will be crucial for effective conversion process ## UX As writers of guides and apps, we first do some explorations (R&D phase). Then, when we have a working app/product, we write the steps in a markdown file, then hero converts the steps into an flist (as per tf-images repo structure). Finally, Hero converts the Flist into a Dashboard app (as per tfgrid-sdk-ts repo structure). - R&D Stage: test on a micro VM the steps to build the app - e.g. dashboard suite, zos boot generator, ubuntu image, nextcloud app, casperlabs, etc. - Manual Stage: write the steps in a markdown file with distinct section - manual section e.g.: - prerequisites - deployment - etc. - hero conversion e.g.: - Dockerfile, zinit, scripts, README.md - each section constitutes logical section necessary for a docker image - e.g. scripts go into /scripts - yaml files go into /sbin/zinit - - Hero conversion Stage: Hero - takes the manual markdown files of the steps and - turns it into a Docker image - it pushes the image to Docker Hub - then use the TF Hub to convert from docker to flist - then creates a dashboard app with all the necessary environment variables and Vue layout ## Two Actions: App Actions, User Actions There are two types of actions, user action, e.g. set the DNS record of a domain (can't be automated directly), and app action, e.g. the app sets the environment (written in scripts, can be automated). - We use Hero AI assistant to easily go from one step to another of the TF App process. - Hero can differentiate when steps are done by the flist/dashboard app OR by the user (e.g. set DNS A Record is done by user, loading latest TF repos is done by the flist, etc.) ## Repos - The structures should follow existing TFGrid repos to ensure flowing CI/CD, e.g. - tf-images: - https://github.com/threefoldtech/tf-images - tfgrid-sdk-ts - https://github.com/threefoldtech/tfgrid-sdk-ts

mik-tf added the

Story

label 2024-07-23 16:01:44 +00:00

mik-tf commented

2024-07-23 16:06:45 +00:00

@despiegk
Pinging you so you can have a look. No rush, might be for 3.20. Didn't set a project yet.

@despiegk Pinging you so you can have a look. No rush, might be for 3.20. Didn't set a project yet.

Sign in to join this conversation.