ConceptsPrompts & Versions

Prompts & Versions

A prompt in VersionR is a named, versioned LLM instruction. Every edit creates a new version — nothing is ever overwritten.


Anatomy of a prompt

Each version stores:

FieldDescription
ContentThe raw prompt text, with {{variable}} slots
Rolesystem, user, or assistant
ModelThe target model (e.g. gpt-4o, claude-opus-4-6)
TemperatureSampling temperature (0–2)
Max tokensResponse length cap (null = model default)
VariablesAuto-detected list of {{variable_name}} slots
Change noteOptional description of what changed in this version

The model and temperature are versioned alongside the prompt text. This is intentional — a temperature change can affect quality as much as a wording change, and both should be reviewable together.


Version lifecycle

Draft → Review → Staged → Production
StatusMeaning
DraftBeing edited in the dashboard
ReviewSubmitted for team approval (optional for solo devs)
StagedDeployed to the staging environment
ProductionLive to real users

You can skip Review and deploy directly from Draft if your project doesn’t require approval.


Deploying a version

To make a version live, deploy it to an environment:

  1. Open the prompt in the dashboard
  2. Select the version you want to promote
  3. Click Deploy and choose an environment

The environment’s active version pointer updates immediately. Any in-flight SDK requests complete against the previous version; new requests pick up the new one.


Rollback

If a deployed version causes problems, roll back to any prior version in one click:

  1. Open the Versions tab
  2. Find the version you want to restore
  3. Click Rollback

The rollback itself creates an audit trail — the dashboard shows who deployed and who rolled back, and when.


Slugs

Every prompt has a slug — a short, URL-safe identifier you use in SDK calls:

const prompt = await pd.get('interviewer-system', { env: 'production' })
//                          ^^^^^^^^^^^^^^^^^^^
//                          This is the slug

Slugs are set when you create a prompt and cannot be changed. Choose something descriptive and stable — it’s the identifier your code depends on.

Slug format: lowercase letters, numbers, and hyphens. Example: onboarding-welcome, support-triage-v2.


MIT 2026 © Nextra.