v0.5.0

Minor Changes

• #177 5e8e96b Thanks @akmalakhpah! - The agents have faces. The OKR Coach is blue, the OKR Champion is green, and the copilot and any custom agent wear a face in the text colour that follows light and dark. A face shows a mood that fits what the agent is saying: a sad Champion on an overdue check-in, a thinking Coach on a quality question, a happy one when every check passes. Faces appear on Agents and runs, the copilot, the inbox, the goal page, the Draft Coach, the alignment panel and the method pages. An idle face blinks and glances on a slow loop; the other moods play on hover and when the mood changes, and loop only while an agent is working. Every face holds still for anyone who asks for reduced motion.

• #165 5438611 Thanks @agungksidik! - The agent setup pill on Home shows the marks of the agents it covers and adds steps for Codex and Cursor. Its close button appears on hover.

• #165 6172032 Thanks @agungksidik! - The agent setup button copies one line. The agent fetches the steps from the instance's new public /agent-setup/prompt.md, which covers Claude Code, Codex, Cursor, GitHub Copilot, OpenCode, Windsurf, Hermes Agent and ChatGPT and holds no token.

• #178 02e97c4 Thanks @agungksidik! - Activity counts unread on the server, so its number no longer stops at the 50 rows a page holds, and the Updates badge says what it adds up ("3 to do, 19 unread"). A check-in group in Activity is named after its objective. A Board card can be deleted, after a confirmation. The first-visit tour outlines its stop with a dashed line that does not read as keyboard focus, scrolls the stop into view, and shows no Back on the first stop.

• #178 0c03fdf Thanks @agungksidik! - The person accountable for a goal is called the Owner by default, where it was the Champion. "Champion" is now the suggested alternative on the terminology card, and a workspace that renamed the term keeps its own word. The OKR Champion agent keeps its name. METHOD.md §0 and §2.5 say Owner, and the quality checks, gates, rule pages and goal exports follow. Goal list exports head the column "Owner".

• #178 ff7b2d5 Thanks @agungksidik! - The setup wizard reads plainly: "Step 1 of 5" with a progress bar, the local time for the chosen zone, a Back button from step 2, "Weekly (recommended)", 44px rhythm buttons, templates as a radio group, and "Add sample data" or "Skip and finish" at the end. The Work Map's three figures explain themselves on hover, and its confidence chip reads "High, 8 in 10" with the bands on hover. Goals and search say objectives where they said goals.

• #168 3dbba8c Thanks @akmalakhpah! - Every setting, profile and comparison row on Admin, Practice has a small "i" that says what it decides, in a sentence or two, with a link to its exact section of the method for the full write-up. It opens on hover, on tap and from the keyboard. The practice read in the API now returns each setting's summary as well.

• #167 570a0a2 Thanks @akmalakhpah! - Admin, Practice has a "Compare profiles" button that shows Recommended, Google-style, Radical Focus, Lightweight and Governed side by side: every setting and threshold one of them changes, with the profile in use marked and each value that differs from Recommended in bold. The page also links to the full method, as the running release shipped it. METHOD.md carries the same table, and the conformance check now holds it to what choosing each profile does.

• #178 c426fe2 Thanks @agungksidik! - An administrator can now choose who may create an account from Admin > General > Registration. The choices are Invitation only, Company email addresses and Anyone. Under Company email addresses, a company address can sign up and start a workspace. A Gmail, Yahoo, Outlook or other free mail address is refused with the reason, unless an invitation admits it. Every change is written to the instance audit chain. OPENOKR_REGISTRATION is now read when nothing has been saved in the application. A free mail provider can no longer be added as a trusted email domain.

• #174 91af1bf Thanks @akmalakhpah! - Every threshold on Admin, Rhythm and thresholds has a small "i". Where its reason cites a principle by number, such as "(principle 5)", the hint says what that principle is. It also names the part of the method the threshold comes from and links to that exact section. Check-in frequency and the anchor day, which show no reason on the card, carry theirs in the hint.

• #175 80354db Thanks @akmalakhpah! - A rule named by its key explains itself where it is named. Resting on a verdict such as "KR-3" on a goal's coach strip, on "Read the rule" in the goals list, or on a nudge's rule on the review inbox and the nudge volume page shows the rule's name, what kind it is, what a check looks for or when a nudge fires and who hears it. A click still opens the rule's page, and a quality check's page now links to its exact section of the method.

• #165 8c0fd5b Thanks @agungksidik! - Administrators see the running version at the foot of every screen, on the right, and a link to the newest GitHub release when the instance is behind it. The check asks GitHub at most once a day, never delays a page and sends nothing about the instance. It is off on a managed cloud, and an administrator turns it off in Admin, General, or the operator with OPENOKR_UPDATE_CHECK=off.

• #165 df256ec Thanks @agungksidik! - Admin, Branding takes a logo and a favicon. A chosen image is previewed in a light and a dark sidebar, or a browser tab, before it is saved. The logo replaces the initials in the sidebar for every member and the favicon is the tab icon on signed-in screens. New action settings.setWorkspaceBrandingImage; the colour save now keeps the images.

Patch Changes

• #180 9eb86cf Thanks @agungksidik! - Every save, create, edit and delete now shows the thin bar across the top of the window while the server works on it, the same bar a sidebar link shows. A fast answer shows nothing; after ten seconds, or offline, a message says the server has not answered yet.

• #178 6f01dbb Thanks @agungksidik! - A new read, notifications.reasonCounts, returns how many unsnoozed notifications a member has in all, unread and per reason. Activity shows the number on each filter pill, counted on the server rather than over the page it lists. It is in the REST, command line and agent surfaces with the other notification reads.

• #176 1248521 Thanks @akmalakhpah! - The cards on the Channels, Nudge volume, AI and Agents admin pages now have space between them instead of touching.

• #171 8f202fc Thanks @dependabot! - The Anthropic provider moves to version 1 of Anthropic's SDK, which uses API version 2026-10-21 and sends the API key as a bearer token. That API version refuses a temperature on every model, so requests to Anthropic models no longer carry one, and the sampling a workspace sets for a tier does not apply to them. Nothing set a temperature on a request before this change, so no request behaves differently. A proxy in front of A...

v0.5.0-rc.1

Minor Changes

• #177 5e8e96b Thanks @akmalakhpah! - The agents have faces. The OKR Coach is blue, the OKR Champion is green, and the copilot and any custom agent wear a face in the text colour that follows light and dark. A face shows a mood that fits what the agent is saying: a sad Champion on an overdue check-in, a thinking Coach on a quality question, a happy one when every check passes. Faces appear on Agents and runs, the copilot, the inbox, the goal page, the Draft Coach, the alignment panel and the method pages. An idle face blinks and glances on a slow loop; the other moods play on hover and when the mood changes, and loop only while an agent is working. Every face holds still for anyone who asks for reduced motion.

• #165 5438611 Thanks @agungksidik! - The agent setup pill on Home shows the marks of the agents it covers and adds steps for Codex and Cursor. Its close button appears on hover.

• #165 6172032 Thanks @agungksidik! - The agent setup button copies one line. The agent fetches the steps from the instance's new public /agent-setup/prompt.md, which covers Claude Code, Codex, Cursor, GitHub Copilot, OpenCode, Windsurf, Hermes Agent and ChatGPT and holds no token.

• #178 02e97c4 Thanks @agungksidik! - Activity counts unread on the server, so its number no longer stops at the 50 rows a page holds, and the Updates badge says what it adds up ("3 to do, 19 unread"). A check-in group in Activity is named after its objective. A Board card can be deleted, after a confirmation. The first-visit tour outlines its stop with a dashed line that does not read as keyboard focus, scrolls the stop into view, and shows no Back on the first stop.

• #178 0c03fdf Thanks @agungksidik! - The person accountable for a goal is called the Owner by default, where it was the Champion. "Champion" is now the suggested alternative on the terminology card, and a workspace that renamed the term keeps its own word. The OKR Champion agent keeps its name. METHOD.md §0 and §2.5 say Owner, and the quality checks, gates, rule pages and goal exports follow. Goal list exports head the column "Owner".

• #178 ff7b2d5 Thanks @agungksidik! - The setup wizard reads plainly: "Step 1 of 5" with a progress bar, the local time for the chosen zone, a Back button from step 2, "Weekly (recommended)", 44px rhythm buttons, templates as a radio group, and "Add sample data" or "Skip and finish" at the end. The Work Map's three figures explain themselves on hover, and its confidence chip reads "High, 8 in 10" with the bands on hover. Goals and search say objectives where they said goals.

• #168 3dbba8c Thanks @akmalakhpah! - Every setting, profile and comparison row on Admin, Practice has a small "i" that says what it decides, in a sentence or two, with a link to its exact section of the method for the full write-up. It opens on hover, on tap and from the keyboard. The practice read in the API now returns each setting's summary as well.

• #167 570a0a2 Thanks @akmalakhpah! - Admin, Practice has a "Compare profiles" button that shows Recommended, Google-style, Radical Focus, Lightweight and Governed side by side: every setting and threshold one of them changes, with the profile in use marked and each value that differs from Recommended in bold. The page also links to the full method, as the running release shipped it. METHOD.md carries the same table, and the conformance check now holds it to what choosing each profile does.

• #178 c426fe2 Thanks @agungksidik! - An administrator can now choose who may create an account from Admin > General > Registration. The choices are Invitation only, Company email addresses and Anyone. Under Company email addresses, a company address can sign up and start a workspace. A Gmail, Yahoo, Outlook or other free mail address is refused with the reason, unless an invitation admits it. Every change is written to the instance audit chain. OPENOKR_REGISTRATION is now read when nothing has been saved in the application. A free mail provider can no longer be added as a trusted email domain.

• #174 91af1bf Thanks @akmalakhpah! - Every threshold on Admin, Rhythm and thresholds has a small "i". Where its reason cites a principle by number, such as "(principle 5)", the hint says what that principle is. It also names the part of the method the threshold comes from and links to that exact section. Check-in frequency and the anchor day, which show no reason on the card, carry theirs in the hint.

• #175 80354db Thanks @akmalakhpah! - A rule named by its key explains itself where it is named. Resting on a verdict such as "KR-3" on a goal's coach strip, on "Read the rule" in the goals list, or on a nudge's rule on the review inbox and the nudge volume page shows the rule's name, what kind it is, what a check looks for or when a nudge fires and who hears it. A click still opens the rule's page, and a quality check's page now links to its exact section of the method.

• #165 8c0fd5b Thanks @agungksidik! - Administrators see the running version at the foot of every screen, on the right, and a link to the newest GitHub release when the instance is behind it. The check asks GitHub at most once a day, never delays a page and sends nothing about the instance. It is off on a managed cloud, and an administrator turns it off in Admin, General, or the operator with OPENOKR_UPDATE_CHECK=off.

• #165 df256ec Thanks @agungksidik! - Admin, Branding takes a logo and a favicon. A chosen image is previewed in a light and a dark sidebar, or a browser tab, before it is saved. The logo replaces the initials in the sidebar for every member and the favicon is the tab icon on signed-in screens. New action settings.setWorkspaceBrandingImage; the colour save now keeps the images.

Patch Changes

• #180 9eb86cf Thanks @agungksidik! - Every save, create, edit and delete now shows the thin bar across the top of the window while the server works on it, the same bar a sidebar link shows. A fast answer shows nothing; after ten seconds, or offline, a message says the server has not answered yet.

• #178 6f01dbb Thanks @agungksidik! - A new read, notifications.reasonCounts, returns how many unsnoozed notifications a member has in all, unread and per reason. Activity shows the number on each filter pill, counted on the server rather than over the page it lists. It is in the REST, command line and agent surfaces with the other notification reads.

• #176 1248521 Thanks @akmalakhpah! - The cards on the Channels, Nudge volume, AI and Agents admin pages now have space between them instead of touching.

• #171 8f202fc Thanks @dependabot! - The Anthropic provider moves to version 1 of Anthropic's SDK, which uses API version 2026-10-21 and sends the API key as a bearer token. That API version refuses a temperature on every model, so requests to Anthropic models no longer carry one, and the sampling a workspace sets for a tier does not apply to them. Nothing set a temperature on a request before this change, so no request behaves differently. A proxy in front of A...

v0.4.0

Minor Changes

• #161 af6432b Thanks @agungksidik! - The review inbox can be cleared.

  Each acknowledgement row shows the check-in's status and the first line of what
  it says, and "Acknowledge all" acknowledges every listed one in one press. A
  check-in nobody acknowledged within 14 days lapses: it leaves Review and the
  acknowledgement ladder stops asking, and it is not recorded as acknowledged.
  The window is the new §11 threshold cadence.acknowledgementWindowDays, which
  a workspace can change in Admin, Rhythm.

• #150 709abed Thanks @agungksidik! - A member can connect their own AI agent by pasting one prompt.

  The Work Map shows a pill that copies a setup prompt for Claude Code, Hermes
  Agent and ChatGPT in developer mode. The prompt names this
  instance's agent endpoint and signs in with OAuth, so it carries no token and
  the member approves the connection in the browser. Closing the pill hides it
  until the next sign-in.

  The API tokens screen copies the same prompt, and the answer that mints an
  agent token copies a version with that token in it. ChatGPT accepts OAuth only,
  and the token version says so.

• #160 99c5680 Thanks @agungksidik! - A new action, blockers.raise, lets a key result's owner (or their agent)
  raise a blocker at check-in time without a session. It takes the same type,
  owner and next action as a session blocker, is due by the goal's next
  check-in, and shows on the space's blocker board for the weekly session to
  discuss. It is in the default core agent toolset, and the weekly check-in
  prompt uses it for a key result that is actually blocked.

• #161 4406071 Thanks @agungksidik! - A champion watching their goal hears when the reviewer acknowledges its
  check-in: the acknowledgement reaches the goal's watchers in the inbox, as a
  published check-in already did. Pressing acknowledge again on the same
  check-in tells nobody.

• #161 76abcf5 Thanks @agungksidik! - A member can check in on any day, not only when the goal is due.

  The check-in page lists the member's other goals under "Update any time". A
  goal that is not due offers "Post an update", which opens the usual composer.
  An update posted more than the cadence tolerance before the due date is an
  extra update: it moves health and progress, and the due date, its nudge and
  the blocker clock stay where they were. Before this, an early check-in used up
  the next period and that Monday's reminder never came.

  The check-in page and the Review inbox now list the same goals, so the Review
  can no longer say a check-in is owed while the check-in page says nothing is
  due. The rhythm share in the review diagnostic counts an extra update as
  answering nothing, the way the cadence does. goals.due accepts up to 366
  days ahead.

• #161 fd00fd3 Thanks @agungksidik! - A blocker can be raised from the check-in page.

  Under the composer, "Something in the way?" takes the key result, the blocker
  type, an owner and the next action, the same fields as the weekly session's
  form. The blocker is due by the goal's next due date and shows on the space's
  blocker board for the next session. Until now only the session room could
  raise one.

• #160 91034ec Thanks @agungksidik! - A new read action, goals.checkDraft, runs the METHOD.md §4 quality checks
  on an objective and its key results that have not been saved, and returns
  each check's status, the condition it matched, its coaching prompt and the
  strength score. It saves nothing and needs no AI provider. It is in the
  default core agent toolset, so an agent drafting with its own model can
  check a draft before goals.create publishes it, and the
  review-against-the-canon prompt checks each rewrite with it.

• #161 aa2b282 Thanks @agungksidik! - The OKRs list opens on your own objectives and reads as a status.

  Scope defaults to Mine, or to All when you have no objective of your own in the
  cycle. The level, health, kind, champion, space and closed filters sit behind
  one Filter button, which opens on its own while any of them is set. The pickers
  under each row show their chevrons and calendar marks only while the row is
  hovered or being edited, so the list no longer looks like an open form. Health
  chips use the same words as everywhere else.

• #161 214cbb3 Thanks @agungksidik! - A key result that reads a KPI now measures progress from its own baseline to
  its own target.

  The KPI's latest reading is still the key result's current value. What changed
  is the percentage: it used to be the KPI's achievement against the KPI's own
  target, so "Raise deflection from 31% to 35%" at a reading of 31 showed 89%. It
  now shows 0%, the same as any other key result at its baseline. Changing a
  KPI's target no longer moves the key results that read it. Goal progress
  recomputes the next time the KPI is recorded or the goal changes.

• #160 2afccd5 Thanks @agungksidik! - Seventeen AI-assist tools leave the agent catalogue, among them
  goals.draftObjective, kpis.suggest and sessions.draftMinutes. The MCP
  server makes no outbound AI calls, so each of them answered null to every
  agent. They stay on REST, the command line and the browser. The full
  catalogue is now 377 tools.

• #160 0467a32 Thanks @agungksidik! - Two new agent toolsets. room holds the live session's own moves (votes and
  their reveal, the mic, pulse, kudos, retro notes, anonymous process-health
  answers), split from facilitation. integrations holds AI providers, chat
  channels, single sign-on and spreadsheet imports, split from admin, which
  was over the 128-tool cap some clients apply. The default core toolset now
  also publishes and approves a drafted objective. An address that named
  facilitation or admin for one of the moved tools adds room or
  integrations.

• #150 7db1b13 Thanks @agungksidik! - The agent endpoint offers a small catalogue by default. A connection sees the
  57 core tools a member uses every week unless its address names more with
  ?toolsets= (core, work, facilitation, admin or all), and it is shown
  only the tools its token's scopes can call. Twenty actions that exist for the
  importer, the scheduler, the copilot's own bookkeeping or the tokens screen are
  no longer agent tools; they stay on /api/v1. An agent that relied on the full
  list should add ?toolsets=all to the server address. A call to a tool outside
  the chosen toolsets is refused with the toolset that holds it.

• #161 73845cc Thanks @agungksidik! - A member sees their own check-in streak on the check-in page.

  It counts the periods in a row in which they posted a check-in on a goal they
  champion, by the same rule as the space's streak. It is paused, and names the
  goal, while one of their goals reads on track with a key result that has not
  moved past the divergence window, so re-entering the same number does not keep
  it alive. A new read action, goals.myStreak, serves it and is in the agent
  core toolset.

• #161 35acd7e Thanks @agungksidik! - The sidebar is five areas instead of fifteen rows.

  Home, Updates, Goals, Work and the workspace's word for spaces, then Cycle for
  workspace administrators and space coordinators and managers, who also see the
  phase strip. Updates holds To do (the former Review) and Activity (the former
  Inbox) as tabs with one badge; Goals holds OKRs and KPIs; Work holds
  Initiatives and the Board. Search is in the topbar only, and the account pages
  and Admin are in t...

v0.4.0-rc.1

No changelog section was found for v0.4.0-rc.1. See the commit log.

---

Image: ghcr.io/open-okr/open-okr@sha256:a9cc2d116c7637c5c88616cd6c648f92d3d699955a846fe2e989e54f325c76d4

Verify the signature before you run it:

cosign verify \n  --certificate-identity-regexp "^https://github.com/open-okr/open-okr/.github/workflows/release.yml@refs/tags/" \n  --certificate-oidc-issuer "https://token.actions.githubusercontent.com" \n  ghcr.io/open-okr/open-okr@sha256:a9cc2d116c7637c5c88616cd6c648f92d3d699955a846fe2e989e54f325c76d4

v0.3.1

Patch Changes

• #128 2becbe2 Thanks @akmalakhpah! - The first published 0.3 release.

  0.3.0 was tagged and never published: its release job ran every package's
  tests at once on one machine, and a timing test failed under that load. The
  release job now runs them one package at a time. Everything in 0.3.0's
  notes
  ships in this release, and upgrading from 0.2 follows the "0.2 to 0.3"
  section of docs/runbooks/upgrade.md, data changes included.

• Updated dependencies []:

  • @openokr/adapters@0.3.1
  • @openokr/agents@0.3.1
  • @openokr/config@0.3.1
  • @openokr/core@0.3.1
  • @openokr/formats@0.3.1
  • @openokr/method@0.3.1
  • @openokr/ui@0.3.1

---

Image: ghcr.io/open-okr/open-okr@sha256:72ca59b9431730f2f1eb825646767c4a425653a0b2365b304b3991a0e4bcf09b

Verify the signature before you run it:

cosign verify \n  --certificate-identity-regexp "^https://github.com/open-okr/open-okr/.github/workflows/release.yml@refs/tags/" \n  --certificate-oidc-issuer "https://token.actions.githubusercontent.com" \n  ghcr.io/open-okr/open-okr@sha256:72ca59b9431730f2f1eb825646767c4a425653a0b2365b304b3991a0e4bcf09b

v0.2.0

Minor Changes

• #110 d44d1d7 Thanks @akmalakhpah! - A member now hears at most five nudges a week, where it was ten (METHOD.md
  §11). Anything past the ceiling is held and listed in the member's next
  morning summary, with the rule it cites and a link to what it is about, so
  nothing held back is lost. The morning summary itself is not held by the
  ceiling and does not count against it, and an escalation about somebody
  else's work still always gets through.

• #111 70250f5 Thanks @agungksidik! - A cycle and an objective ask for less before they will accept anything.

  A cycle demanded a sponsor, a facilitator and a publication deadline before
  phase one would read green, and every one of those is a question the person
  making it could usually only answer with their own name or a guess. The
  sponsor and the facilitator now default to whoever creates the cycle, and the
  publication deadline to the day before it starts.

  Nothing in the method loosened. Phase one still asks that a sponsor and a
  facilitator be named, and this names them; the publish gate still asks for a
  date before day one, and the default is the latest date that satisfies it,
  which is the only one the product can choose without making a judgement that
  belongs to a facilitator. Any of the three is changed on the cycle screen.

  An objective can be created from a title alone. Champion and reviewer default
  to whoever typed it, wherever an objective is created: the OKR screen, the
  drafting surface, the annual frame, the command line and the API all behave
  the same way, because the default sits in the one place they all pass
  through. A reviewer who is also the champion is reported by the quality checks
  rather than refused, which is what makes it something you can fix.

• #110 af81a77 Thanks @akmalakhpah! - The weekly session's step 2 is now "Discuss what dropped" (METHOD.md §7.2).
  It lists every key result scored low, and every one whose confidence fell
  since the last session that scored it. A low score needs a next action and
  its owner, due by the goal's next check-in, before the session moves on; a
  blocker is raised beside it only where something is actually blocked, and a
  blocker's own next action answers the score too. The session used to ask for
  a blocker on every low score.

  Two new actions, sessions.setNextAction and sessions.lowScores, and three
  columns on a session's confirmed confidences hold the next action, its owner
  and when it is due (migration 0129).

• #111 3ee286a Thanks @agungksidik! - Roles and permissions have a screen.

  The matrix arrived in the previous release with five actions and no page, so
  the only way to answer "who may edit an objective" was the command line. It is
  at Admin, Roles and permissions now, and it takes the same level as
  invitations and support access, because deciding what a role may do is
  deciding who can change what.

  Two cards. The matrix says what each role may do in each domain: objectives,
  KPIs, initiatives, tasks, comments, spaces and the workspace itself, each at
  nothing, view, comment, edit or manage. The list below says who holds which
  role, and offers "no role", which is a real answer: a member with none holds
  exactly what their bindings give them.

  Owner is drawn and drawn as fixed, with the reason on the row. Hiding it would
  leave somebody wondering where the most powerful role went.

  The screen says the rule that surprises people, rather than leaving them to
  discover it: a role raises what somebody may do and never lowers it. Lowering
  Member on objectives does not take away the edit a champion holds on their own
  objective.

  A role somebody still holds cannot be removed, and the screen says who would
  be stranded rather than moving them somewhere by itself.

• #111 edc2863 Thanks @agungksidik! - A space no longer decides who may edit an objective in it.

  It did, and that was the only answer to the question: every member of the
  owning space held edit on every objective in it. The rule could be stated only
  by reading the access tables, it could be changed only by moving people between
  spaces, and no screen showed it.

  The workspace role answers it now, on the roles screen an administrator can
  edit. That is what makes the screen mean anything: before this, lowering Member
  on objectives changed no level anywhere, because the space binding still
  granted the edit underneath it.

  Nobody loses the edit when this release lands. The role backfill runs first and
  gives every active member the Member role, which grants edit on objectives, so
  the level is the same number from a different source. What changes is that an
  administrator can now change it.

  Who can see an objective is untouched. Every member of the workspace reads
  every objective, as they always have.

  An initiative follows the same rule: its space no longer grants edit either,
  and the role decides. The owner keeps full control of their own initiative.

  A space still decides its own membership, its session cadence and everything
  that reads them.

  A review is unchanged for everybody in the room. Scoring a key result,
  writing the narrative, taking the close decision and revealing the votes used
  to ask for edit on the objective, which worked only because a space granted it
  to everybody. They ask to see the objective now, and being in the room is what
  says you may take part. Somebody outside the room still cannot, and somebody
  holding the narrowest role can still score with their team, which is what the
  method asks for.

• #110 9d1b0e4 Thanks @akmalakhpah! - A space's check-in frequency now takes effect (METHOD.md §7.1). It was stored
  and read by nothing, so a team that chose every two weeks went on being asked
  every week. A goal created in the space takes the space's frequency, and
  changing it moves the open goals that follow it, each with its next due date
  counted from the new frequency; a goal set to a frequency of its own keeps it.
  The space card no longer offers "quarterly", which no goal checks in at.

  A quarter's planning now opens four weeks before it starts, where it was
  three.

• #110 9172e0b Thanks @akmalakhpah! - The rhythm streak now counts a space's own check-in periods (METHOD.md §7.4):
  weeks, fortnights for a space on every two weeks, or months for a monthly
  one. A team on every two weeks no longer breaks its streak in the week
  between, and the streak-at-risk warning reads the same periods. Booking a
  cycle's rhythm books one check-in per period, and whether the cycle is booked
  is judged the same way.

• #110 4e38c5a Thanks @akmalakhpah! - A key result may now be saved before its target is known, and an addition
  says what it still lacks (METHOD.md §2.9). goals.addKeyResult no longer asks
  a metric or a maintain key result for its targetValue; one saved without it
  reads "no target yet", fails the completeness check, reads 0% progress and
  forecasts nothing, and its first target asks for no reason. targetValue is
  null in every read until it is set. An objective or a key result added
  mid-cycle is live once it passes the checks set to block, and until then the
  OKR list, the drawer and the diagram mark it "Draft: needs" what is missing,
  such as a target or a due date. goals.tree answers the same as a draft on
  each objective and key result, null when it is live.

• #110 fecd1c1 Thanks @akmalakhpah! - Stage 9 of the quarterly review is "Keep, modify, abandon or defer"
  (METHOD.md §8.8). An objective now closes as one of five decisions:
  achieved, keep, modify, defer or abandon, each with its meaning on the
  screen. An unfinished aspirational objective shows "Proposed: Keep" where
  the workspace carries forward, and the room still chooses.

  A deferred objective is fed forward to the next cycle's issue list when the
  cycle is archived, at the carry-forward impact, once.

  Migration 0136 widens the close decision on goals and on review decisions.
  Every stored decision stays valid.

• #110 39c25e8 Thanks @akmalakhpah! - An objective or key result started mid-cycle now says so (METHOD.md §2.9).
  Once t...

v0.1.2

Patch Changes

• #103 9ea1aee Thanks @akmalakhpah! - 0.1.1 was tagged but never published either, so this is the first release of
  OpenOKR you can install. Everything listed under
  0.1.1
  and
  0.1.0
  in the changelog is in it.

  Before it publishes anything, the release run boots the Compose target from
  nothing and checks it. That check read the logs of a stack named openokr,
  while the release run had started its stack under another name. It found no
  log and reported that migrations had not run, when they had.

  deploy/docker/smoke-test.sh now checks whichever stack ./openokr up
  started: the one COMPOSE_PROJECT_NAME names when it is set, and openokr
  otherwise. If you run the smoke test yourself under a project name of your
  own, it now tests that stack rather than one that is not there.

• Updated dependencies []:

  • @openokr/adapters@0.1.2
  • @openokr/agents@0.1.2
  • @openokr/config@0.1.2
  • @openokr/core@0.1.2
  • @openokr/method@0.1.2
  • @openokr/ui@0.1.2

---

Image: ghcr.io/open-okr/open-okr@sha256:a941e26f4797485054c40ec7fbed0ffa14cc63ae0a4b899883e36ce12f33229a

Verify the signature before you run it:

cosign verify \n  --certificate-identity-regexp "^https://github.com/open-okr/open-okr/.github/workflows/release.yml@refs/tags/" \n  --certificate-oidc-issuer "https://token.actions.githubusercontent.com" \n  ghcr.io/open-okr/open-okr@sha256:a941e26f4797485054c40ec7fbed0ffa14cc63ae0a4b899883e36ce12f33229a

v0.1.2-rc.1

No changelog section was found for v0.1.2-rc.1. See the commit log.

---

Image: ghcr.io/open-okr/open-okr@sha256:1709f668dff57990d8dc968a39c67d0ed655785591d2fe8e6c7d6c21cca071d5

Verify the signature before you run it:

cosign verify \n  --certificate-identity-regexp "^https://github.com/open-okr/open-okr/.github/workflows/release.yml@refs/tags/" \n  --certificate-oidc-issuer "https://token.actions.githubusercontent.com" \n  ghcr.io/open-okr/open-okr@sha256:1709f668dff57990d8dc968a39c67d0ed655785591d2fe8e6c7d6c21cca071d5