When try remove Test env.

List with environments are blank.

List are filled after re login to site and force site refresh (Ctrl + F5).

Hi bullet-train developers, it is a great project.

I am try to run bullet-train-api and bullet-train-frontend locally, and I edited the file env/project_dev.js api to my local server in bullet-train-frontend and run it using npm run dev.

Most of the functions seem working fine, except I found it keep calling https://api.bullet-train.io/api/v1/flags/ to retrieve feature flags. May I know which part of the config I am missing, so that it can get the flags from my local server?

When you do a search in the feature flag page, when you activate/deactivate the first one in the list, it will activate/deactivate the first one in the unfiltered list, not the filtered list.

As today more and more services offer dockerized deployment we would like to have this ability also in the frontend server.

Currently, we are implementing flagsmith behind a gateway in a k8s cluster. For the frontend specifically, we have set the ASSET_URL to the appropriate path. E.g. http://localhost:4200/flags-admin/

In the webpack config, using url.resolve(), this ultimately translates to http://localhost:4200/flags-admin/static/

However, there are several issues when doing this. For example, the app attempts to load images and fonts at /images/ and /fonts/. This is easily resolved by adding matching routes to the gateway proxy.

The main issue is that the app is rewriting the browser history, under the assumption that it lives at the root of the domain, and typical user behavior causes it to fail (such as refreshing a page, etc). Additionally, accessing the primary login screen is also a challenge. The app will not render the login screen despite all requests coming back 200. You can "trick" the app to render the screen by setting t cookie with a bad value, or by navigating to any of the paths defined here. This will force the app to render the screen and add a redirect query param. This however does not resolve the usability issue as the app could easily fail should a user manually refresh or click a bookmarked link, etc.

It would make the most sense to refactor the routing system to respect the ASSET_URL across the codebase. This would
then allow support for "/" as easily as "/custom-path".

Please,

Report which version of the frontend is compatible with the backend.

Hi guys, is there a reason you're using the "regular" 300+ MB node Docker image instead of the ~36 MB node alpine image? Otherwise I think that would be a better choice. And perhaps upgrade to node 12 or 13 at the same time.

I'd love to make a PR if you want me to.

Hi,

In the Dockerfile it is instructed to use
RUN npm ci --only=production
instead of
RUN npm install
if we are building for production.

When I do this (API in project_prod.js configured to be a local deployment of the API service) I get the following error:

> [email protected] dev /usr/src/app
> cross-env npm run env && node_modules/.bin/nodemon --watch server --ignore server/config/config.json --watch webpack --exec node ./server


> [email protected] env /usr/src/app
> node ./bin/env.js

Using project_dev.js
sh: 1: node_modules/.bin/nodemon: not found
npm ERR! code ELIFECYCLE
npm ERR! syscall spawn
npm ERR! file sh
npm ERR! errno ENOENT
npm ERR! [email protected] dev: `cross-env npm run env && node_modules/.bin/nodemon --watch server --ignore server/config/config.json --watch webpack --exec node ./server`
npm ERR! spawn ENOENT
npm ERR! 
npm ERR! Failed at the [email protected] dev script.
npm ERR! This is probably not a problem with npm. There is likely additional logging output above.

npm ERR! A complete log of this run can be found in:
npm ERR!     /root/.npm/_logs/2019-11-05T20_45_49_276Z-debug.log

create pipeline for build production and release docker image for distribution

At the moment, this repository contains both the official Bulet Train web site and the feature management panel.

I suggest to split the repository into two separate projects, web page and feature panel, for several reasons.

• Logical separation: The official appearance of the project and management of the features are not influenced by each other.
• Self-hosting only needs the feature management panel but no YouTube video, Pricing, Project Description, …
• Side effects: Introducing a bug in the panel code could bring down the whole web site
• etc. (I think I do not need to elaborate more because it is probably clear anyway)

In the same run, the API keys from production config files can be moved to environment variables, for example. No commits required anymore when an API key changes. And they are not public anymore …

What is your opinion on that? And is this possible in near future?

Currently, there is audit log capability - which can all logs across feature flags.
Also, there is a search box, which can help filter audit logs for a specific feature flag.

But, it would be nice to have audit logs accessible against each feature flag, for easy access. This can help debug issues - to see when and who enabled a specific feature flag.

I tried to run the frontend behind a reverse proxy with some URL prefix.
Routes being served kind of like this:

• /flagsmith/ui/ to the flagsmith frontend address
• /flagsmith/api/ to the flagsmith api address
• /something/else/ to somewhere else

The prefix is always trimmed in the request to the actual backend.

Unfortunately, this causes problems for the frontend. For example, in the page HTML, there are references to images starting with /images/... which tells the browser to query /images/... instead of /flagsmith/ui/images, which fails.

I could not find instructions to set a certain base url or base path to use, like some projects do.
Is there anything that I missed? Maybe serving flagsmith on its own domain a better idea either way?

Value proposition

As a Bullet Train user, I want to see descriptions in the feature list so that I can gain more context around what the feature flag represents

Context

We use the Jira ticket ID as the feature name so we can automate our feature addition/removal flow (webhooks are fired when stories are transitioned into development and transitioned into feature flag removed states).

We use the description to add a link to the ticket itself, so Jira can be our source of truth for all feature flag work. It's a small hassle to click on each feature when we want to gain access to the link.

frontend breaks and does not start if the GOOGLE_SERVICE_ACCOUNT integration code is not provided.

It only worked after commenting on the bullet-train-api / src / analytics / query.py integration in the get_events_for_organisation method and forcing return 0

I suggest making integration with google analitycs optional.

when building the frontend the build process installs npm packages, which comes with the chromium for testing, resulting in a 491 mb image, I suggest using the build process to generate static files and use the nginx or apache image that will be reduced in size and optimized for production use

for that purpose I also suggest using multistage build which would also optimize the build phase.

see: https://docs.docker.com/develop/develop-images/multistage-build/

Hi folks! First of all, thank you for this product! My team is evaluating using Bullet Train as our main feature flagging/rollout tool.

We loved all we saw so far, only one thing we found out was when trying to disable a specific feature flag override, when we click in the toggle option, the feature flag modal just closes.

We tested on the demo app, so, if you want to try it out, here's the link: https://app.bullet-train.io/project/1016/environment/E6Mf24wZHMVwPr5Rqtnh9M/users/879489#

(don't know if it expires).

Also, here goes a gif showing what I mean:

In this GIF (in specific user settings screen):

1. Open feature details inside user screen
2. Go to "Identity Overrides" setting
3. Try to click on toggle to disable override
4. The modal just closes

Don't know if it helps, but I've seen the following error on my console:

TypeError: e.props.router is undefined CreateFlag.js:270:52
    onClick CreateFlag.js:270
    React 14
    Sn self-hosted:920
    a helpers.ts:77
    (Async: EventListener.handleEvent)
    wrapEventTarget trycatch.ts:123
    React 15
    <anonymous> main.js:64
    a helpers.ts:77
    (Async: setTimeout handler)
    wrapTimeFunction trycatch.ts:32
    <anonymous> main.js:50
    <anonymous> main.js:44
    Webpack 3

Thank you!

Steps to reproduce

1. Open a feature for editing
2. Change the description (or any field, potentially)
3. Attempt to update the feature

Expected behaviour

The feature is updated.

Actual behaviour

The POST request to /api/v1/projects/{projectId}/features/{featureId}/ results in the following 400 response:

{"project":["This field is required."]}

Steps to reproduce

Create 11 features

Expected result

11 features should be displayed in the features list

Actual result

10 features are displayed in the features list

I'm trying to setup and configure a standalone instance of this to run on Heroku and am running into some issues that I can't seem to resolve. I've done my best to configure everything the way the README says to, but when I try to run it in my browser, I get the following error:

Uncaught ReferenceError: Loader is not defined
    at Object.<anonymous> (AppLoader.js:12)
    at n (bootstrap:19)
    at Object.<anonymous> (AccountProvider.js:3)
    at n (bootstrap:19)
    at Object.<anonymous> (project-components.js:20)
    at Object.<anonymous> (project-components.js:94)
    at n (bootstrap:19)
    at Object.<anonymous> (main.js:4)
    at n (bootstrap:19)
    at bootstrap:83
    at bootstrap:83

We have an instance of the API running and it seems to be working as expected, and I've setup the following environment variables on Heroku as well:

API_URL = https://features-api.theartofeducation.edu/api/v1/
ASSET_URL = /static/

Any ideas what I'm missing or what I need to do differently?

We currently get around this missing feature by defining convention like:

• show_new_grid_a
• show_new_grid_b
• show_new_grid_c

Would be nice to put them all in a category for quick visualization

• New_grid_transition
  • show_new_grid_a
  • show_new_grid_b
  • show_new_grid_c

But it would be nice to be able to create category to quickly see without doing a search.

I don't think category should be something related to the API, it should only be used for classification help.