Repo Assist?

[dsyme]

Hi @SchlenkR

Would you like me to add Repo Assist in this repository?

https://dsyme.net/2026/02/25/repo-assist-a-repository-assistant/

Thanks
Don

[SchlenkR]

Hi Don,

thanks for asking. I installed it. Here are some (personal) critical points for this repo:

• Cost is an issue: already spent > 5€ since yesterday for a few planned + triggered runs. Might be possible to bring that down, but anyway: In addition to time, things cost now also money.
• The process of finding and finally making decisions (and reviewing their implementation) still is - as it was - the time and resource consuming part. I see here only a minor help, which also exists when using a "normal" agent at the point when the decision is made towards "Let's implement this".
• In addition to an overkill-count of existing triggers, I now feel having another one requiring attention. This time, it's a bot; not a human being with the potential of bringing "real" joy of working together. I was always excited finding a "[fsprojects/FsHttp]" notification mail in my inbox when someone reached out. Today, I felt the requirement of having react to a long-commented PR from a proactive bot. My personal feeling is: I just don't want that.
• It's a Trap ;) It doesn't hold what it promises (is there a promise? At lease an implicit one I guess). There's a pool of endless interesting things to work on and explore - this repo included - and repo-assist feels like a "pull-factor" that lures me into this, and I currently can't spend time on (unfortunately).

Thank you, and I wish you a good time.

[dsyme]

Hi @SchlenkR,

Thanks for the feedback!

Costs is a factor, yes. For this I can install a token, for 30 days. You can also use a token from a Copilot or Claude or Codex subscription, I assume you're using a Claude one, which is costly. We're planning to add Mistral and others.

Moving to weaker models is possible and reduces costs greatly. On some plans some of the models are near free.

It's interesting it draws you back. I found that too.

[dsyme]

Regarding "working with the bot" - I understand, I actually don't like that either and don't see this as for personal direct productivity, which I'd mostly do from an IDE.

I do like

• the labelling
• the comments and fixes on old issues and suggestions I would never have time to analyze and feel guilty about never having responded to
• the unexpected suggestions about taking the repository forward, eg new engineering validation or cleanup.

But I don't want to work "with" the AI, I just want it to help me achieve my goals of a clean, bug free repository.

I think it could be good if we added a chat on installation where it clarifies with you whet kind of assists you might like (eg focus on fixing old bugs then stop), how often and tailors things.

Thanks again. Don't at all feel like you need to continue using it especially if the repo is dormant, would be happy to keep hearing feedback if you do.