Environments
One branch, one environment
Every branch of your Semantic Repository is an environment with its own model and its own
outputs. main (or whichever branch you marked as production) is the production
environment; any other is a working environment.
Nothing has to be declared: if someone creates a branch from their terminal, it shows up as an environment. Delete it and it goes away.
Working on a branch
Create an environment from Environments → New environment. It branches off production and inherits its model. From then on, everything you edit — definitions, assistant proposals — is committed to that branch, and the switcher at the top reminds you where you are.
Working branches do not count towards your plan’s limit: the plan is measured in production semantic views. Trying an idea should not cost money.
Promoting
Promote opens a pull request from your branch into production. It does not merge: review is in the path on purpose. Changing a metric’s definition changes what the company understands a number to be, and that deserves someone reading it.
The pull request shows exactly what changes: the SQL, the meaning, and who had validated it. Once merged, qontexta hears about it through a webhook, reindexes production, and the semantic engine starts serving the new commit.
Which version is being served
Each environment shows the commit it is serving. There is no compilation step to wait for: the model is the branch’s contents, and the engine reloads when the branch moves.
If compilation fails (a YAML file that is not a valid Cube model), the environment says so and keeps serving the previous version.
The outputs
Connect on an environment gives you its four outputs and a token scoped to that tenant, that environment and you. Each environment has its own, so you can point an agent at a working branch to try something and leave production alone.