HRPulsar

Changelog

New features, improvements, and fixes in each release.

2.5.3

October 16, 2026
Improved
  • •Global search finds every employee, assessment, vacancy and candidate you can see, not just the first page of each — and a division manager finds colleagues outside their own divisions
  • •Employees: the search also matches the job title
Fixed
  • •Search: % and _ typed into a search box match those characters instead of acting as wildcards
  • •Recruitment: saving a candidate card with an email another active candidate already has says the address is taken instead of failing with an error

2.5.2

October 15, 2026
Improved
  • •Section access: view access to Dictionaries opens every dictionary tab read-only, and the defaults now fit the roles better — HR edits dictionaries, division managers read the competence library, and recruiters, hiring managers and employees keep the competence pickers without the library page
  • •Coverage: HR sees every process without being added to it
Fixed
  • •Recruitment: a manager or hiring manager who does not evaluate a round sees its results once it is completed or archived — the progress row, the average and the round average per competence — and evaluator names no longer show above the sheets
  • •Company: a division head sees the people of divisions outside their own, and employees see each colleague's specialization on a division page again
  • •Coverage: the process and project counter follows the search and filters, a process or project needs a description before it is saved, and the access dialog closes after Save access
  • •Talent Market: a division head files a card only under their own divisions

2.5.1

October 13, 2026
Improved
  • •Development plans: ticking items off a plan is free — on HRPulsar Cloud, credits are charged only when a plan is submitted for review or returned
  • •Coverage: when a step breakdown or skill generation fails, it says why in your language — including when there were not enough credits
Fixed
  • •Self-hosted: a fresh install pulls the bundled file storage again — it now runs a pinned community build of MinIO, and existing installs switch over on their next update with nothing to migrate
  • •Public check: the Dictate button asks for the microphone again once you allow it in the browser's site settings, instead of still reporting it as blocked
  • •Employees: opening a colleague's card that was deleted or is outside your view shows a clear "not found" page instead of a loading error
  • •Billing: the duration shown for an AI operation is its real time, not the sum of calls that ran in parallel

2.5.0

October 12, 2026
Improved
  • •Roles: section access is configured right from the Roles page — roles, their people and what each section allows now live in one place
  • •Roles: the capability column wraps on small screens so the people counts stay visible, and the page hints explain how the counters add up
  • •Recruitment: recruiters and HR open the settings tabs their permissions already cover — assessment scales, data retention and roles are no longer admin-only screens
  • •Recruitment: the divergence canvas opens from a visible link in the candidate row and the mobile card — no hover needed, works on touch
  • •Self-hosted: the bundled database image is stock PostgreSQL
Fixed
  • •Recruitment: a manager with view access who is not on a round's evaluator list sees the round's evaluation sheets read-only instead of an empty tab, for active and closed rounds alike
  • •Company: a division head cannot hand their own manager or deputy seat to someone else — reassigning it stays with HR and admins
  • •Employees: deleting an employee locks their account out of the workspace immediately; re-hiring works through a new invitation and grants only the invited role
  • •Coverage: the button that changes a step's executor sits right next to the executor it changes

2.4.0

October 9, 2026
Added
  • •Self-hosted: self-serve workspace sign-up can be switched off for company-internal installs; invitations keep working
  • •AI workforce: when the owner of an agent workflow leaves, the workflow is marked as needing a new owner and admins are notified
Improved
  • •Self-hosted: HRPulsar runs on stock PostgreSQL — no database extensions needed anymore
Fixed
  • •Company: rank-and-file employees no longer see the company Profile, Positions and Specializations tabs
  • •Positions: AI position generation and the drafts-awaiting-review banner show only to admins and HR
  • •Specializations: the pointer to Settings → Dictionaries shows only to those who can manage the dictionaries
  • •Employees: the directory shows each colleague's specialization and status, lists active and on-leave people and filters by status; a division manager's list covers the whole company
  • •Employee profile: a hire on the timeline names the position in your language
  • •Recruitment: the shipped Standard 1–4 assessment scale, its description and level labels follow the interface language
  • •Recruitment: the sample interview question set opens instead of failing
  • •Recruitment: a round scored only by external evaluators shows its progress and can be marked complete
  • •Recruitment: the "Transcribe and run AI analysis automatically after upload" switch keeps its position after a reload
  • •Recruitment: the link in the submitted-assessment email opens the page at the Manager assessments block
  • •Recruitment: the consent link resend confirmation shows only when the link was actually sent
  • •Recruitment: a hiring manager reads and exports interview questions without editing them and fills in only their own evaluation sheet; a vacancy's hiring manager can be named an evaluator on its rounds
  • •Self-hosted: visitors behind your own trusted proxy keep their own IP addresses and per-IP rate limits
  • •Live demo: the Manager role filter no longer lists the HR people who lead a division

2.3.0

October 6, 2026
Added
  • •Recruitment: vacancy status commands — Publish, Pause, Reopen, Close with or without a hire, and Cancel. A paused, closed or cancelled vacancy is read-only, along with its interviews and candidates
  • •Recruitment: Offer accepted and Probation completed stages in the default funnel, and funnel conversion per stage and overall on the vacancy Analytics tab
  • •Recruitment: run Whisper transcription and your own LLM on a self-hosted OpenAI-compatible server, connected by its address, with a connection check for every provider
  • •Exams: download the question template and import questions from Excel, with a count of what was imported and the reason for each skipped row
  • •Talent market: a card posted from recruitment names the vacancy it is based on
Improved
  • •Recruitment: GDPR export and erasure open from the candidate page and the recruitment settings; the export covers contact details, the parsed resume, AI verdicts, evaluator sheets, comments and question sets
  • •Recruitment: AI Insights, question generation, transcripts and their analysis wait for the candidate's signed consent, and the candidate can withdraw it from their consent link
  • •Recruitment: retention settings are applied every night to interviews, resumes, consents and reports
  • •Recruitment: interview analysis keeps only quotes found in the transcript, with their own timecodes; a score without a quote reads "Unconfirmed"
  • •Recruitment: a division manager reads the Manager assessments block and fills in only their own sheet
  • •Employees: the Problems filter also finds incomplete profiles, plans awaiting review and overdue assessments, oldest first
  • •Analytics: people on leave count as employed in turnover, and metrics without data say what they need to appear
  • •Dashboard: the Active assessments, In development and Completed tiles open lists filtered to exactly what they count
  • •Data import: the preview shows each row as the import reads it, XLSX included, and the CSV template opens in columns in Excel in any locale
  • •Date pickers, position pickers, searchable filters and popovers behave the same inside dialogs as everywhere else
  • •Self-hosted: Caddy, PostgreSQL and Redis images are pinned to fixed versions, and the frontend no longer sends Next.js telemetry
Fixed
  • •Recruitment: GDPR erasure changes only the requesting company's data and also removes AI analysis texts, notes, reports, exports and the candidate's details in logs and notifications
  • •Recruitment: the interview page shows interviewer names to everyone who can open it, and the audit log names who moved a candidate or started an analysis
  • •Recruitment: internal candidates show their match against the vacancy's current competences
  • •AI workforce: an agent assignment or workflow always keeps an active owner when an employee leaves
  • •Assessments: growth zones list the lowest match first and strengths the highest first
  • •Employee profile: the Last assessment tile and the timeline read correctly in your language
  • •Daily background jobs run at fixed UTC times and no longer slip by a day after a redeploy
  • •A clear error instead of a server error when the embedding provider fails

2.2.0

October 2, 2026
Added
  • •Section access: admins decide, per role and per menu section, whether it opens nothing, view or manage — or hide a section for everyone. Today's access is the starting point, so nothing changes until you change it
  • •Division heads draft their own competences; only they, HR and admins see a draft until HR or an admin publishes it
  • •An AI call log: every AI call with its model, tokens and response time, and a transaction panel showing who ran it, on what, with which model, how long it took and how many calls it made
  • •Settings → AI lists the outside AI providers allowed to receive your workspace data, and the list can be narrowed per workspace
Improved
  • •Coverage, analytics, exams, assessments, the talent market, development plans, competences, dictionaries, company structure, employees, recruitment and settings all follow the section access setting, each still limited to the part of the org structure a person may reach
  • •HR and division managers can send invitations: HR across the workspace, a division manager into the departments they lead
  • •Recruitment: a candidate is never offered as an interviewer or evaluator on their own vacancy, and an employee who applies to a vacancy no longer sees its candidates, interviews or scores
  • •Recruitment: interview questions are generated in the background, with progress and a Cancel button, while the rest of the page stays usable
  • •Recruitment: archived vacancies are tucked away behind "Show archived (N)"
  • •Recruitment: the manager's percent of scale is computed the same way in every view and in the vacancy report
  • •Coverage: the list of processes and projects has an Owner column and a Clear button for search and filters, and the actions on a step row are plain buttons
  • •Coverage: "Get the skill" asks for a confirmation that names the cost before it starts
  • •Coverage: a description you broke down before comes back at once for thirty days, the same text gives the same steps, and repeat breakdowns cost less
  • •Dashboards: the Completed tile counts development plans finished in the period and how many met their deadline; a competence counts as raised only when an assessment after the plan scores it higher
  • •The live demo's development plans each offer a book, a course, an article and a video
  • •Action menus are as wide as their longest command and wrap long commands instead of cutting them off
Fixed
  • •The employee card's events feed names each change in the interface language, and a role change reads as old role → new role
  • •Recruitment: the manager score on the candidate page updates without a reload, Back returns to the list you came from, and the vacancy attachments block is back
  • •Recruitment: a candidate who already has interviews or scores on a vacancy is moved to Rejected or Withdrew instead of being removed with their data
  • •Recruitment: a recording-consent resend goes to the candidate's current email in the company's language, replaces an expired link, and an old link can no longer be signed
  • •Recruitment: the "transcript updated" banner shows only when a transcript arrives after the analysis
  • •Assessments: saving the same answer twice no longer breaks the survey
  • •Coverage: a hidden or deleted process opened by a link says it was not found, and a step keeps its executor when that person goes on leave
  • •Cancelling an AI analysis returns the credits to the pool they came from
Security
  • •Semantic search only finds and updates entries of your own workspace
  • •Development plans, divisions, salary analytics and talent-market match details are readable only within the reader's part of the org structure

2.1.1

September 25, 2026
Added
  • •The live demo has a "Talk to us" button: leave your name and an email, and we get back to you
Improved
  • •The list of processes and projects counts both under its title and can be searched and filtered by type and status, like the other lists
  • •A step added without capabilities reads "Not classified", with a hint to pick them or reclassify the step, instead of passing for a step that needs none
  • •People on a step show their position under the name, and a step's hourly rate starts at the company rate and can be set before the company has one
  • •The agent setup guide reads as a section of its own, and a competence without an assessment says "No assessment"
Fixed
  • •A page translated by the browser no longer breaks into "Something went wrong" after an action or a navigation
  • •The "capabilities not recalculated" hint goes away once you change the capabilities by hand or accept the step
  • •The demo banner shows the time left in the interface language

2.1.0

September 23, 2026
Added
  • •Coverage: set a step's automation mode and agent type by hand, right on its row. The value you set drives the hours, the shares, the money and the To do list, and you can return to the computed one at any time
  • •Already automated: the hours that an agent you registered already does, on the process and as a percent in the list of processes and projects
  • •An agent setup guide groups the steps by agent type — hours a year, whether you have registered the agent, how many skill files are ready — and downloads every ready SKILL.md with a README as one zip
  • •A step can carry its own hourly rate; the company rate stays the default, and the money of a process is summed step by step
  • •The "assessed" and "expected" labels of a matched person open what the match stands on: the competences, the assessment scores and the passing score
  • •People on a step link to their employee cards, step titles in the summary jump to their row, and an agent type opens a description of what it does
Changed
  • •The Coverage summary shows one bar of yearly hours — to an agent, to review, with people — instead of two scales
  • •Review is counted as it happens: by default half of a reviewed step's hours stay with the person who checks the agent's work, adjustable per step, and the hours freed include what review gives back
  • •After you reword a step, the editor says its capabilities were not recalculated and offers to reclassify it
  • •Clearer wording across Coverage: every figure on the step rows, the summary, the To do sections and the hours editor says what it means
  • •The live demo opens straight on Coverage, with two processes where a real share of the work moves to an agent
Fixed
  • •The reclassify dialog fits narrow screens and longer languages without horizontal scrolling

2.0.0

September 16, 2026
Added
  • •Coverage — a new section that answers who does each step of your work. Describe a process or a project in plain language and the AI drafts the ordered steps, tagging every step with the capabilities it needs from a fixed catalog of seventeen. You correct the draft in place: rename, reorder, add or remove steps, change any capability
  • •Per step, a verdict on who covers it — an agent type from the built-in packs or an agent you registered, a person on the team whose assessments or grade carry the capabilities, or nobody yet — with the automation mode and how well an agent does that particular work
  • •What the work costs: hours a year per step that you correct by hand, the share that moves to an agent, the share that moves to review and the share that stays with people, and the money behind those hours at one hourly rate you set on the company profile
  • •A To do tab with everything the breakdown leaves unclosed: the steps nobody covers, which open a draft vacancy in Recruitment with the steps as its description and the closest competences filled in, and the steps an agent type could take where you have registered no agent and written no skill file
  • •Assign the person who does a step and the person who checks and signs it; an assignment closes a gap and takes the step out of the hiring list
  • •A ready-to-run SKILL.md for any step an agent can perform, built from the agent pack's skeleton and your own competence indicators — preview it, download it, and run it in your own tools
  • •Process access: every process has an owner and is visible either to the whole company or only to the roles, positions and people you name, with a history of who changed what
  • •A catalog of seventeen capabilities, and an AI mapping of your competence library onto them, so people are matched to steps from assessments you already have
  • •An AI workforce registry API: the built-in agent packs, the agents your company actually runs with vendor, cost, data classification and EU AI Act risk level, assignments to employees with a supervision level and an accountable owner, multi-agent workflows, and an audit feed
  • •Imported employees receive an email with a link to set their own password, and an admin can resend it from the employee card until they do
  • •One password reset link now works for an email address that has accounts in several organizations — it sets the password for all of them
  • •On HRPulsar Cloud, an individual organization can get its own workspace branding: its logo instead of ours, a theme preset, an accent colour and no version badge
Fixed
  • •Employee import accepts the CSV template from the import page as it is, semicolon-separated files and DD.MM.YYYY dates saved by Excel included, and imported accounts no longer share a default password
  • •Password reset links work exactly once
  • •A bulk invitation keeps processing the rest of the batch when one address fails unexpectedly
  • •An answer scale can be made the default again, and a recruitment scale configuration can be activated, without the save failing
  • •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 column
  • •A long AI job is no longer started a second time when it outlives its timeout
  • •On HRPulsar Cloud, a data import is charged for every row in the file, and the import page tells a demo visitor what the rows cost
Security
  • •Competence indicators are scoped to the workspace that created them
  • •A password change or offboarding now also closes live connections, not just future sign-ins
  • •Uploads are rejected by size before they are read, and the delivery webhook that reports on outgoing email verifies the provider's signature
  • •Sign-in and the other rate limits are enforced across every application instance, so a busy deployment throttles as one system
  • •External evaluation pages are rate-limited and their free-text fields are bounded
  • •Fetching a company logo by URL pins the address it resolved to, the health endpoint no longer echoes connection details, test-only endpoints stay off on deployed installations, and a self-hosted stack refuses to start on the default signing secret

1.23.2

September 11, 2026
Improved
  • •The default recruitment funnel is named in your workspace's language instead of always in English, on workspaces that already exist as well as new ones
  • •Stage colours are picked from a palette instead of typed in by name, so only real colours can be set
Fixed
  • •Dropdowns name the option you picked rather than showing its internal code — participant roles in a 360° assessment, exam status changes, dictionary filters and company size

1.23.1

September 9, 2026
Improved
  • •A candidate's Manager score comes from the last completed round in hiring order — pre-interview, interviews, final — and its tooltip names that round on the candidate page and in the vacancy candidates table
  • •Every competence in a scoring round shows who gave which score, with a warning only where evaluators genuinely disagree, and missing Pre-interview and Final rounds hold their place in the round strip
  • •Employees can open the Talent Market again and see the cards they are candidates on, with their own row and match breakdown
  • •The vacancy owner is emailed as soon as an external evaluator submits, with a link straight to the candidate's manager assessments
  • •The assessment canvas filters manager and AI scores by round, uses the vacancy's own rating scale, keeps the candidate column on screen while the matrix scrolls, and exports exactly what the toolbar shows to XLSX and CSV
  • •AI dashboard summaries name grades the way your Grades reference does and describe findings in plain words instead of internal codes
  • •The dashboard period switch moves every tile, and demo workspaces report different figures for 30, 90 and 365 days
Fixed
  • •Managers can no longer change candidates, rounds, questions or evaluator invites on another division's vacancy, and next-round question generation reads only the candidate's own interviews
  • •External evaluation links no longer download the resume on their own, stay read-only once the round is completed or archived, and explain why a link stopped working
  • •The vacancy XLSX report includes the AI score, verdict and summary of the candidate's last completed analysis
  • •Vacancy candidate rows switch from "Analyzing…" to the verdict without a page reload, the interview-questions email link scrolls to the right block, a regenerated question set shows its latest date, and Schedule interview no longer offers Pre-interview
  • •Internal candidate names link to employee profiles only for people allowed to open them, internal search stays on once a vacancy has a talent card, and a division page shows the specializations mapped to its empty child departments
  • •The divergence threshold of an existing assessment scale can be viewed and changed on the scales page
  • •Demo workspaces: learning materials link to what they name, the flagship vacancy ranks its candidates in order, resume citations highlight only the quoted words, and employees see why they cannot start a plan themselves

1.23.0

September 7, 2026
Added
  • •Completed assessment results are split into Growth zones and Strengths, and a development plan can be built from the growth zones in one step, linked back to the assessment it came from
  • •HR metrics on the analytics page: turnover, promotions, time between grade changes, plans closed, people at risk, match and engagement over the last 30, 90 or 365 days, each marked exact or estimated — and the analytics page now opens for the HR role
  • •Development dynamics for a chosen period — plans completed and competences raised — on both dashboards and in the employee profile
  • •Assessment results show where the self-assessment and the manager's view diverge, per competence and overall, as soon as the assessment is on review
  • •Employees on a vacancy's internal shortlist can be moved into its candidate pipeline in one click, with their profile mapped onto the candidate's resume
  • •Interviewers are notified by email when they are added to an interview, when it is moved, when they are taken off it, and when it is archived
  • •The recording-consent link can be re-sent to a candidate without invalidating the one already in their inbox
  • •Talent Market candidate rows name which requirement falls short — competences, experience, or both; an employee who reacts below the bar receives a soft reply and can ask their manager for a development plan straight from the card
  • •Issue badges on an employee profile name the date the problem is due to resolve
  • •Assessments started from an employee card arrive with the title, format, criteria and rating scale already filled in
Improved
  • •Dashboard queues and the employee lists they open rank people by problem severity, and the assessed stage shows how many people are still unassessed
  • •Answer scales show only the option name in questionnaires and previews, with the level descriptions in one collapsed legend
  • •Assessment create dialogs explain when to use self, 180° and 360°; the Talent Market board explains vacancies, projects and talent cards and counts cards by type
  • •Saving a vacancy's required competences no longer waits for the internal match to finish; the shortlist is recomputed in the background
  • •Interview emails name the time zone of the interview time
  • •Statuses, types and categories across settings, billing and platform admin screens read as words instead of internal codes
  • •Employee cards, vacancy competence saves and the internal candidates block open faster
  • •Demo workspaces run one end-to-end talent story with named personas, real development materials and candidates whose cards match their list rows
Fixed
  • •Every list — employees, candidates, vacancies, exams, competences, dictionaries, development plans and the talent market — shows a load error with a retry button instead of an empty list when the server does not answer
  • •The employee profile header shows the last completed assessment, so it no longer contradicts the "no recent assessment" badge
  • •Grades and skill levels follow the reference catalog in the interface language on every screen that names one
  • •The salary range refills from the company library when position, specializations or grades change on the vacancy edit page, not only on create
  • •Candidate AI Insights show one banner instead of two, starting an analysis switches to the progress banner from every button and can be cancelled, and the analysis emails name the candidate and open straight on the AI Insights block
  • •Interview scheduling: interviewers are picked from the searchable employee list, the schedule modal fits its fields, Title is required, Duration and Interviewers are editable on the interview page, and a new round appears in the list without a reload
  • •Turning a vacancy's internal search off freezes its shortlist, deleting a required specialization removes its derived competences, and a generated vacancy profile records the language it was actually written in
  • •Vacancy Versions and Questions load for a division head, question generation works for manually added and bulk-imported candidates, and cancelling a running analysis keeps the candidate's AI data readable
  • •Vacancy analytics gained a Withdrew tile with the funnel's own terminal-neutral stage names

1.22.0

September 1, 2026
Added
  • •Hiring can start inside the company: one action posts a vacancy to the talent market as a card carrying its required competences, and the vacancy page lists the employees who matched with their percentage — above the external candidates. A vacancy can also be kept out of internal search with a single switch
  • •Vacancy competences can be picked straight from your competence library on the vacancy page, and a candidate who already works here is marked as an internal candidate everywhere they appear
  • •The employee card starts every action about that person — development plan, assessment, career event — from one menu with the employee already filled in, and shows the same issue badges as the directory
  • •A score below its assessment's passing bar is labelled a gap on the competences tab, with the development plan offered right there
  • •A talent market candidate row states how many required competences the person clears, opens the full breakdown from anywhere on the match cell, and builds a development plan from the gaps — before or after the appointment
  • •The candidate analysis card names, in words, the competences where the manager and the AI disagree and which way the AI leaned
Improved
  • •The dashboard reads as a funnel: connected stages with a share bar, an assessment-cycle card across the width, and a gaps tile that only turns red for gaps with no plan behind them
  • •Every section reachable from the sidebar, and every number on the dashboard, carries a question mark that explains what it counts
  • •The vacancy candidates table leads with the manager and AI match, gives the reason for the AI verdict in the row, and explains an unassessed candidate instead of showing a dash
  • •Manager/AI divergence has one definition everywhere, so a row flag can no longer contradict the count beside it
  • •A bulk invitation answers with a per-address outcome, so an address that could not be invited comes back with its reason
  • •On HRPulsar Cloud, price and pack changes are live everywhere within a minute
Fixed
  • •Statuses, roles, formats and categories read as words instead of internal codes across recruitment, assessments, development plans and the credit price list
  • •Vacancy profiles, interview questions, resume and interview analyses and their reports are written in your workspace's AI content language instead of English, and opening the AI settings page no longer pins that language behind your back
  • •The interview analysis matrix shows scores on your own assessment scale instead of rendering every score as a short red bar
  • •Resume upload works with S3-compatible storage that rejects request checksums, and a file the platform could not store reports the real storage error
  • •Resume fragments an analysis quoted are marked in the parsed resume and click back to their citation, and a verdict based on an interview alone says so instead of claiming there is no AI data
  • •The positions catalogue no longer shows a colleague's grade, specialization or salary band to rank-and-file staff, and the dictionary usage preview is admin-only
  • •Talent market cards score candidates who were assessed without a target skill level, keep required competences when a specialization is edited, and carry the specialization and grade of the vacancy they came from
  • •An assessment whose criteria resolve to no indicators is refused at send instead of producing an empty questionnaire
  • •A throttled sign-up address is blocked for one hour rather than for as long as it keeps retrying
  • •The vacancy candidate list opens faster on large funnels

1.21.0

August 25, 2026
Added
  • •A full role model: HR, Recruiter and Hiring manager join Admin, Manager and Employee. Every account now carries a role, an admin sets a colleague's from their employee card, and the employee list shows and filters by it — with guard rails against changing your own role, demoting the last administrator, or taking Manager away from someone who still leads a division
  • •A read-only Roles page in the Admin area: every role in the workspace, what it can actually do, and how many people hold it — the count opens the employee list filtered to them
  • •An employee directory for everyone: rank-and-file staff can look a colleague up by name, job title and department, with grades shown only if the workspace switches them on. "My profile" opens your own card in full from the header menu and the dashboard
  • •An Issues column and filter on the employee list: no active account, incomplete profile, overdue or stalled assessment, competence gap with no development plan, overdue plan, no assessment in the last six months
  • •Interview transcription adds AssemblyAI and Yandex SpeechKit alongside Deepgram and Whisper. Every key you configure joins one fallback chain, so a provider that refuses a recording hands it to the next one instead of failing the interview
Improved
  • •Dashboard tiles and findings open the employee, plan and assessment lists already filtered to the people they counted, and the names under a finding link straight to the person's card — the number on the dashboard and the rows on the page always match
  • •Interview recordings are prepared before they leave: the video track is dropped and the audio re-encoded, so an hour of interview travels as a few megabytes instead of half a gigabyte, and a recording above a provider's size limit is split into overlapping parts and stitched back into one transcript
  • •When the provider that transcribed an interview does not separate speakers, the turns are split into interviewer and candidate from the transcript itself, so the analysis knows whose words it is scoring
  • •Recordings are transcribed in your workspace's content language, or in the language the provider detects, instead of always English
  • •Demo workspaces are staffed in the language they run in — a German demo shows German employee names and emails — and now span the whole role model instead of managers and employees alone
  • •Applied list filters are marked with the accent colour, and the employee filters follow the order of the columns they sit above
Fixed
  • •The public demo throttles each visitor separately again: behind a reverse proxy every visitor shared one hourly bucket, so a handful of demo starts closed the demo for everyone
  • •Opening a candidate who has no linked person record no longer fails with a server error
  • •Demo development plans carry real statuses, so a finished plan is no longer counted as an open, overdue one on the dashboard
Security
  • •Reading is now scoped as tightly as writing: hiring data — candidates, resumes, interviews, reports and recruitment settings — is reserved for the roles that do hiring, a division manager sees their own department's vacancies, the sections under an employee card open only for the person themselves, their division manager and admins, and position and specialization employee lists no longer hand hire dates, employment status, grades and HR alerts to every member of the workspace
  • •Division managers create, edit and calibrate assessments, development plans and talent cards for their own subtree only, and analytics figures and assessment exports cover their own people rather than the whole company
  • •The shared catalogues — competences, indicators, materials, answer scales and a specialization's grade ladder — are edited by admins and HR; positions stay with the division manager for their own subtree
  • •The sidebar and the company page only offer the controls a role can actually use, and every account carries at least the baseline employee role
  • •A throttled demo visitor can no longer keep its own block alive by retrying, or pick its own throttle bucket with a forwarded-for header

1.20.0

August 20, 2026
Added
  • •The dashboard is rebuilt around your development loop: assessed → gaps → developing → closed, with an action queue that turns detected problems — employees below the bar without a plan, overdue plans, stalled reviews, low assessment coverage — into deep-linked next steps
  • •A personal dashboard for every employee: their own loop stages, action queue, strengths with rare-skill highlights, a growth direction to the next grade, and an assessment-history sparkline
  • •One-click AI summaries on both dashboards: the management view explains what is going on and what to do first, the personal view reads like a coach — both cached per data state, with a fair daily allowance per workspace and per person
  • •YandexGPT joins the platform's first-class AI providers, including bring-your-own-key configurations
  • •On HRPulsar Cloud, manual-invoice billing now issues numbered PDF invoices from your installation's own template: buyer requisites collected at checkout, the document emailed and downloadable, with an operator queue to confirm payments by bank-statement date
  • •On HRPulsar Cloud, you can keep everything you built in a demo sandbox: check "keep my demo data" when requesting access and the sandbox becomes your real workspace on approval
Fixed
  • •An explicitly configured passing bar of 0 is respected across the dashboards instead of being treated as 75
  • •The personal dashboard no longer lists below-the-bar scores as strengths
  • •The daily AI-summary limit now tells you it was reached instead of suggesting a retry
  • •The dashboard shows a clear error with a retry button when the server is unreachable
  • •Invoice dates and numbering follow your installation's timezone, and a payment made while the invoice was valid can be confirmed later without an error

1.19.2

August 18, 2026
Fixed
  • •The demo session banner is readable in light theme again

1.19.1

August 18, 2026
Changed
  • •Demo sessions now start on the dashboard, giving a full overview of the workspace before diving into recruitment
Fixed
  • •The demo's sample interview with a completed AI analysis opens correctly again

1.19.0

August 16, 2026
Added
  • •A feedback widget in the app header: rate HRPulsar with a thumb up or down and leave a suggestion, in your interface language
Changed
  • •Evaluator invitation emails are capped per recipient address per hour on top of the per-invitation resend cooldown, so an inbox can't be flooded
Fixed
  • •The interview page shows readable labels for the analysis verdict, process findings and red flags instead of raw codes, and switching vacancies on a candidate card no longer carries the previous sheet's pending saves over
  • •An evaluator invitation that is revoked after the evaluation was submitted keeps that evaluation readable, marked as no longer counted toward the round score
  • •Onboarding's invite step asks for a name next to each address, reports which invitations failed, and keeps a way forward when some of them do
  • •The status page reads its whole 90-day uptime bar under one formula, so older days are no longer shown next to newer ones counted differently
  • •Signup moderation decisions reach the applicant in the language they signed up in, and the approved account keeps that language
  • •On HRPulsar Cloud, credit packs and annual contracts are priced from the installation's own billing profile, and manual-invoice checkout quotes the legal entity its invoice is issued against

1.18.2

August 14, 2026

1.18.1

August 14, 2026

1.18.0

August 13, 2026
Added
  • •Sign-in, registration and invitation pages got a language switcher, and invitation links now open in the language the email was written in
  • •Bring-your-own-key setups warn when a stored provider key can no longer be read, with a prompt to re-issue it, instead of silently dropping the provider from the list
  • •Company → Specializations shows each specialization's Active/Inactive status at a glance
Changed
  • •Any active member of the company can now be a vacancy's hiring manager — no admin role required, and people who have left are no longer offered
  • •On HRPulsar Cloud, starting an interview upload reserves its credits for the length of the transfer, so a large recording can no longer run out of balance halfway through — and every credit warning now quotes the balance you can actually spend
  • •Custom AI endpoints on HRPulsar Cloud are validated against internal addresses when saved and re-checked on every call; self-hosted installs keep using private endpoints (Ollama, vLLM) as before
  • •Resending an external evaluator's invitation is throttled, so repeated clicks can't flood their inbox
Fixed
  • •Vacancy analytics count Hired, Rejected and In progress by the hiring funnel's own stages, and closing a vacancy as hired moves the candidate to the hired stage
  • •Settings → Prices lists your workspace's actual prices with the model multiplier applied, the confirmation dialog quotes exactly what will be charged, and quoted prices refresh when the workspace model changes
  • •Requesting a credit pack twice no longer raises a second invoice — a repeat request answers with the one already issued
  • •Vacancy → Questions shows each candidate's latest question set with names and a competence filter instead of an empty tab
  • •The Salary range fields are back on the vacancy forms, prefilled from the salary band of the chosen specialization and grade
  • •The candidate page's side blocks follow the sourcing timeline, the DIVERGENCE column compares manager and AI scores on the same scale and explains itself on hover, and saving a score no longer nudges the layout
  • •AI Insights verdicts, run history and recommendations are translated instead of showing raw codes, and Augment indicators follows the workspace content language without losing suggestions
  • •Interview analysis served from cache fills the assessment matrix again, and Analyze states its cost up front and warns when the balance can't cover it
  • •Emails follow the reader's language end to end: verification and password-reset emails match the language chosen on the form, deadline reminders name what they remind about in the reader's language, and notification emails render in the branded layout on white-label installs
  • •A translation missing from a locale renders the English text instead of a raw key name, error messages follow the interface language on self-hosted installs, and salary benchmarks use the site currency
  • •Salary and compensation currency defaults to the installation currency instead of a hardcoded one
  • •Division pages count specializations across the whole division subtree, with a switch to narrow back to the division itself
  • •Round averages and manager scores exclude sheets from revoked or declined invitations, and an evaluator arriving at a round someone else already holds gets a plain explanation
  • •The interactive demo opens on a finished AI evaluation with citations in place

1.17.0

August 9, 2026
Added
  • •Interview scheduling grew up: the Schedule dialog now captures vacancy, round, title, date, duration, interviewers and notes; every interview gets its own editable page; the Interviews list shows status chips and rounds, and archived interviews stay restorable for 90 days
  • •Interview question sets are now round-aware: generate a deeper set from completed interview transcripts, add questions by hand or straight from a competency indicator, edit every attribute inline, and jump to a freshly generated set right from its notification
  • •Candidate reports rebuilt end to end: pick the candidate scope, sheets and audience in one dialog; the Summary, Matrix and Detail sheets carry recommendations, divergence markers and blind spots; and a full-screen assessment canvas compares candidates across rounds with XLSX/CSV export
  • •Mass assessments gained an Analytics block: growth zones, top competences, employees at risk and top performers, with a filterable detailed view and per-grade summaries
  • •Manager assessments now support several evaluators on the same round — each with their own sheet and email invitation, colleagues' scores stay hidden until you submit your own — plus round archive/restore and per-invitation Resend, Revoke and View actions
  • •Assessment results show match percentages as colour-coded chips, so strong and weak matches read at a glance
Changed
  • •AI Insights speaks plainly: one Analyze action with both modes priced up front, a banner that flags an outdated analysis and offers the exact re-run that fixes it, and candidate-table AI columns that reflect what the model can actually see
  • •The AI model catalog is now the single source of truth: disabling a model really retires it everywhere, model pickers stay on curated versions, and bring-your-own-key setups are validated so a provider/model mismatch can't silently break AI features
  • •On HRPulsar Cloud, an audio or video interview upload now costs a flat 8 credits charged only when the upload completes — a failed upload costs nothing, and text transcripts are priced as the light-weight updates they are
Fixed
  • •The status page no longer declares an incident resolved while the platform is still unreachable
  • •Interview transcription with your own speech-to-text key works again
  • •Analysis notifications name the candidate and link straight to the AI Insights block
  • •A long list of fairness fixes in reports: partially scored candidates are marked "Additional check" instead of "Not recommended", divergent scores stay in the totals, and resume-sourced candidates are named properly

1.16.1

July 31, 2026
Added
  • •Manager assessments now show scores at a glance: the round header displays the running Average score, and every competence shows its cross-evaluator average with a warning highlight when evaluators disagree strongly — external evaluators count as soon as they submit
Changed
  • •Candidate scoring is self-consistent: rating indicators always derives the competence overall, while setting the overall by hand resets the indicator answers, so a sheet can no longer contradict itself
  • •Settings → Billing on HRPulsar Cloud is reorganized into tabs — Overview, Transactions, Prices and Settings — with a searchable credit price list back in place
  • •German interface texts received a native-style proofread: dozens of wording and grammar refinements across the app, error messages and emails
Fixed
  • •Scoring a candidate no longer scrolls the page away from the competence you were rating
  • •Interview transcripts (pdf and txt) upload properly again and are immediately ready for AI analysis; uploading a file that doesn't match the interview's planned type now asks before switching it
  • •The recording-consent email greets the candidate by name instead of "(unnamed)"
  • •Admin sections (Dictionaries, Invitations) are hidden from managers and employees, and direct links explain the restriction instead of showing an empty page
  • •Accepting an invitation pre-fills your first and last name from the invitation instead of leaving the email in the name field
  • •The division page regained its specialization filter in the Employees list, and the invitation dialog opened from a division closes properly
  • •The sidebar shows your actual strongest role right after promotions and demotions
  • •Completed exams without a description show "No description" instead of a blank space
  • •Assessments locked by a calibration now explain why the questionnaire is closed instead of hiding the buttons

1.16.0

July 31, 2026
Added
  • •HRPulsar now speaks German: each user can switch the interface language, workspaces can set a default, and error messages, emails and built-in reference data all follow along — English remains the default
  • •AI can generate in German too: competencies, indicators, positions, learning materials and development plans are produced in your workspace's chosen content language, selectable in AI settings and onboarding
  • •Bring your own AI: connect provider API keys per workspace, plug in local OpenAI-compatible servers (Ollama, vLLM, LM Studio) via a custom endpoint, and the AI settings page only offers providers with working credentials
  • •New AI model versions now appear automatically — a daily discovery sweep picks them up without waiting for a product update; on HRPulsar Cloud, newly discovered models go through moderation with explicit pricing before they become selectable
  • •White-label installs get curated UI theme presets, a configurable sidebar logo size and a custom sign-in background color or image
  • •On HRPulsar Cloud, credit packs can now be requested right from Settings → Billing with localized prices — credits arrive as soon as payment is confirmed
Changed
  • •AI tiers refreshed to the latest generation: Balanced now runs Claude Sonnet 5, Thorough runs Claude Opus 4.8, and Claude Fable 5 is available as an optional pick
Fixed
  • •AI generation works reliably on the newest reasoning models instead of occasionally failing with a service error
  • •Self-hosted installs with a misconfigured file-storage address now get a clear error at startup instead of silently producing download links that don't open
  • •Background processing (resume parsing, AI analysis, notifications) no longer degrades after several days of continuous uptime

1.15.5

July 24, 2026
Fixed
  • •Generating competencies with AI across a large competency tree no longer fails with a generic "AI service error" — long generations now complete, and if the output genuinely exceeds the model's limits you get a clear message right away instead of minutes of silent retries
  • •Blog articles no longer show stray styling code from the editor in cards and article text

1.15.4

July 23, 2026
Fixed
  • •Starting a demo no longer gets stuck on the browser check — when an extra confirmation is required, it now appears right on the page
  • •Creating an account or signing in while a demo session is active now opens the real form instead of returning you to the demo

1.15.2

July 21, 2026
Fixed
  • •Re-running an AI candidate analysis no longer fails when the candidate already has a completed one — the previous run is archived and kept in history
  • •AI-generated content in Recruitment (interview questions, vacancy profiles, candidate reports, transcriptions) now defaults to English everywhere; pick another language explicitly when you need it

1.15.1

July 21, 2026
Fixed
  • •White-label branding on self-hosted installs now applies out of the box — the web app picks up your logo, product name and colors from configuration automatically
  • •Uploaded files on self-hosted installs — company logos, interview recordings — now open correctly in the browser
  • •Background jobs that read uploaded files, such as resume parsing, now work on self-hosted installs without extra storage configuration

1.15.0

July 20, 2026
Added
  • •Self-hosted installs can now be white-labeled: set your own logo, product name, accent color and favicon via configuration — applied across the web app and outgoing emails
  • •Self-hosted installs now support self-serve sign-up, with sensible fallbacks when no email provider is configured
Changed
  • •HRPulsar Cloud now lives at app.hrpulsar.com, with hrpulsar.com serving the product site; old links, including those in previously sent emails, redirect automatically
  • •Opening the app's root URL takes you straight to sign-in or your dashboard
  • •The interactive demo sandbox is now exclusive to HRPulsar Cloud
Fixed
  • •Create and edit dialogs for employees, positions and divisions now show server errors right in the dialog — validation messages under the fields, conflicts as a visible banner — instead of failing silently
  • •The self-hosted Docker stack now starts out of the box: corrected service commands and health checks, a ready-to-use proxy config, and automatic storage and database setup on first boot
  • •Email links (verification, password reset, invitations) now follow the configured frontend URL on self-hosted installs
  • •Self-hosted installs show the real release version in the sidebar and the health endpoint
  • •Self-hosted installs no longer send cloud-billing requests, removing console noise on every page

1.14.3

July 16, 2026
Added
  • •Employee names in Assessments, Development plans and Exam results now show the person's current position, with a clear chip when they are no longer active
  • •Draft Talent Market cards can now be cancelled right from the card page
Changed
  • •Every date field in Assessments, Development, Exams and Talent Market uses the same calendar picker instead of the browser's built-in date input
Fixed
  • •Starting to score a Pre-interview round works again for the first evaluator; a second evaluator still gets a clear "single evaluator" message
  • •Development plan grade choices now follow the selected specialization: pick a specialization first, then choose among the grades configured for it
  • •Deactivating a built-in dictionary item now truly hides it everywhere: pickers, AI suggestions, the Specializations page and the public API all respect it
  • •Renaming a position updates it everywhere employees are shown — no more stale titles on cards and lists
  • •Specialization and grade pickers offer only active dictionary items, while a previously saved item keeps its name and stays selectable
  • •Completed and cancelled Talent Market cards are fully frozen — no more edits, candidate changes or appointments after the fact
  • •The exam pass mark you configure is what actually decides pass or fail; changing the mark re-grades finished participants
  • •Exam question images show at their natural size while taking the exam and appear in the submitted results
  • •Cancelling or completing an exam now settles its surveys: unfinished ones close, submitted results stay visible
  • •Declined evaluator invitations stay declined — the link cannot be reopened or extended
  • •Candidate pages opened from a vacancy keep that vacancy's context in assessments and AI insights
  • •Active hiring funnel stages are consistently highlighted across the vacancy page
  • •Restarting the demo resumes your existing sandbox instead of starting a new one
  • •Deleting a dictionary item that is still in use is blocked with a clear message in more places: Talent Market cards, vacancies and recommended grades
  • •Editing a vacancy's competence profile can no longer accidentally overwrite someone else's newer changes
  • •Saved assessment calibration keeps questionnaires closed, and calibrated totals keep their badge when calibration is reopened
  • •Billing charges only for what was actually created in mass assessments and Talent Market bulk-add — duplicates and over-cap items are free
  • •The Prices page lists only actions you can actually trigger from the dashboard

1.14.2

July 13, 2026
Changed
  • •Exam pass marks are no longer set in stone: a pencil icon next to the pass mark opens an Edit dialog, and it works on active exams too — only finished exams stay frozen
  • •The Send exam button no longer goes grey when the deadline has passed — clicking it now tells you exactly what is wrong, the same way Assessments and Development plans do
  • •Creating an exam is calmer: the Title field no longer glows red before you have typed anything
Fixed
  • •Deleting a competence type that is still used by competences is now blocked with a clear message instead of silently leaving those competences untyped
  • •The "no longer considered" Talent Market email no longer offers an Open card button that led nowhere — removed candidates have no access to the card
  • •Talent Market ranking now values experience properly: a candidate who meets or exceeds the required experience ranks above one who merely reacted first — reaction time only breaks exact ties
  • •Vacancy Requirements, Responsibilities and Conditions no longer come back empty after saving a draft
  • •Resumes built as DOCX tables now import correctly instead of failing with "Nothing to import"
  • •Resume parsing now carries each job's start date, end date and description onto the candidate card
  • •Downloading a resume from the candidate page now gives you the original PDF or DOCX file
  • •Each hiring funnel stage now shows in its configured color on the vacancy page, and the Manage funnel stages drawer no longer needs a horizontal scrollbar

1.14.1

July 12, 2026
Fixed
  • •The demo, sign-up and save-access buttons no longer get stuck on "Verifying..." when an ad blocker interferes with browser verification — you now get a clear explanation and a retry

1.14.0

July 12, 2026
Added
  • •External evaluators now get their own evaluation page — no account needed: the invite link opens the candidate's resume and interview questions next to the full competence sheet, scores save as they go, and a warning flags unassessed critical competences before submit
  • •Recruitment grew section tabs: Vacancies, Candidates, Reports, Audit and Settings are one click away from any recruitment page
  • •Vacancies now carry a hiring manager — pick one when creating or editing a vacancy (the creator is suggested by default), and see them right on the vacancy page
Changed
  • •AI vacancy profile generation now asks before overwriting: results land in a Review for save step where you keep, edit or drop each competence — nothing touches the saved profile until you apply it
  • •The vacancy competence block is fully editable in place: change a competence's criticality, add new competences with indicators and interview questions, and create whole groups and subgroups without leaving the page
  • •The vacancy list shows Position and Division columns, the owner's full name, and archived vacancies under the All statuses filter
  • •Evaluator invitations now wait until the vacancy has a competence profile, and the Invite button explains what is missing
  • •Assessment calibration is now priced at 5 credits per run
Fixed
  • •Evaluator invite statuses now track the real journey — pending, opened, in progress, submitted — and stale invites show as expired
  • •The invitation email now names the candidate being evaluated, includes your personal message, and links straight to the evaluation page
  • •The invite evaluators window keeps its buttons in view, clears cleanly, and checks email addresses before sending
  • •Candidate names in the vacancy assessment matrix no longer show as Unknown for resume-sourced candidates, and each name links to the candidate card
  • •Assessment surveys are billed once per participant on completion — saving answers along the way is free

1.13.19

July 10, 2026
Added
  • •Employees can now take their exams: a Take this exam panel opens right from the Exams list with single-choice, multiple-choice and essay questions, answers saved as you go, and a resume-where-you-left-off draft if you close midway
  • •Submitted exams show the earned score in the list, and a results view highlights which answers were right and wrong — managers with access to the employee can open the same review
  • •Exams complete themselves: the whole exam flips to Completed automatically once the last participant submits their results
  • •Taking an exam is billed once per completed attempt — answering and re-answering questions along the way is free
Fixed
  • •The manager assessment sheet on the candidate card now scores indicators individually (with the competence overall computed from them), shows each competence's criticality chip, and unlocks Mark as complete only after every critical competence is assessed
  • •Every evaluator now gets their own assessment sheet — fast scoring across several competences no longer loses saves

1.13.18

July 9, 2026
Added
  • •Interviews now live on the candidate card: schedule rounds, watch transcription and AI-analysis status at a glance, and jump into any interview's detail page
  • •Upload several interview recordings at once — one interview per file, with per-file progress; AVI recordings are now accepted alongside mp3, wav, m4a, mp4, webm and mov
  • •Auto-processing for interview uploads: recordings can be transcribed and sent through the AI analysis automatically as soon as they pass the safety scan — no manual clicks between upload and insights
  • •Interview questions moved onto the candidate card: per-vacancy question sets, a Generate next set button once a round is transcribed, and questions already answered in the interview are ticked off automatically after the AI analysis
  • •Notifications when a question set is ready (or fails), linking straight to the candidate — plus a workspace setting that shows hiring managers the questions above the resume
Fixed
  • •A candidate's signed recording consent now carries over to interviews created later — no more repeated signature requests

1.13.17

July 9, 2026
Fixed
  • •AI generation in Recruitment no longer fails on rich vacancies: large responses (vacancy profiles, resume parsing, question sets, interview analyses) were cut off mid-answer and surfaced as a cryptic error — the output budget is now four times larger, and profile generation automatically retries with a compact profile when even that is not enough

1.13.16

July 7, 2026
Fixed
  • •Assigning employees to an exam now offers Division, Position and Specialization filters with a Select all that respects the active filters — matching the Mass assessment picker
  • •The Position filter in the Mass assessment and Exams employee pickers works again
  • •The manager assessment scoring sheet no longer claims the vacancy profile has no competences
  • •Public share, consent and invite links are now rate-limited per endpoint, protecting them against token guessing
  • •Demo data seeding works reliably on fresh installations

1.13.15

July 6, 2026

1.13.14

July 6, 2026
Fixed
  • •Normalized AI candidate scores now map correctly onto your company's assessment scale, and the divergence indicator compares scores on the same scale — no more false divergence flags
  • •Header counters on the Vacancies and Talent Market lists show the true total instead of only the currently loaded page
  • •Dismissing a failed AI profile generation banner no longer affects a generation started in another tab

1.13.13

July 5, 2026
Added
  • •Create Vacancy now builds on your library — pick a Position, then Specializations, Grades and Division from dependent selects, and Competences & indicators pre-fill from the matching role matrix
  • •AI candidate scores can be viewed raw or normalized onto your company's own assessment scale, with a unified verdict across resume-only and full analysis modes
  • •Dictionaries gain Source and Status filters, a sort index column, filtered item counters, and per-company activation of built-in items
Changed
  • •Competences & indicators on the vacancy page are now edited inline — the dialog is reserved for AI generation
  • •Long vacancy descriptions collapse behind Show more across all Overview fields
  • •Item counters on Development, Talent Market, Exams and Recruitment lists now respect active filters
  • •Changing or resetting a password now signs out every other session, and sign-in endpoints are protected against brute-force attempts
Fixed
  • •Viewing the Analytics page no longer consumes credits — only explicit exports are charged
  • •Adding a new interview round asks for confirmation before it lands
  • •Failed vacancy profile generation shows the error inline with a Dismiss button instead of silently disappearing

1.13.12

June 24, 2026

1.13.11

June 24, 2026

1.13.10

June 22, 2026

1.13.9

June 22, 2026

1.13.8

June 20, 2026

1.13.7

June 20, 2026

1.13.6

June 20, 2026

1.13.5

June 20, 2026

1.13.4

June 20, 2026

1.13.3

June 20, 2026

1.13.2

June 19, 2026

1.13.1

June 19, 2026

1.13.0

June 18, 2026
Added
  • •AI Insights on the candidate card flags a stale analysis the moment a fresh CV is uploaded — a "Resume updated — re-analyze" banner re-runs resume-only (20 credits) and archives the prior run on success, so recruiters never act on a verdict that no longer reflects the candidate's resume
  • •Resume-only AI Insights now ships clickable citation chips — each per-competence quote links straight to the matching block in the parsed-resume editor (auto-expanded, scrolled into view, ring-highlighted) so a recruiter can verify the AI's claim in one click
  • •AI analysis in-flight card renders all six pipeline stages with live progress (resume-only greys out two stages it skips), polls every three seconds and exposes a Cancel button — cancelling before the first LLM call refunds 100 % of the credits
  • •Candidate card AI Insights gains a split-button: primary fires Resume only (20 credits), the chevron dropdown adds Resume + interview (40 credits) with a tooltip explaining when the full-mode option is unavailable
  • •Consolidated XLSX recruitment report rebuilt around a 4-sheet layout — Summary ranking with computed recommendation, Competence Matrix with divergence highlight, per-candidate Detail sheet, and an Incomplete-data sheet; the report wizard now exposes a Hiring-Manager audience that swaps raw process findings for a neutral reframe
  • •Vacancy candidates table gains per-row Manager % / AI % match, a new Divergence column with an amber badge linking to Canvas pre-filtered to that candidate, a tooltip preview of the top-5 divergent competences, and a By Manager / By AI / Custom sort control that persists per recruiter
  • •Recruitment Canvas gains a full Versions panel — vacancy-scoped audit timeline with evaluator / candidate / date / "only divergence-triggering" filters and a Revert button on Manager scores, plus a conflict modal when two recruiters edit the same row in parallel
  • •New Assessments tab on the vacancy page — Compact matrix with side-by-side Manager / AI cells, click-to-footer evaluator breakdown, per-candidate AI vs Manager %-match, and a per-tenant divergence threshold setting
  • •Resume-only AI analysis lands on the candidate card with a top-up upgrade to full (+20 credits within 30 days), bulk fan-out from the vacancy candidates table, an analysis history modal, and [resume only] / [full] sub-badges so recruiters never confuse a pre-screen with a full assessment
  • •Development plans cap at three active per employee — a fourth Create now returns a clear "limit reached" error instead of silently piling on plans
  • •Talent Market lets employees React on Published cards, notifies the managing recruiter, and surfaces a Reacted chip on the list and preview
  • •Talent Market Candidates block now shows the last-match timestamp with a one-click refresh for managers
  • •Assessment Results block is now visible only to the assessee on a Done assessment — peers and subordinates no longer see verdicts on assessments they merely participated in
  • •Mass Exam list gains a title-search + multi-status filter bar matching the Development plans page
  • •Mass Exam questions can be edited and removed in place while the exam is still in Draft
  • •Mass Exam Title, Description and Deadline can be edited inline on the detail page until the exam reaches a terminal status
  • •Employee profile Goals KPI tile now explains the formula in a (i) tooltip — average completion across open Development plans
  • •Development plan Progress Timeline captures a version snapshot when the plan moves to On review, and tooltips now render the human status label instead of a code
  • •Development plan email notifications cover the full lifecycle now — reviewer on submit, owner on return, both parties on completion, both parties on post-launch cancellation
  • •Talent Market removals on a Published card now email the dropped employee ("You are not considered more: [Title]"); removals on Draft cards remain silent
  • •Sign up to HRPulsar through a moderated request — verify your email, the team approves your tenant in Slack, and a one-shot magic-link logs you in with no password gymnastics
  • •Try HRPulsar before signing up — the public sandbox now opens straight onto a populated workspace (8 divisions, 22 competences, 40 employees, 7 assessment cycles, 4 mass exams, recruitment funnel with vacancies and AI-analyzed candidates) so every section lands on real content
  • •HRPulsar can be demoed locally — a self-hosted demo stack via make demo runs on isolated ports so a partner can spin up a populated tenant without touching production
  • •Sandbox visitors who like what they see can save their email straight from the demo banner — it becomes a moderated signup request the team reviews in Slack
  • •Personal data redaction now covers credit card, IBAN, US SSN, UK NIN, Spanish DNI and Italian Codice Fiscale on top of the existing identifiers
Changed
  • •Recruitment onboarding wizard is now a non-blocking sticky banner across /recruitment/* — recruiters can keep working while the stepper rides along and a Skip-tour button retires it for good
  • •Competence matrix single-grade view swaps the narrow table for a list layout with colored level chips, a salary pill in the section header, and a 32×32 delete column — easier to scan, faster to edit
  • •Talent Market match drawer now opens through a dedicated chevron icon, labels the spec row "Current position" when an employee qualifies through their present role, colours per-competence chips by the card threshold, and walks an employee viewer through a clean read-only view that hides every write affordance
  • •Talent Market Appoint candidate now requires a confirm-dialog click so a stray button press cannot pin a candidate; appointment and cancellation emails skip the manager copy when the appointee manages their own division
  • •Date picker across the platform — digit-only yyyy-mm-dd input mask, header buttons that switch the calendar to month or year pickers, and weeks now start on Sunday
  • •AI vacancy profile generation runs in the background with an ETA banner and a Cancel button — recruiters can leave the page and come back; the matrix modal lets them keep / drop per-competence rows, edit groups, competences, indicators and questions inline, and the AI prompt now incorporates Requirements / Responsibilities / Conditions plus any uploaded attachments and asks for ≥3 indicators and ≥3 interview questions per competence
  • •Vacancy detail Overview now shows Requirements / Responsibilities / Conditions blocks and edits them inline so the brief stays editable end-to-end
  • •Mass Exam lifecycle is unified across list, detail and Change-status — Draft → Sent → In progress → Completed with matching chips and per-status action buttons; the list groups active first then completed then cancelled
  • •Talent Market Candidates table and Add/Change picker now render the shared employee summary line — position under the name, status chip when the employee is not active — so an "on leave" candidate is visible everywhere their name appears
Fixed
  • •Recruitment AI analysis now keys assessments by the vacancy profile competence slug so the Compact matrix is populated for every interview-analyzed candidate; every analyze path bumps the interview version for consistent edit-conflict detection
  • •Invitations: Send invitation requires Division + Position before submit, and accepting a Manager invitation with a pre-selected division installs the new user as that division's manager
  • •Division edit: assigning Manager / Deputy Manager surfaces a "promoted to manager" toast so the auto-upgrade is visible without leaving the page
  • •Recruitment: Create vacancy redirects to the saved row immediately and runs profile generation in background — no more frozen Saving spinner
  • •Recruitment: Delete vacancy confirms with a single button instead of the type-the-title input — the type-to-confirm pattern was overkill for the action
  • •Recruitment Stage select on the candidates table now renders the stage name instead of the raw UUID; candidates can be removed from a vacancy directly in the list and the candidate-detail Back button returns to the vacancy that opened it
  • •Division detail and Add-employee dialog selectors now show the human title for Position, Grade and User instead of leaking UUIDs into the trigger label
  • •Compensation tab now accepts the Manager role on employees within the manager's division subtree (view / add / edit / delete) and returns a clear scope-violation error outside it
  • •Employee profile KPI tiles read correctly — Status "since" tracks the last status mutation, Tenure / "joined" anchors on first auth, the Assessments tile spells out N open, and Goals progress only counts non-terminal Development plans
  • •Development plan Return action now requires every item to carry at least one material (matching Send) and rejects past-deadline attempts with a clear server-side error
  • •Development plan Send button stays enabled past the deadline — the backend "Deadline is in the past" error now shows in a snackbar instead of being silently disabled in the UI
  • •Development plan Change-status modal: hovering the disabled In progress option exposes a tooltip explaining manual transition is not allowed
  • •Development plan materials seeded from a (specialization, grade) link now include only materials at or below the target skill level per competence — no more overshooting prep content
  • •Assessment Evaluation criteria sheet reopens with the previously saved Passing score for recommended grade instead of resetting to 75; the disabled In progress option shows a tooltip; the Change status trigger renders the picked status label
  • •Assessment Detailed results "Percent for Skill level" now reflects calibrated Totals, and Cancel calibration refreshes the block without a full page reload
  • •Mass Exam Change status menu now sends the correct field and hides itself on terminal Completed / Cancelled exams
  • •Mass Exam Send action rejects a deadline in the past with a clear error; the Details block highlights the deadline date in red once it passes
  • •Mass Exam Title is required on Create; the legacy "End date" label is now "Deadline" and a pencil control on the detail page edits it inline
  • •Mass Exam choice questions require ≥2 options and ≥1 correct (single choice exactly one) — the Add question button stays disabled until those hold, the Type field shows the human label, and single choice options behave like radios
  • •Mass Exam Pass Mark is now one-per-test — duplicates are rejected, a delete control replaces Add once one is configured, and removing it restores the empty placeholder
  • •Mass Exam Assign employees opens a searchable multi-select; the Results table links to the employee profile when the viewer can see it; assigned employees see the sent exam in their own /exams list
  • •Dictionaries: deleting a Specialization or Grade now blocks with a clear 409 (and the confirm dialog enumerates the references) when assessments, assessment groups, Development plans or exam pass marks point at it, instead of returning a generic database error

1.12.2

June 9, 2026
Fixed
  • •Recruiters can now open the Manager assessments section straight from a candidate card — create rounds, score competencies, and invite external evaluators without leaving the page

1.12.1

June 9, 2026

1.12.0

June 9, 2026
Added
  • •New canonical Recruitment flow — Add candidate modal with manual entry or bulk upload (up to 50 resumes × 10 MB per batch), live duplicate-email detection so an active candidate's email can't slip in twice, an Import preview that shows which files will create new records vs link to an existing one, and the candidate is dropped on the first non-terminal stage of the chosen vacancy
  • •Vacancy candidates table rebuilt around the canonical schema — 10-column layout with inline stage select, manager / AI score divergence flag, AI verdict popover with key strength / key risk / mitigation, and rows sorted with terminals at the bottom and highest manager score first
  • •Inline parsed-resume editor on the candidate card — each section (Summary, Skills, Experience, Education, Languages, Certificates) edits in place with optimistic concurrency so two recruiters can't silently overwrite each other
  • •Per-vacancy stage funnel — drag-and-drop Manage stages drawer (rename, reorder, pick the stage type, restore-default) with a guard that refuses to delete a stage that still has candidates on it
  • •Soft-deleted candidates can re-import the same email once their old record is archived — no more permanent dedup lock on a former candidate's address
Changed
  • •GDPR erase now also drops the underlying resume / interview media blobs from object storage so a Forget-Me request leaves nothing presignable, and blanks the canonical Candidate row alongside the linked Person
Fixed
  • •Candidate list shows externally-sourced candidates (CVs bulk-uploaded without a Person record behind them) and global search matches against both the canonical full name / email and the linked Person fields
  • •Stage moves on the vacancy candidates table refuse to silently clobber each other — two recruiters touching the same row get a clear conflict instead of a last-writer-wins outcome
  • •Bulk-upload modal poll no longer leaks another recruiter's in-flight batch into your vacancy's progress view
  • •Vacancy funnel save refuses an all-terminal funnel or a payload with duplicate stage codes — the form returns a precise error instead of persisting an unworkable lineup

1.11.1

June 5, 2026
Fixed
  • •The /blog page and its image proxy no longer hang when the upstream blog source is unreachable — every fetch tops out at 5 seconds and falls back to an empty section instead of blocking the page

1.11.0

June 5, 2026
Added
  • •Talent Market now emails everyone who needs to know — publish goes out as "Matched" or "Not matched" so candidates see at a glance whether they qualify, candidates added to a Published card get the same notice, an appointment notifies both the employee and their manager, Completed and Cancelled card states reach the matched / not-matched bucket, and every link drops you straight onto the card (signing in mid-click preserves the destination)
  • •Shared employee summary line — the same Name + grayed Current position + status chip rendering now appears wherever an employee is referenced outside their profile (Assessment lists, Development plans, Talent Market matches), so a status like "on leave" is visible everywhere their name is
  • •Custom English date picker replaces every native date input across the Employee profile and Add-employee dialogs — yyyy-mm-dd everywhere, per-dialog min / max bounds, no more browser-locale surprises
  • •Deleting a specialization now lists every position and grade chain currently pointing at it (by name) and locks the destructive button until those references are detached — no more silent cascades
Changed
  • •Talent Market Candidates table splits the legacy nominated bucket into Matched / Not matched so recruiters can scan a card and see which rows actually fit; auto-pool stamps Matched, manual picks compute their status from the current matcher, Appointed is preserved
  • •Talent Market Experience match falls back to the employee's current Position when no qualifying past experience exists — the Match cell shows a greyed "has experience" and the breakdown drawer labels the spec row "Current position"; longer tenured matches still win and min-experience floors still apply
  • •Talent Market visibility for Employees is scoped to cards they're a candidate on (Draft only when appointed); every write affordance — pencils, Match%, Add / Change candidate, Appoint, spec / competence editors, status transitions — is hidden, "(it's me)" follows the viewer's own row, and the match drawer arrow only appears for self
  • •Talent Market Appoint candidate requires a confirm-dialog click so a stray button press can't pin a candidate, and Required competencies render as inline wrap-to-fit chips so the card-detail page stays compact
  • •Talent Market match cell now opens the breakdown drawer through a dedicated chevron icon (chips are no longer the click target) and the icon stays active for the row whose drawer is open; the drawer's Required-competences row colours the level percent green / yellow / red the same way the employee Competences tab does and surfaces a separate Match / Below chip against the card threshold
  • •Match cell chips colour Competencies and Experience independently against their own thresholds — "no assessments" and "no experience" are now muted instead of red, and the copy switches to "no assessments" when the card has no Required specializations
  • •Development Plans pick up Assessment-style status names ("On review", "Done"), auto-promote from Sent to In progress on the owner's first item tick (manual transition is disabled in the Change Status dialog with a tooltip), tighten Returned so it can only go to Submit-for-review or Cancel, and the list + employee Development tab now sort active rows first by Created, then Done by completion date, then Cancelled by cancellation date
  • •Development Plan specialization and grade are editable only in Draft — pencils sit next to each value, the bottom Change button is gone, and the dialog mirrors the Jira flow ("Development specialization & grade" / "All items will be replaced with the new specialization & grade set"); saving rebuilds items + materials from the new (spec, grade) link
  • •Assessment lifecycle: manual Sent → In progress is no longer allowed — the auto-flip from the first submitted answer is the only path. Details page hides the In progress button on Sent, the Change status / Change status (all) modals grey the option with a "Manual transition is not allowed" tooltip, and the backend rejects single + bulk attempts with a dedicated reason
  • •Employee profile Experience tab enforces Company / Position / Start / End on Previous employment (past-only, Start ≤ End) and Start ≤ End on Current employment
  • •Employee profile Education tab requires Institution / Degree / Field of Study / Start, requires Title + past-only Completed date on Courses, and the Courses dialog now picks Competences developed via the same group-tree picker used by Assessments
Fixed
  • •Compensation payouts above the previous ~$21.4M ceiling (INT32_MAX cents) save instead of returning a 500 — payouts in JPY / IDR / COP or annual share grants now fit
  • •Editing a compensation record no longer 500s with a half-saved row — the audit event stores date fields as ISO strings into JSONB instead of raw datetime.date
  • •Compensation Add / Edit rejects an end date that lands before the effective date (422 + inline error) instead of saving an invalid range
  • •Talent Market Required Competencies are rebuilt from the union of current Required Specializations on edit (grade change) and removal — not just on initial add
  • •Talent Market card detail drops the duplicated "Cancelled: yyyy-mm-dd" row (the Status badge already names the terminal state), keeps the "Completed: yyyy-mm-dd" row, and renders "No description" on terminal cards while keeping the italic "No description yet" + pencil on active cards
  • •Talent Market Publish refuses cards whose End date has already slipped — the Draft → Published transition surfaces "End date is in the past" instead of failing the request later
  • •Development plans refuse to launch when the deadline already elapsed (Draft → Sent returns "Deadline is in the past" with a tooltip on the Send button), Review → Returned applies the same gate, the progress percent recomputes when admins add items after launch (no more stale ratios), passed items + their materials are read-only with the API returning 403 on bypass attempts, and item order is stable — new items always land at the bottom regardless of the value the client sent
  • •Assessment lifecycle emails (Evaluate yourself / Evaluate your employee / Evaluate employee / Your assessment has completed / Your employee assessment has completed) now deep-link to the assessment page instead of the generic list, and the auth middleware appends ?next= for unauthenticated visitors so the login form bounces back to the link after sign-in; cancellation also reaches pending peer / subordinate / external reviewers, and those reviewers now receive their own "Evaluation cancelled" copy instead of the manager template
  • •Detailed results calibration pre-fills the Total Select with the auto-computed answer, the trigger renders option titles instead of UUIDs, the Results table collapses Avg Score + Calibrated into a single calibration-aware Avg Score column, the per-level popup follows the calibration override, and Save only sends Totals that actually moved off baseline
  • •Evaluation criteria sheet reopens at the default Passing score for recommended grade (75) on every open instead of rehydrating the stored value — the chart still consumes the DB value, only the edit form resets
  • •Mass Assessment action menu "Sent" with every child past-deadline now shows the "Deadline is in the past" toast instead of a generic "No assessments updated"
  • •AI competence generation: the Refine request panel is no longer hidden behind the Indicators list's sticky counter and the Expand-to-detailed-form toggle no longer trips the "This page couldn't load" reconciler crash — the panel lives in its own DOM subtree below the suggestions list
  • •AI generation banner, per-target session badges, the competence-drawer status, and background task progress all keep polling regardless of WebSocket connectivity so finished / cancelled sessions update without requiring a page refresh
  • •Assessments list filters (search + Type + Status) are now applied server-side and reset paging to page 1, so the "X of N" counter and pager follow the filtered slice
  • •Assessments filter dropdown and Change status / Change status (all) modals display readable status names ("In progress", "On review", …) instead of raw codes
  • •Division detail page no longer crashes when opening the Add-employee dialog — the positions fetch unwraps the paginated envelope
  • •Specialization matrix save returns the freshly persisted snapshot — bulk upsert now mutates the relationship so the editor no longer renders an empty matrix until a refresh
  • •Recommended grade chart hides grades whose required skill level sits above what the assessment actually covers — a Middle-targeted run no longer paints a phantom Senior bar fed only by Basic + Intermediate answers
  • •Employee Competences tab keeps the latest lower-level assessment visible under Other competences when the employee's position is bumped above the assessed level — a Junior → Middle promotion no longer hides the existing Basic result
  • •The /blog HTML now goes through DOMPurify with an explicit allowed-URI scheme list, and the Confluence attachment proxy parses the requested path with a strict origin + pathname recheck — extra defence in depth on a public page that renders external content
  • •Login redirect rejects backslash-prefixed next= paths that some browsers normalise into a protocol-relative URL

1.10.0

June 3, 2026
Added
  • •Schedule interviews from a candidate card and attach the recording: chunked uploads up to 500 MB survive a refresh, audio / video / text-transcript all work, the in-app player has speed control, and an antivirus pass runs on every file before it becomes playable
  • •Each candidate now gets a tailored interview question set generated by AI from the vacancy competency profile: per-round regeneration skips what you've already covered and probes the blind spots, edit any question by hand or pull a fresh one from an indicator, export to PDF in three layouts (Compact / Full / Cards); sample mode keeps the feature usable when the balance is low
  • •Build your own evaluation scales (2–10 weighted levels) once per tenant, then bind a scale to each vacancy — the scale snapshots at the first score so live edits never rewrite a closed evaluation
  • •Run a structured manager evaluation per round (pre-interview, interview-N, final) with multiple evaluators scoring in parallel: divergence between sheets is highlighted, indicator scores roll up into the overall, autosave protects the draft, and the Manager score in the candidate table follows the last completed round
  • •Invite external evaluators by email — they open a dedicated link, accept consent, fill in the same form, and submit; tokens are SHA-256-hashed in the database, per-IP rate limits and a brute-force deny-list protect the public page, and expired or revoked links surface a friendly error page
  • •Division detail Employees block picks up Position and Grade filters that AND-combine with the existing Specialization plates: active filters become removable chips, the header carries an "X of Y" counter, and Clear all appears once 2+ filters are active
  • •Add an employee straight from a division: pick an existing person (with a cross-division confirm) or create a new one with the division pre-locked — no detour through the Employees list
  • •AI competence generation modal now lists every specialization + division with linked ones pre-checked, marked, and tooltip-explained; the result preview puts existing items side-by-side with new suggestions and blocks Save when nothing new came back
Changed
  • •Vacancy page collapses into a single inline-edit layout (Overview + Competences & indicators + Candidates) with Position-linked Specialization and Grade multi-select and a foldable Description
  • •Employee tables across positions, divisions, and specializations now share a single 7-column row (Name → Position → Specialization → Grade → Division → Status → Hire date) with sticky headers, tooltip-on-overflow, and a wider drill-down modal so nothing wraps
  • •Assigning or removing a division manager / deputy keeps the user role in sync: losing the last division you manage auto-downgrades you to Employee (Admin and Platform Admin are preserved)
  • •Generating positions with AI now drops you into the specialization AI-generate page with the position name in a banner and grade locked; an active-session badge lights up the position and specialization rows so you can find an in-flight matrix at a glance
  • •Generate learning materials with AI gets a full context picker: indicators per level, the full specializations + divisions lists with linked-row highlighting and search once the list crosses 15 rows, sibling competences, a company-description toggle, and a live materials/context summary
  • •AI matrix Divisions picker shows every tenant division with linked ones pre-checked, an "(N selected, K linked)" counter, plus search and Show-linked-only once there are 15+ rows
  • •Competence detail page is rebuilt with shared level colour tokens (Basic / Intermediate / Advanced / Expert) across Indicators and Materials, per-level counters and a "N materials for M indicators" gap warning, jump-to-section nav, a more visible Generate-with-AI button, suspicious-text warnings, click-to-edit Weight with a tooltip, and a Context badge with Reset on the Materials-by-context panel
Fixed
  • •Vacancy edit with Library refs no longer crashes on save (UUIDs are stringified before write)
  • •Hitting AI Generate while another session is in flight now opens a clear decision dialog (Open active / Cancel and try again) instead of silently bouncing you elsewhere
  • •The Generate button on Competences shows a Checking session… state while it asks the server, rechecks before opening the confirm dialog, and the drawer keeps polling for ~30s after Start so it catches sessions that land a beat late — no more empty "no active session" panel

1.9.5

May 31, 2026
Added
  • •Inline edit Match% on Draft Talent Market cards: a pencil affordance flips the value (50–100) without leaving the page; locks once you Publish
  • •Click any Match cell in the Talent Market Candidates list (or the Add/Change picker) to open a per-candidate breakdown drawer with one row per Required Competence and per Required Specialization
  • •Vacancies now carry a full lifecycle: kebab menu on every row + on the detail page, Edit / Archive / Restore / Delete dialogs, an Archived filter on the list, and a 412/428 guard against concurrent edits
  • •Assessment lifecycle notifications land in both email and the in-app bell: per-role Evaluate-yourself / Evaluate-your-employee / Evaluate-employee on Sent (and on a late-add), Your assessment has completed on Done, and matching cancellation copy on Sent / In Progress / On Review cancellations
  • •Detailed results gains a per-indicator Total calibration: reviewers pin a Total directly on each indicator and the per-competence percent / level inherit the override; a calibrated chip surfaces every pinned row; participants are locked out of edits while calibration is open
  • •Create / Edit Vacancy adds optional Library refs — multi-select pickers for Positions, Specializations and Divisions — that seed the vacancy competence matrix from your company grade ladders
  • •Vacancies accept manual Requirements / Responsibilities / Conditions text plus per-vacancy file attachments (PDF / DOCX / PPTX / XLSX / TXT / MD / CSV / JSON / images / ZIP / EPUB, 25 MB / 10 files); the Profile tab carries an upload widget that feeds the AI prompt
  • •Vacancy → Candidates gains a Quick candidates section: separate Recruiter and AI scores, a Δ-marker when they diverge, source / experience / current position fields, and an inline Add candidate dialog — no full Person flow needed
  • •Vacancy competence list — manual library picks, library inheritance, or AI-proposed — with a replace-set save flow
Changed
  • •Talent Market card detail hides the Dates row in Details on Completed / Cancelled cards — the terminal-date row right below already carries the timestamp
  • •Required Specializations block on Talent Market cards is now plural and asks for an explicit confirmation when you add a second spec, edit an existing one, or delete one — because Required Competencies are recomputed on save
  • •Talent Market Match cell renders real Competencies % and Experience years with green / orange / red / "no experience" colour rules, and the pool plus candidate list rank qualifying-first then comp-only then exp-only
  • •Recommended grade chart appears at On Review (was Done only), supports Employees' current positions and Specialization-All-grades, and per-grade match % is the mean of competence percents across the levels up to the required one
  • •Detailed results hides skill levels above each competence's required level so only the indicators actually evaluated render
  • •Criteria sheet drops the Passing score block when criteria is Individual competences, labels it Passing score for recommended grade for Target / Current positions, and the (i) tooltip + formula toggle are gone
  • •Active assessment cap per assessee lowers from 5 to 3; Mass assessment does a partial creation when only some employees are at the cap and rejects the launch as a conflict when every employee is full
  • •Draft → Sent transition with an elapsed deadline returns 409 instead of 400 — the request itself is valid, the stored deadline is the conflict
  • •Grouped Assessments list sorts active rows first by Created, then Done by Completed, then Cancelled by Cancelled date — matching the single-list ordering
  • •The Draft → Sent button label now reads send (imperative) instead of sent
Fixed
  • •Talent Market matcher projects assessment results to the required skill level using the per-level breakdown when a Done assessment ran at a higher level
  • •AI vacancy profile generation runs inline (no fire-and-forget Celery dependency) so the UI returns the saved profile or a real error
  • •AI profile generation tracks an active session and the vacancy detail page polls it — a returning user sees Generating… or Last generation failed without re-clicking
  • •Vacancy save no longer surfaces a spurious If-Match required toast (the cross-origin browser couldn't read the ETag; we now expose it)
  • •Delete vacancy from the list now refreshes the rows in place instead of leaving the deleted row visible until reload

1.9.3

May 23, 2026
Added
  • •AI matrix generation gets a new preflight context picker — positions linked to the specialization, tenant divisions, the company description, and existing matrix competences render as chips with × controls so you can trim what the LLM sees before generating
  • •Generate learning materials with AI from any competence: pick the skill levels, optional specializations, and how many ideas per level, then review the LLM's proposal and bulk-save the entries you keep
  • •Add multiple indicators at once on every skill-level block of a competence — one line per indicator, saved in a single transactional bulk insert
  • •Development Plans list grows the same filter bar as Assessments — title search plus a multi-select Status with a Clear button
  • •Specialization-All-grades assessments now render a per-grade match bar chart in the Results block from On Review onwards; the top-scoring grade(s) are highlighted as recommended
  • •Talent Market card Details block surfaces status transition commands per status: Draft → Publish, Published → Complete or Cancel; terminal statuses show nothing; Publish now also requires at least one candidate
  • •Detailed Results block on the assessment page surfaces every competence's required level, per-skill-level percent breakdown, and per-indicator role answers with author-tagged comments and a Total row
  • •Surveys now open in a side sheet with vertical answer options and a per-indicator comment field; answers autosave on every change and rehydrate when you reopen the sheet
Changed
  • •Date pickers and date display switch to English locale with yyyy-mm-dd format across the app
  • •AI competence library: group-level AI generation now offers an "Augment existing items in the group" toggle — keep fresh-only behaviour off, or extend existing subgroups with new competences and existing competences with new indicators (matched nodes render locked in the review tree)
  • •Assessment detail page swaps the Deadline label for "Completed dd.mm.yyyy" or "Cancelled dd.mm.yyyy" on terminal assessments, and renders an overdue Deadline in red on active assessments
  • •Talent Market list view always shows the action menu (no hover gate), hides it on Completed / Cancelled cards, collapses to Change-status + Delete with capitalised labels; the detail page gains pencil-icon inline edit for Title / Description / Dates on non-terminal cards
  • •Position detail page's Competences & indicators block renders as a compact two-level tree (group → competence with the required level badge); indicators sit behind a per-competence expander instead of a flat list
  • •Headcount drilldown modal on the Positions page widens and scrolls inside so every column stays legible
  • •Employee Experience tab's first block renamed Current Employment, captures Division/Position/Start with defaults from the card, and renders an elapsed-duration badge (days / months / years)
  • •Employee Competences tab rebuilt as two blocks — Current position competences (tree from the grade-specialization, % from the latest qualifying Done assessment) and Other competences — replacing the opaque radar chart
Fixed
  • •AI-generated learning materials now consume credits — the LLM round-trip is no longer free on Enterprise plans
  • •Competence overview on the employee page no longer fires per-assessment SQL: a single batch query handles every Done assessment regardless of count
  • •Stuck AI competence-generation sessions are auto-reaped after 15 minutes so an OOM-killed worker can never block a user from starting another generation
  • •LLM refinement input is stripped and sanitised against role-tag injections and "ignore previous"-style triggers before reaching the model
  • •AI competence generation "active session already exists" 409 now carries the live session id and scope, a global banner surfaces in-flight sessions from every page, and worker crashes flip the row to error instead of leaving it stuck as running
  • •PDP item checkboxes now toggle both ways for the plan owner in Sent / In progress / Returned, and for the division-manager reviewer (implicit reviewer when no reviewer is set) while the plan is Under review
  • •Talent Market Candidates block now auto-fills from the card's requirements — the last Done assessment per Required Competence at the matching skill level, average percent versus the card-level Match %, plus a Required Specialization position check
  • •Talent Market Candidates block renders the employee name as a link to the profile instead of the raw UUID
  • •Division detail page now lists assigned employees and the Edit dialog renders Manager / Deputy names instead of raw UUIDs
  • •Preliminary assessment results now recompute when a new respondent submits after On Review so Avg Score reflects the freshest answers while calibration overrides keep Percent and Level
  • •Manage grades dialog on the Specialization page now toggles already-attached grades (untick to detach) and is also reachable from the Company → Specializations row menu
  • •Specialization matrix with a single grade now uses the full table width, the indicator dialog groups levels with bullet markers and per-level scroll, competence titles open in a new tab, and tooltips snap in at 100 ms
  • •Add-competences dialog on the specialization matrix gains per-branch chevrons plus Expand all / Collapse all controls so deep trees stay manageable

1.9.2

May 22, 2026
Added
  • •Talent Cards now carry a card-level Match % (50–100) that drives the Required Competences matcher threshold across the card — editable until you publish, locked afterwards
  • •Specializations can be deleted from the Company → Specializations list and the specialization detail page; the existing guards against linked positions and grade chains still protect you from cascading damage
  • •Bulk-add competences to a group: pick multiple rows at once from a tenant-group dropdown and attach them in one go
  • •Assessment Results table shows each competence's required skill level inline; a new Results by level popup breaks the percentage down by Basic / Intermediate / Advanced
  • •PDP plans in terminal statuses now show the closing date — "Completed at" / "Cancelled at" — instead of the original deadline, in both the list preview and the detail page
  • •Add-material dialog: Format is a 17-option select, Type is a required 3-option select with icons (Theoretical / Practical / Feedback); each material row now displays the type icon
  • •Edit-employee dialog lets admins and managers rename a person — Name and Last name in a single update
Changed
  • •Required Competences picker rewritten as a tree with search and group-level select-all; existing rows are pre-selected and the action button reads Change once the block has entries, replacing the whole set in one save
  • •Activating a competence group now cascades both up the parent chain and down to every descendant; new children inside an inactive group default to inactive; Delete is hidden once the group is referenced by matrices, assessments, or development plans
  • •Edit-group dialog drops the activate-by-default checkbox and shows a read-only «Used by client» affordance with a tooltip when the group is in use elsewhere
  • •AI generation preflight modals now expose per-item context chips: for whole-base you can drop individual specializations, divisions, or existing tree nodes from the prompt; for group and indicators you can drop related specializations, ancestor groups, descendant subgroups, or sibling competences
  • •AI competence and indicator generation now requires at least 3 indicators per skill level and ships explicit per-level calibration guidance
  • •Evaluation criteria summary collapses Specialization + Grade into the Type line for target-position assessments, drops the auto-pick hint on single assessments, and hides the aggregated Competences list on mass-assessment parents that use current positions
  • •PDP items in Under review can be toggled by admin, platform admin, or the assigned reviewer; the Return button disables once every item is accepted
Fixed
  • •Required Specialization rows no longer render "Min. experience: 0 year(s)" when the value is 0; deleting a row now asks for explicit confirmation
  • •Mass-assessment bulk-launch surfaces a clear "Deadline is in the past" error when every child fails the deadline guard, instead of a misleading "Updated 0 of N" success toast
  • •Editing a Work Experience or Course/Certification entry no longer fails when Competences developed keeps an existing competence — the link sync diffs the set instead of recreating every row
  • •Division managers see PDP plans in Draft within their managed subtree, matching the visibility admins already had when authoring plans
  • •PDP item completion is one-way for the plan owner: once a checkbox is ticked it locks (admin/reviewer keep full toggle in Review)
  • •AI competence generation no longer fails with "AI returned an invalid response" or a false "Background worker offline" banner during long runs
  • •Competence cascade operations are tenant-scoped end-to-end, so a stray cross-tenant node can never be silently activated or deactivated through the tree
  • •PDP plan owners can submit a Returned plan back for Review and toggle item checkboxes in Sent / In progress / Returned
  • •PDP material modal aborts an in-flight file upload when you click Cancel or outside the dialog, so the next open is clean instead of stuck on "Uploading…"
  • •Assessment calibration now refreshes Percent, Level and the overall result from the calibrated score and keeps the override across the on-review → done transition
  • •Add-indicator and Add-material dialogs show the selected skill-level name instead of the raw value
  • •Competence page «Materials by context» dropdown shows the selected specialization name
  • •Employees list search hits the backend and finds matches across every page, not just the visible one
  • •Recruitment AI profile generation works again — backend route now matches the frontend callers

1.9.0

May 19, 2026
Added
  • •AI competence generation preflight: a new dialog lets you drop context categories before the LLM call — Specializations, Divisions, Company description, the existing competence library, sibling competences, or existing indicators — and attach a free-text refinement note. Available for whole-base, group, and per-competence indicator generation
  • •Specialization-matrix brief grows a "Generate indicators for existing competences" opt-in; matrix sessions emit a deduped "ready" toast with an Open review action so background work doesn't get lost
  • •Expand all / Collapse all on the Competence DB tree and the Assessment competence picker — no more clicking through dozens of groups one by one
  • •Assessment participants are clickable: when the viewer is allowed to open a teammate's profile, the participant name becomes a link to that employee's page; otherwise it stays plain text
  • •Employees can now keep their own Education and Courses entries up to date without HR babysitting — every other field on their profile stays admin/manager-only
  • •Division detail page shows employee count per specialization (click to filter the Employees block) and gains an inline Edit dialog for name, description, parent division, manager, and deputy manager
  • •Talent Card create/edit asks for Start date (required, default today) and End date (optional); list preview and detail page render the period in plain language ("since DD.MM.YYYY", "DD.MM.YYYY – DD.MM.YYYY", "Closed: DD.MM.YYYY")
Changed
  • •Assessment status workflow exposes the On Review checkpoint everywhere — list-page Change Status modal, the In Progress action button (Done → On Review), and the On Review action surfacing Done / Cancelled
  • •Titles and descriptions across Assessment, Mass Assessment, PDP, Talent Card, and Mass Exam forms now enforce sensible character limits (100 / 250) on both sides
  • •Assessments and Talent Market list filters for Types and Statuses became multi-select dropdowns — pick any combination instead of being locked to one value at a time
  • •Talent Market list grows the same filter block as Assessments: title search plus Clear
  • •Positions list surfaces AI-generated drafts in a dedicated card at the top so they get reviewed before the approved-positions table
Fixed
  • •Dashboard pages stay browser-translatable again — the auto-translate guard is now scoped to the AI generation drawer only, instead of the entire dashboard
  • •AI generation drawer no longer flashes a false "worker offline" banner during long LLM calls — it waits for three consecutive health-check failures (~45 s) before raising the alarm
  • •AI Generate matrix Apply now surfaces a warning when nothing was actually persisted, instead of a misleading "Matrix applied" toast
  • •Specialization matrix UI polish: salary becomes a bordered pill, grade removal goes through a confirm dialog, cells tint by skill-level rank as a heat map, and competence rows missing indicators or development materials at assigned levels are flagged
  • •Specialization matrix gets a proper Add-competences modal with the full group tree, group-level cascade checkboxes, breadcrumb search, and a per-row link to the competence detail page — replacing the legacy inline dropdown
  • •Launching an assessment refuses an already-elapsed deadline with a clear error so it stays in Draft until the operator extends the date; bulk launch lists which assessments were skipped and why
  • •Positions list status menu respects the lifecycle: a closed position only offers "Reopen → Active" instead of every other status
  • •Inline-edit fields on entity detail pages — assessment title and deadline, position title, headcount, and description — now commit on Enter
  • •Talent Market Add candidate: searchable employee picker showing the computed match score and a matched / not matched status, with bulk attach, replaces the old UUID input
  • •Assessment results endpoint now respects the same access scope as the rest of the assessment views — admins and participants are unaffected

1.8.2

May 18, 2026
Added
  • •Specialization-matrix cells deep-link straight to the competence detail page; coming back via "Back to matrix" preserves spec and grade context, AI indicator sessions launched from there reuse the spec / grades in the prompt, and any indicator change on a referenced competence prompts you before saving or before matrix Apply
  • •AI matrix brief form (specialization page and Position AI card) persists your typed brief locally so a refresh or remount doesn't lose your work; a Clear draft button wipes it on demand
Changed
  • •Position pages: list shows clickable specialization and grade, with inline "Set specialization" / "Set grade" CTAs when missing; the Overview block grows its own Edit button; the Competences & indicators block surfaces a state-aware CTA instead of a generic empty hint; deep-links between positions, specializations, and matrices preserve breadcrumb context
Fixed
  • •AI generation regenerate / refine no longer fails with a server error when a parallel session is already running — the conflict resolves to a clean message
  • •AI generation refinement panel re-hydrates from the last submitted form when you reopen the drawer, refresh the page, or hit Try again; Clear data wipes both the form snapshot and the composed prompt
  • •Participant completion now mirrors the questionnaire's skill-level cascade — answering every visible question actually flips you to Completed=Yes even when the assessment uses skill-level scoping
  • •Specialization-matrix AI generation no longer burns retries on a degenerate prompt: it short-circuits to a clear "insufficient data" message when the spec has no grades, and realistic matrices now fit (LLM context bumped)
  • •AI generation failure messages now include the underlying error type and a short detail snippet so the cause is distinguishable from the UI
  • •Position page: when AI Generate runs but the position has no grade set, the success toast points you to the specialization page (the only place the saved matrix is visible), and the matrix-empty state surfaces a "set a grade or view on specialization" hint
  • •Live AI generation sessions (your own and other tenant admins') light up the affected group, competence, specialization, or position row with a clickable badge that opens the drawer; AI Apply routes you back to the entity it just populated
  • •Competence detail page renames the AI button to "Augment indicators (AI)" when at least one indicator already exists, and indicator generation no longer errors out on specs without grades
  • •AI drawer stops surfacing a false "worker offline" banner while a worker is busy with a long generation
  • •Auto-translation extensions (Google Translate, etc.) no longer break the dashboard — the guard is scoped properly

1.8.1

May 15, 2026
Changed
  • •Position detail page now supports inline editing for Title, Description, Headcount, and Status, with one-click links from Specialization and Grade in Overview straight to the matrix
  • •Assessment detail tightens its action buttons: "Preview questions" lives in the Details block as soon as criteria and scale are picked, and "Take this assessment" is hidden in Draft so reviewers don't try to evaluate before the assessment is published
Fixed
  • •Regular employees can no longer see other people's Draft assessments — list, grouped list, and lazy-loaded group cards now respect the same access scope as the single-assessment view

1.8.0

May 14, 2026
Fixed
  • •Recruitment vacancy detail page renders correctly for fresh tenants — the candidates section no longer breaks into a server-error screen
  • •Public consent links and shared report links work for recipients who aren't signed in — those routes are properly anonymous instead of bouncing visitors to the login page
  • •Creating a new vacancy no longer kicks you back to the login screen after saving — a hidden background-fetch race that mistook a normal navigation for a logout is fixed

1.7.5

May 13, 2026
Added
  • •Talent Market cards now collect Requirements as two structured blocks instead of free-text notes: Required Specialization (specialization + grade + optional minimum years) and Required Competencies (competencies with skill levels and a shared % match). Picking a specialization auto-fills its competences from your career ladder so you don't re-type the matrix
  • •AI-generation drawer for competence indicators gains an explicit Cancel button with a confirm step, so discarding generated suggestions is a deliberate choice
  • •AI-generation drawer header turns the target competence title into a link back to its detail page — one click and you're there
  • •Bulk select / deselect for generated indicators inside the AI-generation drawer
Changed
  • •"Add to library" inside the AI-generation drawer now works for single-competence indicator sessions (previously the action buttons stayed disabled). Choose to publish indicators immediately or add them as drafts, and the drawer takes you straight to the updated competence page
  • •"Generate indicators (AI)" button on a competence page now stays clickable while another AI session is in progress: it shows the live status and opens the active session drawer, with a tooltip explaining why a new generation has to wait
  • •Talent Market cards can only be published once at least one Required Competence is set; published cards lock both Required blocks to keep candidate matching stable

1.7.4

May 13, 2026
Fixed
  • •Multi-select filters across the app (Divisions / Statuses / Positions / Specializations / Grades) now record every consecutive pick — earlier selections are no longer silently dropped when you tick a second value

1.7.3

May 13, 2026
Added
  • •Specialization page now exposes a full cold-start path for a brand-new specialization: an Add Grades modal picks grades from the dictionary, matrix columns are drag-and-drop sortable, and salary ranges edit inline directly in the column header
  • •AI Generate matrix brief now asks for the grades you want covered and pre-creates the Specialization × Grade pairs from the brief, so a blank specialization can be configured end-to-end via AI in one pass
  • •Position AI Generate prefills the specialization's grades and locks the position's own grade as required, so applying always lands a matrix the Position page can render
Changed
  • •Matrix columns now follow the order you set on the Specialization page (not the global dictionary order) — drag-and-drop reorder is what every page reads back
Fixed
  • •Multi-select filters on the employees list keep the dropdown open between picks so you can select several values in one go instead of having to reopen after each tick

1.7.2

May 12, 2026
Fixed
  • •Admins linked to a plan's employee see the full Development Plan transition graph again — single-tenant demo accounts and the e2e lifecycle walk no longer get stuck on Draft

1.7.1

May 12, 2026
Added
  • •Assessment results now show an overall percent next to the Results heading on completed assessments, computed as the average of per-competence percents
  • •Assessment lifecycle gets an On Review step between In Progress and Done so calibrators can adjust scores before the assessment is finalised; the system auto-enters On Review once every participant has answered
  • •Company logo card supports drag-and-drop and a paste-by-URL flow, with the right MIME type sniffed from the bytes so images render correctly after upload
  • •New Position dialog now lets you add a Specialization or Grade inline — no more leaving the form to seed a dictionary value
Changed
  • •Development Plan workflow simplified to Draft → Sent → In Progress → Review → Done with Returned and Cancelled side states; the Awaiting approval / Approved hops are gone, and an overdue plan is now a red-deadline cue rather than a separate status
  • •Development Plan detail page now shows the assigned employee a single "submit for review" action that activates once every item is ticked
  • •Item checkboxes on a Development Plan can only be ticked by the assigned employee; the first tick on a Sent plan auto-promotes it to In progress
Fixed
  • •Development Plan list rows show the assigned employee, hide the action menu on completed plans, and highlight past-due deadlines in red
  • •Development Plan attachments open inline again — file links now use a fresh signed URL that doesn't expire mid-session
  • •Development Plans cannot be sent until every item has at least one material — the Send button stays disabled with a tooltip until the plan is complete
  • •Development Plans in Done or Cancelled are fully read-only: no new comments, no item check-offs, no editing
  • •Regular employees and managers no longer see Draft Development Plans in the list — those stay in the admin authoring queue
  • •Plans entering Review freeze the items+materials block for the assigned employee so admin edits during review don't disturb the snapshot
  • •Deadline pickers on Assessment, Mass assessment, Development Plan and Mass exam refuse dates in the past, both in the form and on the server
  • •AI Generate matrix picker shows a styled "Choose files" button with a clear file-count counter instead of the un-localized browser file input
  • •Assessment per-competence percent now follows the spec algorithm (per-role indicator average → weighted level percent → mean across skill levels → mean across participating roles); existing Done assessments are recomputed in place
  • •Deleting an employee with timeline events no longer crashes — the events are cascaded out cleanly

1.7.0

May 10, 2026
Added
  • •Position pages now inherit salary range and competence matrix from the (Specialization × Grade) pair — the detail page shows a live matrix banner with a deep-link back to Specialization when the pair is missing a profile, the headcount line gets a vacancies pill, and the Overview block adds a salary row plus a ●●●○○ headcount visual
  • •One unified Edit dialog drives both Position creation and editing — pick a Specialization, the Grade list cascades, the Title auto-fills with Spec Grade, and a salary preview from the chosen pair lets you sanity-check before saving. Editing a position with assigned employees raises a clear impact warning if the profile changes
  • •Positions list gets three new filters — lifecycle status, "has vacancies" (positions with open headcount), "matrix unconfigured" (Spec × Grade pair without competences) — alongside a Matrix column with a ✓ / ⚠ icon and a client-side group-by toggle (None / Division / Specialization)
  • •Specialization detail page reorganises into four tabs — Grades & matrix / Positions / Employees / AI history — with shareable URLs. Employees tab aggregates everyone in any Position of the specialization, in the same row format as the Position drill-down
  • •Specialization Grades tab bundles the competence matrix and per-grade attributes into a single Builder so operators no longer hop between sub-pages
  • •AI competence generation gets a tri-state per-suggestion review (Accept / Edit / Reject) with a pending-count banner, Accept-all and Reject-all bulk actions, and inline title rename — your renames flow through to the materialised competence tree on apply
  • •AI Generate page replaces the static "Status: running" block with a live streaming-progress checklist (Thinking → Grades → Competences → Indicators 1/N → Matrix) driven by WebSocket events
  • •AI history tab on Specialization detail lazy-loads recent generation sessions with initiator, brief, counts and status, plus a one-click «Repeat» that spawns a child session — Repeat now works on applied and cancelled sessions, not just successful ones
Fixed
  • •Deleting a Specialization or Grade that is still in use on a Position or grade chain now returns a clear conflict with the affected counts, instead of silently orphaning the position profile or cascading the matrix away
  • •Deleting an employee with no assessments, PDPs, exams or talent records — and who isn't a division manager — now hard-deletes the row instead of silently flipping the status to terminated; if any link exists, the API replies with a clear conflict naming the employee
  • •AI Generate matrix page (Specialization + Position) no longer fails with "Not authenticated" — the multipart upload now flows through the shared API client
  • •AI history tab is now visible to any admin in the company, so the team log is shared across operators (mutating actions still require the original session author)
  • •Live AI generation progress shows a "paused — refreshing" hint when the WebSocket disconnects, so you always know the page is still tracking the run
  • •AI review modal warns before discarding unsaved title edits, so a misclick on Cancel can't wipe out renames you typed
  • •Position edit dialog hardens the cascading Specialization → Grade fetch — a fast switch back-and-forth between specs no longer commits stale grade options

1.6.3

May 8, 2026
Added
  • •Generate indicators with AI directly from the competence detail page — a single button kicks off a tailored AI session for that competence, uses your existing context (group, description, specializations, grades) and shows status inline (in progress / open active session / busy elsewhere)
  • •Mass Assessment is now renameable in one click — the title cascades to every child assessment in the group, and per-child rename is locked while a parent group exists
  • •Assessment detail exposes a "Preview questions" view that mirrors exactly what reviewers will see, including the per-competence skill level
  • •Tooltip primitives wired across the dashboard for consistent hover hints
Fixed
  • •Skill-level, specialization, grade and material pickers across the competence detail and assessment-group pages now show the selected entry's title in the dropdown trigger instead of leaking a UUID or __base__ literal
  • •Starting an AI competence generation session no longer fails with a 500 — the response is correctly serialised after the celery task is enqueued
  • •Single + Mass Assessment lock the Evaluation criteria and Rating scale once they leave draft, and the Evaluation criteria affordance is renamed Edit → Change for consistency
  • •Adding a participant to an assessment now rejects the same human in any role with a clear conflict, and the picker hides everyone already in the assessment
  • •A new assessment is rejected when the assessee already has 5 active ones — no more drowned queues
  • •Assessment detail returns 404 to restricted callers who are neither in the assessee's scope nor a participant — closes the URL-fishing leak
  • •Participant questionnaire only shows indicators of the picked skill level and lower for Individual-competences criteria
  • •AI competence generation no longer fails on the worker with Received unregistered task of type; cancelling a session mid-flight now revokes the queued/running Celery task and skips billing for the cancelled run
  • •Competence tree group/competence dropdown menus stop wrapping labels, and the competence edit dialog widens to fit the indicators block
  • •Every AI-generation entry point on /competences (group menu, competence menu, indicators editor) shows a tooltip explaining the active-session lock instead of relying on a native title that browsers suppress on disabled buttons
  • •Competence tree legend reads competences: N next to group titles, and the published / not-published / origin / in-use icons each carry an explanatory tooltip
  • •Rating scale options no longer leak empty () for the neutral answer; admin view keeps weights, participant view drops them, and the neutral option always reads (not counted)
  • •Bulk status change with "Apply to all → Cancelled" leaves already-completed assessments alone and reports them under skipped reasons instead of silently overwriting them
  • •Evaluation criteria sheet: specialization picker lists every active dictionary item instead of only those wired into a grade chain; assessment summary shows the chosen skill level next to each Individual competence

1.6.2

May 7, 2026
Added
  • •AI Generate on position detail page — launch a position-scoped competence generation that reuses the specialization-matrix flow; applied competences land in a single group named after the position
Fixed
  • •Mass Assessment now lets you pick a Rating scale at the group level — the chosen scale propagates to every draft child in one click; locked once any assessment in the group leaves draft so launched scales stay frozen

1.6.1

May 7, 2026
Added
  • •Position detail page (/company/positions/{id}) — resolved competence matrix and the assigned employees rendered through the unified Name / Position (Specialization · Grade) / Division / Status / Hire Date columns
Fixed
  • •Assessment detail keeps the rating scale visible after the assessment has been Sent — the snapshotted scale is embedded in the response so reviewers always see the right options

1.6.0

May 6, 2026
Added
  • •AI competence framework generator — describe your business, drop in role/spec documents, and generate a full competence tree with cascading checkboxes, refinement prompts, regenerate-and-compare, and a preview before applying. Works at three scopes: the entire base, a single group, or just the indicators of one competence
  • •AI specialization matrix generator — upload role descriptions (PDF/DOCX/XLSX/RTF/TXT, up to 10 MB per file) and produce a full Competence × Grade matrix in one pass, with the same refine / regenerate / preview-before-apply flow
  • •Specializations as first-class entities — list, detail, grades with descriptions, requirements, salary range and currency; positions grouped by division with assigned/plan and lifecycle status; shared Company tabs (Overview / Positions / Specializations)
  • •Competence × Grade matrix editor with dirty-cell highlighting, indicator tooltip (showing every indicator up to the selected level) and an unsaved-changes bar; the editor cascades changes across grades automatically
  • •Position lifecycle and occupancy — Active / On-hold / Frozen / Closed badges with an on-row status menu, Filled / Plan drill-down with a single highest-priority alert per employee, an "Unassigned employees" admin view, and a "matrix not configured" banner that deep-links into the Spec × Grade editor
  • •Material specialization overrides — competence detail page exposes a "Materials by context" panel where admins can hide or add learning materials per specialization; Development plans automatically pick up the right material set based on the employee's specialization
  • •Per-workspace AI settings page (/settings/ai) — pick effort tier (Fast / Balanced / Thorough), Provider + Model, sampling temperature, retry budget, content language, and company context that's merged into every AI prompt; deep-linked from every AI generation flow
  • •Cost confirmation dialog — any AI action that meets your workspace's credit warning threshold now requires explicit confirmation with the cost shown up-front; the "Confirm & start (N credits)" button leaves no room for surprise
  • •Generation context panel inside the AI confirm dialog — shows the filtered specializations / divisions chips, the company description preview, and direct links to fix any missing context (dictionary, company profile, AI settings) before starting
  • •Real-time updates — in-app notifications and AI generation status now update over a WebSocket channel with automatic reconnect and a polling fallback; long-running tasks broadcast started / done / failed live, no page refresh needed
  • •"Cancel generation" mid-flight — cancel a running AI session and the status returns to idle immediately, no waiting for the next polling tick; a Background worker offline warning shows up when a session is queued and the worker isn't responding
  • •Inline indicators editor on the competence edit dialog — list, group by skill level, add / edit / delete from a side sheet, with an inline help tooltip and a deep-link into AI generation
  • •Competence detail page with publish / unpublish / hide controls and level-scoped indicator and material editors with drag-and-drop reordering; two-step wizard for creating new competences
  • •Drag-and-drop in the competence tree to move groups and competences between branches
  • •Inline edit for pending invitations — admins can rename invitees, change division/position, or rotate the email and re-send (rate-limited, audited); Name is now required and a Division column is shown in the invitations table
  • •Inline edit for assessments — title and deadline can be changed in place (admins/managers, blocked on completed assessments)
  • •Bulk status change on the assessments list now reports skipped reasons broken down by category (terminal / same-or-lower status / already cancelled / missing criteria or scale) instead of a single "skipped N" toast
  • •Employees list redesign — combined Name+Email column, Specialization · Grade column, multi-select filters for Divisions / Statuses / Positions / Specializations / Grades, and manager-scoped division pickers
  • •Company profile page (/company/profile) — admins manage logo, description, industry, size and website in one place
Improved
  • •AI competence-generation confirm dialog displays the credit cost up-front and adapts its start button when the cost meets the warning threshold
  • •All AI generation tasks (competences, indicators, PDP goals, positions, specialization matrix) now honour your workspace AI settings — language, company context, model and temperature
  • •Development plan (PDP) detail page now allows renaming, reordering and deleting items and materials inline, picks the right material format from a select, and surfaces the reviewer (or falls back to the employee's division manager)
  • •API error toasts now format FastAPI 422 validation payloads as readable lists instead of rendering "[object Object]"
  • •PDP draft → sent transition now emails the assigned employee with the plan title, deadline, and a direct link
  • •RBAC: assigning someone as a division manager auto-upgrades their role; removing them prompts to confirm the downgrade; invitation roles are capped at the inviter's level
Fixed
  • •Item checkboxes on draft PDPs are no longer interactive — completion can only be marked once the plan is sent
  • •PDP draft → sent email now uses the plan title in the subject and heading instead of a generic placeholder
  • •AI competence generation tooltip "Edit company profile" link now opens the right page (was pointing at user settings)
  • •draft → sent transition on assessments now requires both evaluation criteria and a rating scale, with a clear precondition error instead of a silent no-op
  • •Database connections recover faster from stuck "idle in transaction" backends — every engine now has a 5-minute idle-transaction timeout, and migrations fail fast on contended locks instead of queuing indefinitely
Security
  • •Terminated and inactive employees are blocked from authenticating, status changes trigger an email to the user, and the profile-update endpoint enforces a strict field allowlist
  • •Managers can only edit employees inside their division scope; regular employees can no longer edit profiles via the employees API
  • •Cross-tenant scan in the employee-alerts engine is no longer possible — the bulk computation now requires an explicit tenant id and short-circuits any off-tenant rows in a mixed cohort

1.5.0

April 30, 2026
Added
  • •Custom rating scales — HR can now create, edit and soft-delete answer scales with 2–10 scoring options, an optional "I don't know" neutral choice, and 0–100 percent levels. Each assessment can pick its own scale
  • •Rating scale picker on the assessment detail page — empty block ships an "Add scale" button (and "Change" when one is assigned), with drag-and-drop ordering of options, inline preview, and edit/delete confirmation dialogs
  • •Snapshot-on-launch for rating scales — once an assessment leaves draft, its scale is frozen so subsequent edits never bleed into running assessments
  • •Evaluation criteria selection sheet on single and mass assessments — pick by current positions, target position, or hand-pick competences from the tree, with cascading checkboxes, search, and per-competence skill levels. Mass criteria propagate to every active child assessment in the cycle
  • •Assessment results now expose a percent score and a resolved level per competence on completion (neutral answers excluded). The Results table shows Percent / Level columns with a description tooltip
  • •Passing score per grade in the specialization dictionary (default 75%) and per assessment criteria. On completion, an automatic grade recommendation is produced (% match per grade with hierarchy + "not confirmed" fallback) and surfaced via the API
  • •PDP plans now carry a title with inline edit, auto-generate items from grade-specialization competences, and surface employee + formatted deadline. Specialization/grade can be changed after creation, with auto-items recomputed while passed and custom items are preserved
Improved
  • •AI generation (positions, competences, indicators, PDP suggestions) and AI wizards on Positions and Competences are now asynchronous by default — the request returns immediately with a task id, and the UI polls for the result. No more page-blocking spinners on long generations
  • •Outgoing emails (waitlist confirmations, notifications, external-reviewer invites) now go through a background queue instead of blocking the originating request
  • •PDP detail page status flow aligned with the backend lifecycle: sent / in progress / review / on approval / approved / done / returned / cancelled / expired
  • •Cloud pricing repriced and restructured into four credit packages (Starter / Growth / Scale / Volume) at progressively cheaper per-credit rates. Monthly free tier is now 2,000 credits. New Enterprise Annual plan available on request
Fixed
  • •Five list endpoints (employees, assessments, PDPs, exams, talent-market search) now properly filter by role: admins and HR see the full tenant, managers see their own subtree, regular employees see their own record. Previously every authenticated user could list everything in the tenant
  • •Talent Market: a regular employee now also sees unpublished cards where they are a candidate, not only the published ones
  • •Connections to the database recover faster after network blips — the pool now pings before reuse and keeps TCP keepalives alive, and pool capacity is large enough to absorb load spikes
  • •Adding a participant to an assessment no longer crashes with a 500 when the underlying user has been deleted; the API returns 404 instead and orphaned rows are cleaned up
  • •Division edit modal now exposes the Manager field (previously only the Deputy Manager field was visible), and both selects work without the spurious "Invalid manager" error
  • •Regenerating AI position drafts no longer 500s on a unique-constraint violation when the AI returns a title that already exists as a draft
  • •Cross-domain redirects between the marketing site and the app no longer cache stale routing rules in the browser
  • •Status page (status.hrpulsar.com) no longer gets stuck on an empty placeholder when an upstream probe returns a non-JSON response

1.4.2

April 28, 2026
Improved
  • •Pricing calculator on hrpulsar.com/pricing now lists exactly which actions drive the monthly estimate, sorted by spend, before recommending a top-up package — no more opaque "≈ N credits" total
Fixed
  • •Cloud status indicator on status.hrpulsar.com is back to "operational" — the underlying job-availability heartbeat is publishing again

1.4.1

April 28, 2026
Added
  • •Pricing page now leads with a curated "What things cost" table — eight headline actions priced per credit, instead of the full eighty-line dump
  • •New "Estimated monthly spend" calculator — pick your team shape (solo / 10 / 50 / 200 / 500+) and see live monthly credit estimates plus the right top-up package
  • •Cloud admins get a new Billing layout with three tabs (Overview / Transactions / Prices) and a searchable, in-app price list
  • •Assessment detail page now shows the evaluation criteria with an inline "Add competences" button, and a clear read-only view of the rating scale
  • •"Take this assessment" call-to-action appears at the top of an assessment for the people who still need to fill it in — no more hunting through the participant list
  • •An "Evaluatee" badge highlights the person being assessed in the participant list
  • •Employees can now jump straight to an assessment from their profile's Assessments tab
Improved
  • •Creating any assessment automatically adds the assessed person as a self-participant, so the participant list is never empty out of the gate
  • •Self-assessments no longer auto-include the employee's manager (it was confusing — you were the only person evaluating yourself)
  • •Done and cancelled assessments hide the action buttons that no longer apply (Add participant, Calibrate, Evaluate, status change)
  • •Mass-cancelling an assessment cycle now reports how many sub-assessments were updated vs. already-finished, instead of just counting the changes
Fixed
  • •Regular employees no longer see admin-only buttons (status transitions, Add participant, Calibrate) that would fail with "Insufficient permissions" when clicked
  • •Submitting answers to someone else's evaluation row is now properly rejected by the API — previously any user in the same workspace could write into anyone's row
  • •Public pricing endpoints normalised under /api/billing/* so the marketing site and dashboard share one base URL — fixes the dashboard's missing free-tier limit and credit packages on first load
  • •Cloud pricing page on hrpulsar.com is back online — the live credit cost table loads instantly again, and the "Failed to fetch" error is gone

1.4.0

April 27, 2026
Added
  • •Public waitlist for early access — sign up on the landing page to be notified when HRPulsar opens up. Confirmation email included
  • •Refreshed brand identity — new chronograph mark, Geist wordmark, light and dark lockups, high-resolution social card
Improved
  • •Dashboard redesign — KPI strip with sparklines, headcount by department, active assessment cycle progress, an inbox of items needing attention, and a calendar strip for the week
  • •Sidebar reorganised into Workspace / Talent / Discover / Admin sections with the tenant switcher moved here from the header
  • •Header pared down — auto-generated breadcrumbs, ⌘K hint in the search bar, leaner 52px row
  • •Employee detail pages lead with an identity card showing status, tenure, assessment progress, goals, and the last review at a glance
  • •Dark theme overhauled — KPI bars, charts, focus rings and active navigation are visible in dark mode (they were greyscale on dark before)
  • •Transactional emails now feature the real HRPulsar lockup in the header instead of the placeholder mark
  • •Public status page icons match the main app's branding
Fixed
  • •Web app manifest no longer triggers a console error on the landing site

1.3.0

April 25, 2026
Added
  • •Structured job positions — create and manage positions with optional grade, specialization, and division links
  • •AI-powered position generation — describe a role and let AI create the position with a draft/approve workflow
  • •Position management page at Company → Positions with search, filtering, and headcount tracking
  • •In-site documentation — all docs are now rendered directly on the website instead of linking to GitHub
  • •Cloud & Enterprise features page — learn what's included with a Cloud subscription
Improved
  • •Employee and invitation forms now use a position selector instead of free-text input
  • •Docs page reorganized by user journey — getting started, features, cloud, deployment, API, and GitHub
  • •OpenAI and Gemini API keys now documented in the deployment guide

1.2.10

April 24, 2026
Improved
  • •Employee creation form now uses a dropdown selector with user names and origin badges instead of manual ID entry
Fixed
  • •Background task worker stability improvements

1.2.6

April 23, 2026
Added
  • •End-to-end test infrastructure for improved quality assurance
Improved
  • •Company page loads gracefully when some data sections are unavailable
  • •Assessment participant selection now uses employee names instead of raw identifiers
  • •Human-readable labels in mass assessment filters for divisions and specializations
Fixed
  • •Authentication no longer loops when entering an incorrect password
  • •Development plan card clicks now respond to the correct target areas
  • •Email verification and invitation links work correctly for new users

1.2.5

April 23, 2026
Added
  • •Billing dashboard with credit balance cards, spending analytics by category, and transaction history
  • •Frontend can now detect deployment mode (SaaS vs self-hosted) for adaptive UI
Improved
  • •Human-readable labels throughout the interface — no more raw IDs in dropdown menus

1.2.0

April 22, 2026
Added
  • •Mass assessment creation — create assessments for entire teams or departments at once
  • •App version display across all pages (landing, auth, dashboard)
Fixed
  • •Human-readable labels in Select components across the platform