feat: server-side A/B experiment assignment (Phase 1)#699
Merged
Conversation
m2broth
reviewed
Jun 11, 2026
| // so neither server-router nor ClientRouter merges an empty `experiments` into every | ||
| // app's appProps, and nothing extra is inlined into the page. | ||
| if (!experimentsEnabled()) { | ||
| return; |
Contributor
There was a problem hiding this comment.
Shouldn't we expire stale cookies when experiments are disabled? eg. users who already have x-ab-* cookies keep readable variant cookies for up to 90 days. Clients can continue seeing/applying a variant even while the kill-switch is expected to put everyone in control.
stas-nc
reviewed
Jun 12, 2026
| protocol: 'ILC_CLIENT_PROTOCOL', | ||
| }, | ||
| experiments: { | ||
| enabled: 'EXPERIMENTS_ENABLED', |
Member
There was a problem hiding this comment.
prefix with ILC_? Not sure, but it looks like half of the variables follow this convention.
| const { experiments } = this.#ilcState; | ||
| const experimentsProps = experiments ? { appProps: { experiments } } : {}; | ||
|
|
||
| return deepmerge.all([appConfigProps, routeProps, experimentsProps]); |
Contributor
There was a problem hiding this comment.
a bit expensive operation. I do not see alternative now but it could be a bottleneck potentially
| * | ||
| * @param options.secure emit cookies with `Secure` (set when the site is https). | ||
| */ | ||
| export function assignExperiments( |
Contributor
There was a problem hiding this comment.
Do you use function by purpose ? I could be complex to scale functions maybe it make sense to create responsible class ?
Contributor
Author
There was a problem hiding this comment.
since it's a pure resolver without any internal mutable state to encapsulate and safe to call once per request with no lifecycle. constructor would add additional step without owning any state
| import { createHash } from 'node:crypto'; | ||
| import type { ExperimentId, ExperimentVariant, VariantName } from './interfaces'; | ||
|
|
||
| const BUCKET_COUNT = 100; |
Contributor
There was a problem hiding this comment.
Does it make sense to put it into configuration ?
Contributor
Author
There was a problem hiding this comment.
I don't think to be honest at the moment it's valuable enough, it's actually should be considered in percentage mass - that's why it's like not any number is fit here
| // ILC mints its own session id (gateway-agnostic, ILC-public like `ilc-i18n`), so the | ||
| // feature works in OSS ILC without depending on any deployment's identity headers. | ||
| /** Stable per-visitor session id minted by ILC; seeds deterministic bucketing. */ | ||
| export const SESSION_COOKIE = 'ilc-sid'; |
Contributor
There was a problem hiding this comment.
Who is responsible for establish the cookie ?
Contributor
Author
There was a problem hiding this comment.
added additional comments regarding this
wRLSS
changed the title
- kill-switch now expires stale x-ab-* cookies so disabling puts mid-experiment visitors back on control instead of leaving a readable variant cookie for up to 90 days - rename env var EXPERIMENTS_ENABLED -> ILC_EXPERIMENTS_ENABLED to match the ILC_ convention used by the other ILC-specific settings - clarify in cookies.ts that it only defines cookie names/options; index.ts is what actually writes them - fill docs/ab-testing.md: a Phase 1 guide for product (what you can/can't do) and developers (defining/reading experiments)
m2broth
approved these changes
Jul 7, 2026
Coverage ReportIlc/serverCommit SHA:61d34c18a7cc506825097d72a844f3ca81bc5726 Test coverage results 🧪File details
Ilc/clientCommit SHA:61d34c18a7cc506825097d72a844f3ca81bc5726 Test coverage results 🧪File details
RegistryCommit SHA:61d34c18a7cc506825097d72a844f3ca81bc5726 Test coverage results 🧪File details
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Introduces server-side A/B experiment assignment in ILC (Phase 1).
ILC resolves a single experiment variant per visitor in the
onRequesthook, before any fragment renders, and propagates it to every fragment viaappProps.experimentsand to the browser via inlinedilcState+ sticky cookies. Assignment is deterministic (hash-based), cookie-sticky, config-driven, and fail-safe — any failure falls back to the baseline variant. A global kill-switch turns the whole layer off without a deploy.See
docs/ab-testing.mdfor the full design: what Phase 1 supports (and what it intentionally does not do yet), how to define experiments in configuration, and how apps read the assigned variant.The consuming side — a
useExperimenthook plus genericappPropsdelivery — ships as a companion change in the React application library.