It looks like the problem with sporadic disconnects in telegram integration is now resolved. But we will take a close look on the matter, to be fully sure.

[aav via huly.io]

Now egress messages routed to Telegram channels, are optionally amended with the tear line. Configuration parameter is introduced for this purpose: HULY_MESSAGE_TEARLINE

[Alexey Aristov via https://huly.io]

📊 Create Mermaid diagrams directly in Huly for better collaboration! Check out our guide: https://github.com/hcengineering/platform/wiki/Simple-diagrams-in-Huly-with-Mermaid

Review status: SLA is 3 business days for the review process.

Yes, in the end, building Huly as an MSIX package turned out to be too costly. Our auto-update system can't be stuffed in there without modifications - we'd need to make a separate build for Windows with a special code path.

But in the Microsoft Store, as it turns out, you can submit the existing installer for publication, albeit with some minor limitations.

Waiting for the Review results 🤞

[Denis Gladkiy via https://huly.io]

Huly is now officially in the Microsoft Store! ( https://apps.microsoft.com/detail/xpdbz19bxsd9dv )
If you're on Windows, you can install it with a single shell command: winget install Huly

[Denis Gladkiy via https://huly.io]

Hi everyone , as you might know we are working on AI agents, and we are working on several use cases. We have described one use case on how this will tremendously help the prospection on social medias - feel free to have a look at short video recorded, and if you want to see what we are working on and will be coming, we are happy to share more here is the 2min video

[Charles Rollet via https://huly.io]

As many as three people on the team caught a cold. Even though we have a fully distributed team, it's starting to seem like viruses have learned to spread through wires and Wi-Fi.

I've recovered a bit and knocked out some improvements for the drawing tool: shapes, minor bugs ( https://github.com/hcengineering/platform/pull/10142 ).

[Denis G]

Blue (or red?) stripe for me. Got back to working on 3D engine tasks, πD.

Right now we have enough features implemented to build interesting scenes with them: SDFs and their composition, 3D texts, animated procedural and bitmap textures, object morphing animation, lighting.

The immediate goal is to achieve proper performance on high-end GPUs. The first step I tackled was removing data transfer to the denoiser and back through data copying.

Turns out people have already figured out how to punch a hole between WGPU and Intel OIDN ( https://github.com/Vecvec/oidn-wgpu-interop ). Now I need to pull this into our codebase. That's what I'm working on.

[Denis Gladkiy via https://huly.io]

Although it's not a breakthrough, the result is encouraging. I managed to completely eliminate memory copies between video and system memory when working with OIDN (Intel Open Image Denoiser). The code isn't just undebugged, it's also "held together with duct tape," but the current performance is 7 ms for denoising per frame (on a GeForce 3070). This is good because the performance floor is 33 ms per frame. So I have both a time budget reserve (about 26 ms) and a technological one (on a GeForce 5070 and more powerful cards, everything will be faster).

Note: eliminating memory copies reduced the denoiser time from 12 to 7 milliseconds—a 1.7x speedup for this stage.

The aforementioned budget, 26 ms, is how much I have for ray tracing. Not much, but it's something workable. Right now I don't fit within this budget, plus there's synchronization between the denoising stage and the tracing stage. They need to be separated: denoise the result of the previous frame while the current frame's tracing is being computed. Plus the ray tracer's architecture is currently very poor—an "Uber Shader". It needs to be converted into separate stages.

[Denis Gladkiy via https://huly.io]

Meet our newest contributor to the Huly Platform: Ryzhik the Cat! 🐾

Ryzhik has already made some impressive changes:

i
\\\\
Pm

I’m not sure what these mean, but the code smells like tuna and curiosity.

I can't commit this to our codebase at this point, but I hope he continues to develop his talents — maybe one day he’ll master rushx build (or at least stop sitting on the keyboard)!

[Artyom Savchenko via https://huly.io]

This morning, Alexander Onnikov and I set out to build the production version of Huly.
But the universe had other plans!
All previous builds went smoothly, but starting today, even 30 GB of disk space just wasn’t enough.
I wouldn’t call it a tough problem—we’ll optimize our build process soon (maybe with a little magic).
But today, I had a flight booked to celebrate my father’s 70th birthday.
Tomorrow, I’ll be back, ready to tackle these challenges with fresh energy.
[Artyom Savchenko via https://huly.io]

A Week of Fires, Fixes, and a Plan for Better Releases

This week was a testament to resilience. After facing downtime from the global Cloudflare outage and a separate hosting provider issue, I stabilized our Docker pipeline and shipped a new version on Thursday.

True to form, the release uncovered new bugs, which we've mostly fixed. A patched version is imminent.

This process highlighted flaws in our release strategy. Here’s how we’ll improve:

Separate Release Trains: Clearly distinguishing major releases from bug-fix ones.

Better Communication: For major releases, we will provide clear change logs and data migration notes.

Increase Frequency: Shipping bug fixes more rapidly.

Enhance Testing: Implementing more thorough pre-release testing.

We're turning this week's challenges into a more robust and predictable release process.

[Artyom Savchenko via https://huly.io]

Woke up at 4 am? Perfect Time for a "Quick" Prod Deploy.

It's 4 AM. You're wide awake for no reason, and your brain, instead of dreaming about peaceful landscapes, decides it's the ideal time to think about bugs in production. "I'll just do a quick deploy," you tell yourself. "It's a tiny update, just a few changes. What could possibly go wrong?"

So, there I was, pushing the new version of Huly to production. The deployment itself was smooth — a clean, five-minute job. However, then I noticed that messages were taking a long time to load in chats.

I dove into the SQL server metrics. The culprit? A massive spike in the execution time of queries that count the number of messages in chats.

The graph told a sobering story. It turned out these counting queries had always been sluggish, sometimes taking up to 10 seconds to run.

The diagnosis was a classic one: a missing database index.

The query was doing a full table scan every single time it needed to count messages for a chat. As the messages table grew, this became exponentially more painful.

The Fix? Simplicity itself. A well-placed index on the workspaceId and space columns was all it needed. Query time plummeted from over 10 seconds to just 5 milliseconds.

As for me? After adding the index and watching the graphs return to normal, I went back to bed. However, in the morning there were already other problems that had to be dealt with.

[Artyom Savchenko via https://huly.io]

Hi everyone!

We have published a new release v0.7.375 with the following changes:

Internationalization
- Add Brazilian Portuguese translation
- Enable Turkish translation

Workspace invites & login
- Redesigned workspace join flow
- Show Google login on mobile devices

Guest collaborations
- Guests can now upload attachments, use reactions, and configure notification settings
- Allow notifications for guests with collab security

Docs & export
- Ability to export multiple docs to pdf
- Allow to copy all table content as markdown table for documents and issues
- Add import notifications

Tracker
- Added ability to group issues by project
- Add project breadcrumb
- Fix issues layout

Cards and processes
- Approve requests
- Add unlock card and section functionality with UI updates.
- Implemented process import/export, improved process execution flow, and added CancelSubProcess action
- Card notifications

Controlled docs
- Added an activity section for controlled documents.
- Improve controlled documents export from one workspace to another
- Support browser translations for read only documents

Permissions & security
- Add drive permissions

Diagrams
- Allow to open mermaid diagram in fullscreen

Infrastructure improvements
- PostgreSQL support for KVS service
- Support in-memory mode for hulypulse
- Support minified dev deployment
- Draft Windows development support without WSL

Self-hosting updates
New infrastructure allows to simplify Huly platform deployment.
We plan to update the self-hosting documentation and add additional deployment options soon.

Full Changelog: v0.7.353...v0.7.375

[Artyom Savchenko via https://huly.io]