Skip to content

v2.0.0

Choose a tag to compare

@github-actions github-actions released this 17 Sep 06:24

Added

  • Coverage: break a process or a project down into steps with an AI draft and edit the result in place (HRP-754, HRP-755, HRP-756, HRP-757, HRP-775, HRP-802)
  • Coverage: per-step verdict - an agent type, a person on the team, or a gap - with the automation mode, three buckets and weighted shares once hours per run and runs per year are set (HRP-758, HRP-760, HRP-762)
  • Coverage: the tab opens on two plaques - what share of the work moves to an agent and how well an agent does it, and what the work costs in hours a year and in money once an hourly rate is set on the company profile (HRP-795, HRP-796, HRP-797)
  • Coverage: a To do tab collects what the breakdown leaves unclosed - the steps nobody covers yet, ready to hand to hiring, and the steps an agent type could take where no agent is registered and no skill file is written (HRP-800)
  • Coverage: the first breakdown applies itself, a regeneration shows the current and the proposed steps side by side before you decide, and a long run shows its phase with a running clock (HRP-798)
  • Coverage: capability chips carry the model's confidence and the fragment of the description behind them; doubtful chips are suggestions until confirmed, and a step can be reclassified by a comment (HRP-775, HRP-776, HRP-777)
  • Coverage: gaps open a hire need - a draft vacancy in Recruitment with the steps as its description - for the checked steps or for a single step from its row (HRP-759, HRP-801)
  • Coverage: assign the person who does a step and the person who checks it; an assignment closes the gap even without matching skills (HRP-809)
  • Coverage: process visibility for the whole company or for chosen roles, positions and employees, edit rights for the owner, and an access history (HRP-810)
  • Coverage: a SKILL.md for any step an agent can perform, built from the agent pack's skeleton and the company's competence indicators, generated in the background with preview and download (HRP-760, HRP-761, HRP-799)
  • Capability catalog of seventeen primitives, AI mapping of competences to capabilities, and an AI workforce registry API for agents, packs, assignments and workflows (HRP-748, HRP-749, HRP-750, HRP-751, HRP-752, HRP-753)
  • Imported employees receive an email with a link to set their password, and an admin can resend it from the employee card until they set a password (HRP-806)

Fixed

  • Employee import works again and accepts the CSV template from the import page, including semicolon-separated files with DD.MM.YYYY dates saved by Excel (HRP-806)
  • Imported accounts no longer share a default password (HRP-806)
  • Password reset links work only once (HRP-806)
  • Forgot password works for an email address with accounts in several organizations: one link sets the password for all of them (HRP-815)
  • Imported employees whose set-password email could not be sent are counted on the import job (HRP-806)
  • Bulk invitation requests keep processing the batch when one address fails unexpectedly, instead of answering with an error (HRP-833)
  • An answer scale can be made the default again, and a recruitment scale configuration can be activated, without the save failing (HRP-811)
  • Interview question generation no longer fails when the model returns a question without an explanation
  • Dialogs keep the width they were designed for instead of collapsing to a narrow default (HRP-746)
  • Re-applying a Coverage draft asks for confirmation before it discards generated skills or edited steps
  • Long AI jobs are no longer started again after a broker timeout
  • Upgrading to this release no longer fails on duplicate agent names, and it keeps duplicate pending AI assignments instead of deleting them
  • Coverage: an empty breakdown cannot be accepted, and a step is handed over to hiring only once (HRP-746)
  • Coverage: reordering steps keeps edits made in the meantime, switching between breakdowns no longer shows the previous one, and the hours fields refuse out-of-range values before saving (HRP-746)
  • Coverage: generating several skills at once reports how many failed, and a ready skill with no content is not regenerated automatically (HRP-746)
  • Coverage: the access dialog names a person by email when their card has no name, searches the owner and rule pickers separately, and offers only active employees as owners (HRP-810)
  • File uploads answer with an error when the storage refuses the object instead of recording a file without a URL, and a file record stays when the storage refuses to delete the object (HRP-746)

Security

  • Competence indicators are scoped to the workspace that created them
  • WebSocket connections are refused for tokens revoked by a password change or by offboarding
  • The email delivery webhook verifies the provider's signature
  • Uploads are rejected by size before they are read, and the file category in the storage key is validated
  • Login and the other rate limits are enforced across all application instances
  • External evaluator pages are rate-limited and their free-text fields are bounded
  • Test-only seed endpoints stay disabled on deployed tiers
  • Downloading a company logo by URL pins the address it resolved to
  • /health no longer echoes connection details
  • The self-hosted stack refuses to start with the default JWT secret
  • Email verification and invitation acceptance are rate-limited
  • Email subjects are sanitized for every sender
  • Browser cookies carry the Secure flag on HTTPS
  • Feedback rate limiting fails closed when Redis is unavailable
  • Swagger UI is served only on self-hosted and debug installations; the OpenAPI schema stays available

Changed

  • Public API keys record their last use at most once every five minutes (HRP-746)