#general
Hey, my name is Anmol-I'd love to work on issue https://github.com/archestra-ai/archestra/issues/3406 (Claude models failing through the Bedrock proxy with
body/system/1 Invalid input). Mind if I pick it up?
New release! 🚀🚀🚀
1.2.21 (2026-04-22)
Features
- OpenAI Complations - Bedrock converse compat layer (#3996) (6ea9c8a)
- unified connect page (#3980) (f198391)
Bug Fixes
- gate team vault-folder queries on byosEnabled flag (#3984) (dc698bc)
- IdP JWKS auth selection (#3989) (3bd3dcf)
- persist chat errors on conversations (#3982) (b46167e)
- render user chat messages as plain text (#3992) (33d5853)
Miscellaneous Chores
Hey team!
I've had a few feature ideas that I think could really improve the experience for users. Wanted to check if they align with the current roadmap. If they look good, I'd be happy to open issues and take them on!
1. Agent Cloning (One-click duplication): Right now, if you want to fork an agent to test a new prompt or slightly different toolset, you have to recreate the whole config (tools, KBs, sub-agents, labels) from scratch. I’ve seen this in a few other products and it's a huge time-saver. It would be a pretty self-contained backend addition using the existing model associations.
2. Knowledge Base Search Feedback Loop: Right now, we can see sync stats and query volume, but we have no way to know if cited documents actually helped the user. Adding small 👍/👎 buttons to citations in chat would give us a direct signal on which docs are authoritative vs. which ones are just noise. This helps admins find content gaps (frequent 👎) and identify outdated docs without manually auditing thousands of files. In the long run, we can feed this data back into
rrf.ts to automatically boost documents that users find consistently helpful. I think it adds a lot of value for anyone managing enterprise knowledge at scale.What do you all think?
Hey team 👋. I'm Sasha)
I’m interested in contributing to this issue: https://github.com/archestra-ai/archestra/issues/3838
Before I jump in, I wanted to ask - is there already a defined vision or some expectations for how this should be approached? Or is this more of an open-ended task where contributors are encouraged to propose their own direction/implementation?
Happy to align with any existing ideas, or draft a proposal if it’s still exploratory.
🙌1👋1
7 replies
Can't we attempt the the issue with the tag Reserved for SE Interview?
Read-only live mirror of Archestra.AI Slack
👋Join the discussion withAI enthusiasts!Hey team!
I've had a few feature ideas that I think could really improve the experience for users. Wanted to check if they align with the current roadmap. If they look good, I'd be happy to open issues and take them on!
1. Agent Cloning (One-click duplication): Right now, if you want to fork an agent to test a new prompt or slightly different toolset, you have to recreate the whole config (tools, KBs, sub-agents, labels) from scratch. I’ve seen this in a few other products and it's a huge time-saver. It would be a pretty self-contained backend addition using the existing model associations.
2. Knowledge Base Search Feedback Loop: Right now, we can see sync stats and query volume, but we have no way to know if cited documents actually helped the user. Adding small 👍/👎 buttons to citations in chat would give us a direct signal on which docs are authoritative vs. which ones are just noise. This helps admins find content gaps (frequent 👎) and identify outdated docs without manually auditing thousands of files. In the long run, we can feed this data back into
rrf.ts to automatically boost documents that users find consistently helpful. I think it adds a lot of value for anyone managing enterprise knowledge at scale.What do you all think?
6 replies
One more thing on the KB front: I was also thinking we could extend that into a Knowledge Base Document Browser.
Right now, once a connector syncs, it’s a bit hard to see what actually made it into the index. Having worked with the connectors here before (Slack, Google Drive, etc.), I’ve noticed it’s a challenge to verify if the parsing worked correctly or to prune out noisy/stale data without jumping into the database.
I'm imagining a simple 'Documents' tab on the KB page where admins could:
• Browse the full list of ingested documents (titles, source URLs, last updated).
• Text search across document titles to find specific files.
• Manually delete individual documents to clean up the index instantly.
I think this would go a long way in making the KB feel less like a 'trust me, it's in there' system and more like a manageable data asset.
hey! these are all great ideas I think. if you wouldn't mind creating some github issues for these 3, I can assign you :thumbsup_all:
agent cloning and document browser make sense
I'd have to think a bit more about #2 (@user any thoughts on that one and how that'd potentially improve query side of RAG?)
:archestra-love:1
Thanks @user! I've created the issues for them. We can discuss in the comments as needed... 👍
I would say one at a time for now, I can only review so many things at once 😉 I assigned 4022 to start, sound good?
Perfect, thanks