Also thinking deeper, I think your question also is very broad on aspects to answer. Technically in times where there is a gathering of ai and people I think a single chat surface is not the right experience but yeah having a UI saying agents as tabs felt tooo boring and actually fed up.

Just wanted to keep it fun but at the same time can we elevate the experience when there is a common surface where a bunch of ai-npcs and people can meet.

I hope it answers, happy to dig more in a different direction if you felt I understood it wrong

yeah a question with which I started :)

I definitely love the format, but the real reason is each town comes with it's own creative story, theme and interaction rhythm. I first thought about making one town that held them all which I wanted to have for myself or say my own town. But then later thought different people want to have their own representation of it, since it's english, different themes and all about creativity it has no bounds.

In the hindsight I also wanted to figure out which experiences it improves vs which it decorates. Curious what you felt when you experienced it.

hey thanks for the feedback, agreed there are a lot to get right

1. If you have self hosted it yes then we need OpenAI api key, we are also working to get things working with llama so should be available soon (but if from cloud you shouldn't have got this error)

2. We also have moved away from openAI embedding (BGE-M3) to our own hosted in the cloud.

In self host we already have claude code support happy to help there

We faced the same challenge while building SOL (https://github.com/RedPlanetHQ/sol) — a personal assistant that relies heavily on memory for context and continuity.

Getting LLMs to invoke memory tools at the right time is definitely trickier than just wiring up MCP correctly. We're still refining it, but we've made good progress by explicitly guiding the assistant within the system prompt on when and how to use memory.

You can see an example of how we structure this in SOL here: Prompt instructions for memory usage (https://github.com/RedPlanetHQ/sol/blob/964ed23c885910e040bd...)

Using something on similar lines as rules in claude/cursor etc has been working better. It’s not perfect yet, but this combination of prompt engineering and structured tool exposure has been moving us in the right direction.

ps - one of the authors of CORE

Hey thank you, you can find this here https://github.com/RedPlanetHQ/core?tab=readme-ov-file#core-..., it's a docker compose based setup

What’s even the point of calling yourself open source if you can’t bother to support LLaMA models? Kinda defeats the purpose, don’t you think?

We are working on it, LLaMA models are not working good when it comes to finding facts from the message. We should have this out in the next couple of days

I was working on something similar https://github.com/harshithmullapudi/tsukuru, so to create a personal website by just writing mdx files. Here is my website https://github.com/harshithmullapudi/harshith.sh

Could you help with how the workflow looked like? How does it ensure that the code is tested?

Strictly speaking it doesn’t.

It ensures someone else has moved the card to the next status (or, that automated tests have passed and moved it to the next status by automation)

You can permission who can move to what status.

As a random dev, I personally wouldn’t recommend it unless there are legal requirements for such separation (or your org has lawyers that think such requirement exists)

Ahh, yeah paper trail makes sense. But the ownership aspect of where it went wrong can we understand without the workflow restrictions also right?

Really appreciate the thoughts. The places where we believe AI matters a lot is

1. Ensuring ticket has enough information 2. Offloading some amount task before someone picks it.

These are the 2 key areas we are focusing and we will launch some features around these in the coming weeks. Would love to hear if there are some other places you feel we can focus on?

What about, instead of using AI to summarize the issue text for the developer, using AI on the _submission_ end to notice if the bug report doesn't contain enough details and ask the submitter clarifying questions until the submission is a really good bug report?

We tried different approaches, we made a recent change where we started using cohere for re-ranking and embedding which did give really good results.

Happy to create an account for you to try it out. harshith [@] tegon.ai