fix(cli,studio): adopt the CLI's canary decisions in a launched Studio

Closes both cross-surface findings with one mechanism. The CLI publishes
window.__HF_CLI_CANARY_DECISIONS ({ name: boolean }); a CLI-launched
Studio takes it as authoritative over its own seed, URL override and the
registry percentage.

Studio re-deriving could not agree with the CLI in three cases:

  - Telemetry off. The CLI resolves telemetry_opt_out, but Studio's
    opt-out is a separate localStorage flag it cannot see, so it would
    evaluate normally and could enrol on a render the CLI excluded. The
    previous commit gated each surface independently; that fixed silent
    enrolment per surface but NOT the disagreement between them.
  - HF_CANARY_* override. Env vars never cross into the browser — Studio
    reads only its URL param / sessionStorage — so a support session
    forcing a canary on got the CLI forced and Studio guessing.
  - No seed injected. Studio falls back to a different unit id, i.e. a
    different bucket.

Shipping the decision instead of the inputs makes divergence structurally
impossible: one evaluation, two surfaces. It also exposes strictly less —
booleans about features, rather than the seed buckets derive from — which
is why it is safe to publish with telemetry off, the case it exists for.
Studio still evaluates locally when standalone, or for a canary the CLI
did not publish, and ignores a non-boolean value rather than trusting it.

Tests: 6 Studio (CLI-off wins over unset local flag, CLI-on with no URL
param, beats contradicting override, beats seed, falls back per-canary,
rejects non-boolean) and 4 CLI (decisions with telemetry off and no
identity, alongside identity when on, script-tag escaping on a hostile
canary name, throwing resolver degrades to identity only). Four existing
identity tests asserted the old "nothing when telemetry off" contract and
were updated; the registry is now mocked there so string assertions don't
move when a canary is added or ramped. Fault injection: dropping the
adoption fails 4.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
This commit is contained in:
Vance Ingalls
2026-07-30 16:29:31 -07:00
co-authored by Claude Opus 5
parent 4f464dc424
commit 98b23a8850
6 changed files with 271 additions and 34 deletions
+16 -2
View File
@@ -206,8 +206,22 @@ once and no two rollouts could be read apart.
**Cohorts are keyed to the install directory, not to `config.json`.** The
bucketing unit is a dedicated seed in `install-state.json`, inherited across
`config.json` re-mints (see the calibration section) — distinct from the
telemetry id, never emitted, and shared with a CLI-launched Studio so both
surfaces agree. It does not outlive `~/.hyperframes`.
telemetry id and never emitted. It does not outlive `~/.hyperframes`.
**A CLI-launched Studio adopts the CLI's decisions rather than re-deriving
them.** The CLI publishes `window.__HF_CLI_CANARY_DECISIONS` — a plain
`{ name: boolean }` map — and Studio takes it as authoritative over its own
seed, URL override and the registry percentage. Re-deriving cannot agree in
the cases that matter: telemetry off (the CLI resolves `telemetry_opt_out`,
but Studio's opt-out is a separate localStorage flag it cannot see), an
`HF_CANARY_*` override (env vars never cross into the browser), or no seed
injected (Studio falls back to a different unit, so a different bucket). One
render spanning both surfaces must not run half-enrolled.
The decisions map is published even when telemetry is off — that is the case
it exists for. It is safe to expose where the seed is not: booleans about
features, not the value cohorts are derived from. Studio still evaluates
locally when opened standalone, or for any canary the CLI did not publish.
## Removing your canary state
@@ -6,6 +6,9 @@ import { describe, expect, it, vi, beforeEach } from "vitest";
const shouldTrack = vi.fn();
const readConfig = vi.fn();
// Pinned rather than using the real registry, so these string assertions
// don't move every time a canary is added, ramped, or retired.
const canaryDecisions = vi.fn<() => Record<string, boolean>>();
vi.mock("../telemetry/client.js", () => ({
shouldTrack: (...args: unknown[]) => shouldTrack(...args),
@@ -13,6 +16,9 @@ vi.mock("../telemetry/client.js", () => ({
vi.mock("../telemetry/config.js", () => ({
readConfig: (...args: unknown[]) => readConfig(...args),
}));
vi.mock("../telemetry/canary.js", () => ({
canaryDecisionsForStudio: () => canaryDecisions(),
}));
const { resolveCliTelemetryDistinctId, buildCliIdentityScript, buildStudioHeadScripts } =
await import("./telemetryIdentity.js");
@@ -21,6 +27,8 @@ describe("resolveCliTelemetryDistinctId", () => {
beforeEach(() => {
shouldTrack.mockReset();
readConfig.mockReset();
canaryDecisions.mockReset();
canaryDecisions.mockReturnValue({});
});
it("returns the CLI anonymousId when telemetry is enabled", () => {
@@ -56,6 +64,8 @@ describe("buildCliIdentityScript", () => {
beforeEach(() => {
shouldTrack.mockReset();
readConfig.mockReset();
canaryDecisions.mockReset();
canaryDecisions.mockReturnValue({});
});
it("emits a script that sets window.__HF_CLI_DISTINCT_ID when telemetry is on", () => {
@@ -74,11 +84,56 @@ describe("buildCliIdentityScript", () => {
);
});
it("emits an empty string when telemetry is disabled (nothing to seed)", () => {
it("emits an empty string when telemetry is off and there are no canaries", () => {
shouldTrack.mockReturnValue(false);
expect(buildCliIdentityScript()).toBe("");
});
// The cross-surface fix: with telemetry off the CLI resolves every canary
// to telemetry_opt_out, and Studio cannot see that from its own separate
// localStorage flag. Publishing the DECISIONS (not the identity) is what
// stops Studio evaluating independently and enrolling anyway.
it("still publishes canary decisions when telemetry is off, but no identity", () => {
shouldTrack.mockReturnValue(false);
canaryDecisions.mockReturnValue({ "de-parallel-router": false });
const script = buildCliIdentityScript();
expect(script).toBe(
'<script>window.__HF_CLI_CANARY_DECISIONS={"de-parallel-router":false};</script>',
);
expect(script).not.toContain("__HF_CLI_DISTINCT_ID");
expect(script).not.toContain("__HF_CLI_BUCKET_SEED");
});
it("publishes decisions alongside the identity when telemetry is on", () => {
shouldTrack.mockReturnValue(true);
readConfig.mockReturnValue({ anonymousId: "machine-uuid", bucketSeed: "seed-uuid" });
canaryDecisions.mockReturnValue({ "de-parallel-router": true });
expect(buildCliIdentityScript()).toBe(
'<script>window.__HF_CLI_DISTINCT_ID="machine-uuid";' +
'window.__HF_CLI_BUCKET_SEED="seed-uuid";' +
'window.__HF_CLI_CANARY_DECISIONS={"de-parallel-router":true};</script>',
);
});
it("escapes a canary name that tries to close the script tag", () => {
shouldTrack.mockReturnValue(false);
canaryDecisions.mockReturnValue({ "</script><script>alert(1)": true });
const script = buildCliIdentityScript();
expect(script).not.toContain("</script><script>alert(1)");
expect(script).toContain("__HF_CLI_CANARY_DECISIONS");
});
it("survives a throwing canary resolver — telemetry must never break preview", () => {
shouldTrack.mockReturnValue(true);
readConfig.mockReturnValue({ anonymousId: "machine-uuid" });
canaryDecisions.mockImplementation(() => {
throw new Error("registry blew up");
});
expect(buildCliIdentityScript()).toBe(
'<script>window.__HF_CLI_DISTINCT_ID="machine-uuid";</script>',
);
});
it("JSON-encodes the id so it can't break out of the script literal", () => {
shouldTrack.mockReturnValue(true);
readConfig.mockReturnValue({ anonymousId: "</script><script>alert(1)" });
@@ -93,6 +148,8 @@ describe("buildStudioHeadScripts", () => {
beforeEach(() => {
shouldTrack.mockReset();
readConfig.mockReset();
canaryDecisions.mockReset();
canaryDecisions.mockReturnValue({});
});
const ENV_SCRIPT = "<script>window.__HF_STUDIO_ENV__={};</script>";
@@ -105,7 +162,7 @@ describe("buildStudioHeadScripts", () => {
expect(head.indexOf("__HF_CLI_DISTINCT_ID")).toBeLessThan(head.indexOf("__HF_STUDIO_ENV__"));
});
it("returns just the env script when identity is suppressed (telemetry off)", () => {
it("returns just the env script when there is no identity and no canary", () => {
shouldTrack.mockReturnValue(false);
expect(buildStudioHeadScripts(ENV_SCRIPT)).toBe(ENV_SCRIPT);
});
+56 -10
View File
@@ -18,6 +18,7 @@
import { readConfig } from "../telemetry/config.js";
import { shouldTrack as telemetryShouldTrack } from "../telemetry/client.js";
import { canaryDecisionsForStudio } from "../telemetry/canary.js";
/**
* The CLI's anonymous distinct id to hand to Studio, or null when CLI telemetry
@@ -34,13 +35,6 @@ export function resolveCliTelemetryDistinctId(): string | null {
}
}
/**
* `<script>` tag to inject into the served index.html `<head>`, publishing the
* CLI distinct id as `window.__HF_CLI_DISTINCT_ID` before the studio bundle
* runs. Preferred over a URL param so the id never leaks into `$current_url` /
* `url_hash` telemetry or browser history. Empty string when there's nothing to
* seed (telemetry off / no id).
*/
/**
* The CLI's canary bucket seed to hand to Studio, or null. Injected alongside
* the distinct id so a CLI-launched Studio buckets canaries on the SAME unit
@@ -67,13 +61,65 @@ function encodeInlineScriptValue(value: string): string {
return JSON.stringify(value).replace(/</g, "\\u003c").replace(/\//g, "\\/");
}
/**
* Same escaping, for a structured value. Separate from the string form
* because that one stringifies its argument — passing an object through it
* would double-encode into a quoted JSON blob.
*
* The keys here are registry canary names, so they are developer-authored and
* ASCII by convention rather than user input; the escaping is belt-and-braces
* for the same reason it is on the ids.
*/
function encodeInlineScriptJson(value: unknown): string {
return JSON.stringify(value).replace(/</g, "\\u003c").replace(/\//g, "\\/");
}
/**
* The CLI's resolved canary decisions for a launched Studio, or null when
* there are none to publish. Fail-silent, like everything else here.
*/
function resolveCliCanaryDecisions(): Record<string, boolean> | null {
try {
const decisions = canaryDecisionsForStudio();
return Object.keys(decisions).length > 0 ? decisions : null;
} catch {
return null;
}
}
/**
* `<script>` tag to inject into the served index.html `<head>`, publishing the
* CLI distinct id as `window.__HF_CLI_DISTINCT_ID` (plus the bucket seed and
* resolved canary decisions) before the studio bundle runs. Preferred over a
* URL param so the id never leaks into `$current_url` / `url_hash` telemetry
* or browser history. Empty string only when there is nothing at all to
* publish.
*/
export function buildCliIdentityScript(): string {
const parts: string[] = [];
const cliId = resolveCliTelemetryDistinctId();
if (!cliId) return "";
const parts = [`window.__HF_CLI_DISTINCT_ID=${encodeInlineScriptValue(cliId)};`];
if (cliId) {
parts.push(`window.__HF_CLI_DISTINCT_ID=${encodeInlineScriptValue(cliId)};`);
const seed = resolveCliBucketSeed();
if (seed) parts.push(`window.__HF_CLI_BUCKET_SEED=${encodeInlineScriptValue(seed)};`);
return `<script>${parts.join("")}</script>`;
}
// Emitted even when telemetry is OFF and the identity block above is empty —
// that is the case it exists for. With telemetry off the CLI resolves every
// canary to `telemetry_opt_out`, but Studio's opt-out is a separate
// localStorage flag it cannot see, so left to itself Studio would evaluate
// normally and could enrol on a render the CLI had already excluded. Same
// for an `HF_CANARY_*` override, which never crosses into the browser.
//
// Safe to publish unconditionally: these are booleans about features, not
// identity, and strictly less than the seed they replace as Studio's input.
const decisions = resolveCliCanaryDecisions();
if (decisions !== null) {
parts.push(`window.__HF_CLI_CANARY_DECISIONS=${encodeInlineScriptJson(decisions)};`);
}
return parts.length === 0 ? "" : `<script>${parts.join("")}</script>`;
}
/**
+26
View File
@@ -118,6 +118,32 @@ export function isCanaryEnabled(name: string): boolean {
return resolveCanary(name).enabled;
}
/**
* Every registered canary's resolved on/off for this process, for handing to
* a CLI-launched Studio.
*
* Studio adopts these wholesale instead of re-deriving, because re-deriving
* cannot agree in three cases:
*
* - **Telemetry off.** The CLI resolves `telemetry_opt_out`; Studio's own
* opt-out is a separate localStorage flag on a different machine-level
* switch, so it would evaluate normally and could enrol.
* - **`HF_CANARY_*` override.** Env vars do not cross into the browser at
* all — Studio only reads its URL param / sessionStorage — so a support
* session forcing a canary on got the CLI forced and Studio guessing.
* - **No seed injected.** Whenever the seed is withheld Studio falls back
* to a different unit id, i.e. a different bucket.
*
* Shipping the decision rather than the inputs makes divergence structurally
* impossible: one evaluation, two surfaces. It is also strictly less to
* expose — booleans about features, not the seed the buckets derive from.
*/
export function canaryDecisionsForStudio(): Record<string, boolean> {
const out: Record<string, boolean> = {};
for (const canary of CANARIES) out[canary.name] = resolveCanary(canary.name).enabled;
return out;
}
/**
* Canary assignments as PostHog flag properties — `$feature/canary-<name>`
* set to `"true"` / `"false"` for every registered canary. Spread onto every
@@ -225,3 +225,53 @@ describe("telemetry opt-out is canary opt-out", () => {
});
});
});
describe("CLI-launched Studio adopts the CLI's decisions", () => {
const OPT_OUT_KEY = "hyperframes-studio:telemetryDisabled";
afterEach(() => {
delete window.__HF_CLI_CANARY_DECISIONS;
});
// The divergence this exists for: CLI telemetry off resolves every canary
// to telemetry_opt_out, but Studio's opt-out is a SEPARATE localStorage
// flag it cannot see — left to itself it would evaluate and could enrol.
it("stays off when the CLI opted out, even though Studio's own flag is unset", () => {
expect(localStorage.getItem(OPT_OUT_KEY)).toBeNull();
window.__HF_CLI_CANARY_DECISIONS = { "on-everywhere": false };
expect(resolveCanary("on-everywhere").enabled).toBe(false);
});
// HF_CANARY_* never crosses into the browser, so before this the CLI was
// forced on and Studio silently guessed from the percentage.
it("turns on when the CLI forced it on, with no URL param present", () => {
window.__HF_CLI_CANARY_DECISIONS = { "off-everywhere": true };
expect(resolveCanary("off-everywhere").enabled).toBe(true);
});
it("beats a contradicting URL override — one render must not run half-enrolled", () => {
window.__HF_CLI_CANARY_DECISIONS = { "on-everywhere": false };
setSearch("?hf_canary_on_everywhere=on");
expect(resolveCanary("on-everywhere").enabled).toBe(false);
});
it("beats the seed-derived bucket", () => {
window.__HF_CLI_BUCKET_SEED = "5f1c9d2e-0000-4000-8000-aaaaaaaaaaaa";
window.__HF_CLI_CANARY_DECISIONS = { "on-everywhere": false };
expect(resolveCanary("on-everywhere").enabled).toBe(false);
});
it("falls back to local evaluation for a canary the CLI did not publish", () => {
window.__HF_CLI_CANARY_DECISIONS = { "off-everywhere": true };
expect(resolveCanary("on-everywhere").enabled).toBe(true);
});
it("ignores a non-boolean value rather than trusting it", () => {
window.__HF_CLI_CANARY_DECISIONS = { "on-everywhere": "false" } as unknown as Record<
string,
boolean
>;
// Falls through to local evaluation: on-everywhere is at 100%.
expect(resolveCanary("on-everywhere").enabled).toBe(true);
});
});
+62 -18
View File
@@ -53,6 +53,34 @@ export function canaryParamName(name: string): string {
declare global {
interface Window {
__HF_CLI_BUCKET_SEED?: string;
/** Resolved on/off per canary from the launching CLI. Authoritative. */
__HF_CLI_CANARY_DECISIONS?: Record<string, boolean>;
}
}
/**
* The launching CLI's decision for this canary, if it published one.
*
* When present this WINS over everything Studio could work out locally —
* seed, URL override, registry percentage. Studio re-deriving cannot agree
* with the CLI in the cases that matter most:
*
* - telemetry off: the CLI resolves `telemetry_opt_out`, but Studio's
* opt-out is a separate localStorage flag it cannot see from here;
* - `HF_CANARY_*` override: env vars never cross into the browser;
* - no seed injected: Studio falls back to a different unit, i.e. a
* different bucket.
*
* In all three the CLI has already decided, and one render spanning both
* surfaces must not run half-enrolled.
*/
function cliDecision(name: string): boolean | undefined {
try {
if (typeof window === "undefined") return undefined;
const decision = window.__HF_CLI_CANARY_DECISIONS?.[name];
return typeof decision === "boolean" ? decision : undefined;
} catch {
return undefined;
}
}
@@ -142,6 +170,39 @@ export function __resetStudioCanaryCacheForTests(): void {
decisions.clear();
}
/** The uncached decision. Split out so `resolveCanary` is purely the memo. */
function decideStudioCanary(name: string): CanaryDecision {
const definition = findCanary(name);
if (!definition) return { enabled: false, reason: "out_of_cohort" };
// The launching CLI's decision, when there is one, is the whole answer:
// it already applied telemetry opt-out, HF_CANARY_* overrides and the
// percentage against the shared seed. Re-deriving here is what let the two
// surfaces disagree on the same render.
const fromCli = cliDecision(definition.name);
if (fromCli !== undefined) {
return { enabled: fromCli, reason: fromCli ? "forced_on" : "forced_off" };
}
const override = readOverride(definition.name);
// Standalone Studio. Opting out of telemetry opts you out of canaries —
// same rule as the CLI (see packages/cli/src/telemetry/canary.ts). A
// profile that sends nothing can't be compared against anyone, so
// enrolling it changes that user's code path for no signal. An explicit
// override still wins: a deliberate local choice, not silent enrolment.
if (override === undefined && isOptedOut()) {
return { enabled: false, reason: "telemetry_opt_out" };
}
return evaluateCanary({
feature: definition.name,
unitId: resolveBucketUnit(),
percentage: definition.percentage,
override,
exclude: isAutomatedBrowser(),
});
}
/**
* Full decision including the reason. An unregistered name resolves to off
* rather than throwing — a typo in a rollout control must never break the
@@ -151,24 +212,7 @@ export function resolveCanary(name: string): CanaryDecision {
const cached = decisions.get(name);
if (cached) return cached;
const definition = findCanary(name);
const override = definition ? readOverride(definition.name) : undefined;
// Opting out of telemetry opts you out of canaries — same rule as the CLI
// (see packages/cli/src/telemetry/canary.ts). An install that sends nothing
// can't be compared against anyone, so enrolling it changes that user's
// code path for no signal. An explicit override still wins: that is a
// deliberate local choice, not silent enrolment.
const decision: CanaryDecision = !definition
? { enabled: false, reason: "out_of_cohort" }
: override === undefined && isOptedOut()
? { enabled: false, reason: "telemetry_opt_out" }
: evaluateCanary({
feature: definition.name,
unitId: resolveBucketUnit(),
percentage: definition.percentage,
override,
exclude: isAutomatedBrowser(),
});
const decision = decideStudioCanary(name);
decisions.set(name, decision);
return decision;