Absence Approvers, PM Agent Protocol, and UX Enhancements
This release introduces mandatory approvers for absence types, establishes a structured JSON protocol for PM Agent tasks, and improves overall user experience with loading indicators and automatic company branding. We've also enhanced PM Agent's conversation context understanding and fixed worklog calculations for extended time entries.
Release dates: 10/21/2025
Improvements to Absences
Copy link
- Added a mandatory approver system for absence types. We created a new Mandatory approvers table in Settings → Absences settings that allows you to assign mandatory approvers for each absence type. The table lets you configure rules to assign approvers only among full-time employees or for everyone. You can also configure automatic addition of approvers to requests and handle edge cases (for example, an employee cannot be an approver for their own request). HR specialists now have a flexible tool for controlling leave approvals; you can centrally configure who must approve specific absence types (for example, all vacation requests through a designated HR specialist). This ensures compliance with corporate policies and prevents approval procedure bypasses for critical absence types.
Improvements to PM Agent
Copy link
- Developed and implemented JSON protocol for PM Agent Periodic Tasks. We created a structured JSON protocol for creating PM Agent periodic tasks, including name_of_task, instruction, schedule (type, datetime, timezone, recurrence), platforms (email, Telegram, Slack, team_chat), and language. PM Agent now has a formalized protocol for creating tasks with support for multiple notification delivery platforms. Users can now configure automatic reminders and reports with flexible channel selection (email, direct messages in Telegram/Slack, or group chats), ensuring notification system versatility for various corporate scenarios.
- Fixed message relationship detection logic in the PM Agent context. We revised the prompt and expanded the message sample for similarity checking. The system now correctly determines user request relationships without limiting itself to only human messages and properly connects dialogue context. PM Agent has improved dialogue context understanding and now more accurately determines when a new question is related to a previous discussion. This ensures more natural conversations with context preservation and eliminates situations where the agent loses the thread of conversation when switching between related topics.
Other improvements
Copy link
- Implemented automatic logic for displaying the company logo/name in the menu panel. We configured an automatic display system in the menu panel: when an uploaded logo is available, it displays the logo; when absent, it automatically displays the company name from the Company Name field. Users now have a simple and intuitive interface branding system without manual switching. The system automatically determines what to display in the menu panel, ensuring constant company identification in the interface regardless of whether a graphic logo is available.
- Added loading indicators for all tables. We implemented visual loaders (loading indicators) for tables throughout the system, clearly showing the data loading process. Users no longer experience uncertainty when working with tables: it's now immediately clear whether data is loading, whether there's no data at all, or whether a technical problem has occurred. This eliminates frustration from unclear interface behavior and increases overall trust in the system.
- Fixed the Total calculation per day in the Global Worklogs report for long worklogs. We resolved an error in calculating the Total per day for worklogs longer than 24 hours. The system now correctly parses and displays entries up to 200+ hours long, properly summing them in the overall Total and daily breakdown. Users of Azure DevOps, where approximate worklog tracking with large values at once is practiced, now have accurate total work time calculations. The match between Total in the list and the breakdown restores trust in the data and eliminates the need for manual hour verification for reporting.