Skip to content

Bring docs.authkestra.com up to v0.13.0 - #3

Merged
marcjazz merged 2 commits into
mainfrom
docs/v0.13.0-sweep
Sep 25, 2026
Merged

marcjazz merged 2 commits into
mainfrom
docs/v0.13.0-sweep

Conversation

@marcjazz

Copy link
Copy Markdown
Member

Two commits.

1. Mechanical: stale versions and one stale claim

Every install snippet site-wide was pinned to an old authkestra-engine/authkestra version (0.9 or 0.11) — 23 occurrences across 13 files. Bumped to 0.13, the current release.

concepts/architecture.md's persistence-boundary description named TOTP/WebAuthn as the only CredentialStore-backed secrets. Recovery codes (below) are CredentialStore-backed too and were missing from that line.

2. Three new provider pages

Magic link, email/SMS one-time codes, and recovery codes all shipped in v0.13.0 (marcjazz/authkestra#400, #402, #404) and had no docs at all. Written to match the existing totp.md/passkeys.md register — intro, an "Enabling" section with the feature flag, configuration, authenticating — with :::note/:::caution for the same class of easy-to-miss behaviour those two pages already flag.

Every code sample was checked against the actual source, not copied from RFC prose:

  • RFC-011's own "Proposed surface" section names OtpAuthMethod::mint's max_attempts as u8 — the shipped signature is u32. otp.md uses the real type and doesn't repeat the stale draft figure.
  • recovery-codes.md's generate() sample matches RecoveryCodeAuthMethod::generate's argument order and types exactly (identity, gate, count: u8) — the one signature here worth getting precisely right, since it's the only method gated by ReproofRequirement.
  • Recovery codes are framed as a NIST SP 800-63B §5.1.2 look-up secret authenticator throughout, never "TOTP backup codes" — that framing is explicitly wrong per the design record and the page says why.

One thing I caught in review after the pages were drafted: magic-link.md's frontmatter description called it "email/SMS magic-link authentication" — backwards, since magic link is delivery-channel-agnostic by design (RFC-010), and email/SMS-specific delivery is what the other new page covers. The page body already had this right; only the one-line description overclaimed. Reworded to "delivered however you choose".

Verification

  • Every function signature, struct/enum name, and field name in all three pages checked directly against crates/authkestra-engine/src/auth/{magic_link,otp,recovery}.rs in the source repo — not trusted from the pages' own drafting.
  • Sidebar entries added for all three; confirmed with a real pnpm run build that all three routes render (/providers/magic-link/, /providers/otp/, /providers/recovery-codes/), alongside the other 25 existing pages, no build errors.

Companion PRs

  • marcjazz/authkestra#410 — the mdBook and RFC status equivalent of this PR, in the main repo.
  • A landing-page PR (adding these three methods to the marketing site's method grid) is ready but not yet opened — its card links point at the three routes this PR adds, so it should merge after this one deploys, or the links will 404 in the interim.

🤖 Generated with Claude Code

https://claude.ai/code/session_01XaoYHiT9Vb85XhBdVCg9Xu

marcjazz and others added 2 commits September 26, 2026 00:32
Every install snippet site-wide was pinned to an old authkestra version
(0.9 or 0.11) — 23 occurrences across 13 files. Bumped to 0.13, the
current release, the same class of fix already applied to the
top-level READMEs and the mdBook.

architecture.md's persistence-boundary description named TOTP/WebAuthn
as the only CredentialStore-backed secrets; recovery codes (shipped in
v0.13.0) are CredentialStore-backed too and were missing.

Sidebar entries added for three provider pages that don't exist yet
(Magic Link, Email/SMS One-Time Codes, Recovery Codes) — the pages
themselves are the next commit.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XaoYHiT9Vb85XhBdVCg9Xu
Magic link, email/SMS one-time codes, and recovery codes all shipped
in v0.13.0 (marcjazz/authkestra#400, #402, #404) and had no docs.
Written to match the existing totp.md/passkeys.md register: a short
intro, an "Enabling" section with the feature flag, a configuration
section, and an authenticating section, using :::note/:::caution for
the same class of easy-to-miss behaviour those two pages already flag.

Every function signature, struct/enum name, and field name shown in a
code sample was checked against the actual .rs source in
marcjazz/authkestra, not copied from RFC prose — RFC-011's own
"Proposed surface" section names OtpAuthMethod::mint's max_attempts as
u8, but the shipped signature is u32; the otp.md page uses the real
type. recovery-codes.md's generate() sample matches
RecoveryCodeAuthMethod::generate's argument order and types exactly
(identity, gate, count: u8), which is the one signature on this page
worth getting precisely right — it is the only method here gated by
ReproofRequirement.

One thing caught in review after the pages were drafted: magic-link.md's
frontmatter description called it "email/SMS magic-link authentication",
which is backwards — magic link is delivery-channel-agnostic by design
(RFC-010), and email/SMS-specific delivery is what the *other* new page
covers. Reworded to "delivered however you choose". The page body
already got this right; only the one-line description overclaimed.

Sidebar entries for these three (added in the previous commit) verified
against a real `pnpm run build`: all three routes
(/providers/magic-link/, /providers/otp/, /providers/recovery-codes/)
render, alongside the other 25 existing pages, no build errors.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XaoYHiT9Vb85XhBdVCg9Xu
@marcjazz
marcjazz merged commit 388bf1e into main Sep 25, 2026
1 check passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant