v0.122.10

## Highlights

- **Clearer run failures.** When an agent run fails in the playground, the failure now reads like an agent answer. The full server message is shown, and the ways to recover (add a key, buy credits, change the plan) stay next to the run.
- **Fewer stuck question forms.** A question form with a bad schema no longer leaves a stuck card in the composer. The agent's lead-in text no longer shows as the answer while the run waits for you.
- **No false "add your key" prompt.** Workspaces that run on credits no longer see the add-key strip, and the strip no longer flashes on reload.
- **New link preview card.** Links to the website, the docs and the app now show the same preview image.
- **Sign-in rate limit.** The API now limits sign-in code and sign-up requests per client IP. Behind a proxy or load balancer, set `AGENTA_ACCESS_TRUSTED_PROXY_HOPS` to the number of proxies in front of the API, so the API reads the real client IP. `AGENTA_ACCESS_TRUST_CLOUDFLARE_IP` is off by default. `AGENTA_AUTH_CODE_RATE_LIMIT_ENABLED` turns the limit off.
- **Sign-up blocklist cache.** `AGENTA_ACCESS_BLOCKLIST_CACHE_TTL` sets how long the API keeps the blocked-domains list. The default is 60 seconds.
- **Abuse checks for free credits (Enterprise).** The Enterprise edition can check new sign-ups and new API keys for credit farming with TypeSafe Jev. It runs only when `TYPESAFE_API_KEY` is set, and it allows the request when the check service does not answer. When the work-domain sign-up cap refuses a free grant, the wallet floor is raised so the account can't spend free credits.
- **Branded emails held back.** The branded sign-in, invitation and password reset emails (#7668) were reverted before this release (#7720). Sign-in codes are still sent by SuperTokens, as in v0.122.9.

## What's Changed

* [fix] Hide unrenderable question forms and keep lead-in text in the fold by @ashrafchowdury in https://github.com/Agenta-AI/agenta/pull/7625
* [fix] Hide the add-key strip while loading or when credits fund runs by @ashrafchowdury in https://github.com/Agenta-AI/agenta/pull/7699
* [feat] Show run failures as an agent message with in-place recovery by @ashrafchowdury in https://github.com/Agenta-AI/agenta/pull/7684
* [feat] Use the workspace card for website, docs and app link previews by @ashrafchowdury in https://github.com/Agenta-AI/agenta/pull/7685
* fix(api): make the sign-up blocklist flag cache configurable, default 60s by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/7706
* [feat] Check signups and API key creation for credit farming with Jev by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/7707
* [fix] Rate-limit sign-in code and sign-up requests per IP in the API by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/7710
* fix(api): raise the wallet floor when the work-domain signup cap refuses by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/7709
* fix(api): read the fraud check policy and thresholds from the flag payload by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/7714
* [feat] Brand the sign-in, invitation and password reset emails by @ashrafchowdury in https://github.com/Agenta-AI/agenta/pull/7668 (reverted by #7720, not in this release)
* [revert] Keep sign-in code emails on SuperTokens for v0.122.10 (reverts #7668) by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/7720

**Full Changelog**: https://github.com/Agenta-AI/agenta/compare/v0.122.9...v0.122.10