Tigran Kostandyan in #general: "While working on error processing in web chat I noticed we started showing duplicate error cards – it wasn't like this a few days ago. Traced it to th..." | Archestra Community

#general

Apr 22April 23, 2026

Hi Team, I am Palak

I am interested in contributing to Archestra, but I found that commenting on issues is currently blocked. Could anyone please guide on how can we claim an issue? Should I work on reproducing and provide a video of working fix first?

Ps: I’m about to graduate and have prior experience through internship at MNC as well as open source contributions. I’d also like to know if I’m eligible to participate in the interviews.

🙌1:archestra-love:1🚀2

1 reply

Phil Filippak7:03 AMOpen in Slack

Hi everyone! I'm Phil, have been working in software engineering for 15 years, will try and contribute to Archestra in the next few days, then we'll see how it works out

❤️1🙌1🔥1👋1

Ilya Korovin10:09 AMOpen in Slack

Hi, is the A2A API supposed to work right now, specifically with API keys managed on the Settings page? API keys in Settings are handled via ApiKeyModel, and it looks like that model is not used anywhere. For the A2A API, I only see UserTokenModel and TeamTokenModel.

14 replies

Alexander Balashov11:42 AMOpen in Slack

Hey all,

I’ve been poking at upgrading the repo from TypeScript 5.9 to 6.0, and it looks like a pretty smooth change.

I only ran into two things:

• a tsconfig deprecation warning, already addressed in this PR

• a small TS bug that the stricter compiler checks caught, already fixed in my TS 6 PR

So it feels like we're in a good spot to move to 6.0 without much churn. It also seems like a good prep step for eventually switching to the faster TypeScript 7 toolchain.

TypeScript 6 PR is here https://github.com/archestra-ai/archestra/pull/4035

😮1

10 replies

Archestra App5:02 PMOpen in Slack

New release! 🚀🚀🚀

1.2.23 (2026-04-23)

Features

*a2a:* accept arbitrary JSON payloads in addition to JSON-RPC envelope (#4037) (6d15f0c)
add onboarding wizard (#4013) (12f8a5a)
add Outline knowledge connector (#3938) (0bd17f3)
interweave persisted chat errors (#4002) (e5e3fec)

Bug Fixes

include scheduled- prefixed session IDs in sessionId search (#4040) (a1822eb)
scroll to bottom when user sends a message (#4005) (2ac31a6)
Slack setup slash command slugs (#4014) (0570258)
treat UUID search input as sessionId filter on sessions list (#4029) (e580151)

Documentation

fix stale mcpServer:admin reference in access-control docs (#3994) (a1fc7bf)
update agent docs (#4038) (315ffeb)

Miscellaneous Chores

clear agent identity provider config (#4020) (0bb5954)
knowledge base - make advanced-access control an enterprise feature (#3783) (10b4ff0)
remove deprecated downlevelIteration from tsconfig (#4027) (77bae6a)
upgrade TypeScript 5.9 -> 6.x (#4035) (2b7737e)

Tigran Kostandyan5:06 PMOpen in Slack

While working on error processing in web chat I noticed we started showing duplicate error cards – it wasn't like this a few days ago. Traced it to the AI SDK invoking onError twice after we added persistence inside that callback. Proposed fix here: https://github.com/archestra-ai/archestra/pull/4047

Join Slack to see the image

👀1

3 replies

Kyle Skutt6:45 PMOpen in Slack

Hi Kyle looking to be more involved with open source. Experience in building LLM, environments and various programs. I noticed most of the new bounties appear to be for in house members. If there is anything i can jump on let me know!

2 replies

Read-only live mirror of Archestra.AI Slack

👋Join the discussion withAI enthusiasts!

Thread

Tigran Kostandyan5:06 PMOpen in Slack

Join Slack to see the image

👀1

3 replies

joey (archestra team)5:22 PMOpen in Slack

hey! the bug and proposed fix make sense on the surface, I'm curious why onError would be getting fired twice in the first place. I see in your comment:

toUIMessageStream invokes onError twice for the same upstream
// error (once when formatting the error chunk's errorText, once
// as a notification when the chunk is walked downstream)

are there any vercel docs references to support that?

Tigran Kostandyan5:52 PMOpen in Slack

@user I'll move discussion to PR, but honestly this just comes from an empirical evidence 🙂 I just debug-logged it.

:archestra-love:1

Tigran Kostandyan6:25 PMOpen in Slack

I replied to you in the thread