fix(web): regenerate facts.generated.ts for the two v0.9.8 providers

[Lstarsky0]

check:facts has been failing on main since v0.9.8, and it sits inside Lint & Type Check — a required check, so every web pull request inherits the red no matter what it touches. My #5397 is currently red on exactly this and nothing else.

53af08ce3 registered Google Gemini and 531b64ed9 registered Google Antigravity; web/lib/facts.generated.ts was never regenerated. Same pair that left the CLI assertion counts behind in #5383, and the same family as the docs/PROVIDERS.md row #5394 refreshed — this looks like the remaining artifact from those two commits.

This is the output of cd web && npm run prebuild, which is what the checker prints as the remedy. derive-facts.mjs reads the Rust sources directly and needs no build.

[derive-facts] version=0.9.8 crates=21 providers=44 sandboxes=2 default-model=deepseek-v4-pro node=>=18 tools=74 license=MIT

The diff is the two provider rows plus generatedAt. The script deliberately preserves that stamp when no checked fact has moved, so it changing here is the signal that one did.

Checks

Branched from ad102ec44. Gates run twice, identical both rounds:

Gate	exit
npm run check:facts	0 — was 1 on the unmodified head
npm run check:locales	0
npm run check:docs	0
npm run lint	0

npx vitest run stays at 10 failed / 250 passed, and the failing set is byte-identical to the same suite on an unmodified ad102ec44 — seven Cloudflare deploy-preflight tests that need credentials, plus one public-copy and two public-surface-contract parity tests. This regeneration does not address those three; they look like the changelog/release-facts side of the same v0.9.8 drift, and I did not want to bundle a guess with a mechanical regeneration.

Analysis drafted with local tooling; every change and test verified by hand.

No-Issue: unbreaks the check:facts gate on main; the drift itself is already covered by the v0.9.8 follow-ups.

[github-actions]

Thanks @Lstarsky0 for taking the time to contribute.

This repository is observing a maintainer-managed PR intake gate in dry-run mode, so this pull request is staying open. This note helps maintainers prepare the allowlist before any enforcement is considered.

Please read CONTRIBUTING.md for the expected contribution shape. A maintainer can grant recurring PR access by commenting /lgtm on a pull request.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: f2825bf9e5

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Exclude credential-only Antigravity from provider routes

When these facts render on the models page, every entry is presented and counted as a supported provider route, but crates/tui/src/client.rs rejects every Antigravity request because its wire protocol is unimplemented; adding this row therefore advertises a route that users cannot run and inflates the provider count across the homepage, FAQ, and facts API. Exclude Antigravity from the source derivation (or model credential-only entries separately) and regenerate this file rather than publishing it as a normal provider.

AGENTS.md reference: web/AGENTS.md:L5-L8

Useful? React with 👍 / 👎.

[Lstarsky0]

This clears check:facts but the gate stays red, and the remaining three are worth listing together because they're the same v0.9.8 drift and one of them isn't mine to write.

Lint & Type Check runs check:facts then vitest. With facts regenerated it gets past the first step and fails on:

missing 0.9.8 contributor ledger: expected undefined to be truthy
expected '0.9.7' to be '0.9.8'
expected '# Installing Codewhale…' to contain 'v0.9.8 source candidate'

Traced to three stale artifacts:

Where	Now	Should be
docs/public-surface-facts.json → sourceCandidate.version	0.9.7	0.9.8
docs/public-surface-facts.json → sourceCandidate.providerCount	42	44
docs/INSTALL.md:11	**v0.9.7 source candidate**	v0.9.8
CHANGELOG.md 0.9.8 entry	no ### Contributors section	one is needed

One thing that took checking and is easy to get backwards: latestPublishedRelease should stay at v0.9.7. v0.9.8 is a tag but not a GitHub release — gh release list still shows v0.9.7 as latest. So the matrix's published <= sourceCandidate invariant is satisfied by moving only the candidate side; touching the published side would make the site advertise a release that doesn't exist, which is the exact thing that assertion guards.

The two docs/ rows are mechanical and I can add them here — sourceCandidate derives from Cargo.toml (0.9.8) and the provider registry (44, same Gemini + Antigravity pair as the facts file). Say the word.

The contributor ledger I've deliberately not touched. AGENTS.md puts credit under CI enforcement with canonical identities in .github/AUTHOR_MAP, and I'd be guessing at who belongs in a release I only partly worked on. That one needs you regardless of what I do with the rest.