ciscosystems / cognitive Goto Github PK

Users need to save and reload already created workspaces.

• Create the list of saved workspaces
• Create dialog page for creating new workspace
• Get the nodes and edges information when existing workspace is opened

Will add 'gulp' for watching and compiling *.ts to *.js

We need some automated test execution like: tox -e py27
Also need to define the location of test codes and it will be better if we have some sample test codes, because it enable future contributor to add new test easily.

currently GET /api/v1/operation/{some_component_name}/{id} does not have the detail parameters for the component. Let's expose the component parameters, which can be useful for front end and other services as well.

It should be supported for edit and delete functionality in the component graph of main UI

These components should be able to be removed via UI actions.

• change the way to manage the connection links between these components
  • Remove the links related to the component when it's removed
• call delete APIs when deleted

Results are not shown when only one component is present(input data component).

Reason: Workflow graph is not created when only one component is present in the experiment

Currently [POST|GET] /api/v1/oprations/* APIs do not have the consistent json format. So let's modify the response format for example:
{ params: {constant_value: 1, target: [1,2]...}, outputs: [{id:1, name:"age"}, ...], success: true, ... }
for params, it will describe component-dependent values, for outputs, it will show the column indices and names for the next component.

This is just a idea, any ideas to improve this is appreciated.

Let's use travis CI or Circle CI or Jenkins as a CI environment.
To enable travis CI for example, one has the rights to manage this repository have to activate travis service. after that, we can add .travis.yml file.

front-end send whole topology information to backend, and backend will calculate each path

like check mark for component

1. anyone can contribute via pull request, not directly push to master
   • if we have review systems, it's possible to push master, but for me, I prefer to keep current way.
2. how to execute test
   • some setup or dependencies and execution command

For future contributor, it's likely to say that the changes they have to do to add new functionality should not be widely. For reducing the change areas, we would like to support the plugin which is the set of backend logic with UI part. This allows contributors to easily add new component as a plugin.

Need to discuss:

1. the location of plugin directory which can compose Javascript, html, css, and python logic.
2. Sometimes new component would like to use Database, so how to support.
3. how to load plugins

this is not all, when you have ideas which I'm not aware of, please pose here.

For Column Selection and Remove Duplicates, minus button of the UI does not work correctly.

Selected train.csv from temp folder to try system. System doesn't respond to the add button (does respond to the file selection, just not add).

When selecting a capability from the list (e.g. "math formula"), it's not clear that selections are pulldown lists. This looks like a simple UX glitch.

Use Jasmine.js

Currently UI and backend supports only for one edge for destination and one edge for source node when you add some components from UI.
It's likely to need to support multiple edge support for each component

Let's separate file upload from creating file component

• API /api/v1/data
  • return data_id
• Modify POST /component/inputfile/
  • remove csv centric parameter
  • instead, use data_id

I seem to be able to pick any random input value against any of the "types" of flow components. There's no obvious correlatoin capability.

Perhaps a data browser to select, rather than just a selection of titles as an option for the answer to "what are you working on"?

Currently UI only supports creation of components. Edit and delete components support is required if user adds any components by mistake

develop branch is not used correctly now so renew the develop branch next weekend(26/09/2015). if there are someone who needs to use current develop branch, post here.

To support for multiple user, UIs for Login and Logout are needed

I think it's better to modify the name from MLaaS to cognitive.
Please post comments

For keeping good maintainability especially when project will be bigger, switch from Javascript to Typescript, which allow us OOP way.

Angular JS can be alternative way for good maintainability. but it's still not mature and version 2 will be completely different one. and it's said v2 will support typescript as well

Can't run the server without django.core.management

We are able to automate the generation of documentations for this project using sphinx and tox.

Users need to track the status of components when an experiment is executed. This is particularly useful when the topology takes time to finish execution.

REF: https://github.com/nvbn/django-bower

• install django-bower
• install javascript libraries using django-bower
• remove current javascript libraries
• write the chages on the README.md

• Add Angular Material
• Modify Index page
• Modify Whiteboard page
• Remove bootstrap template

need to add token header for API