Major concerns with rovodev

Rovo
1

I have been using rovodev cli for last 2 days only, trying it our comparing it to my claude code subscription. following are my observations.

  1. there are many EOF File errors while writing .md files that has json data within. if this error does pop up, rovo just crashes and tells to re-initialte the session with – restore. which is a bit excessive to be honest.
  2. The /prune doesnt work. I asked it to create a basic boiler plate, and it did, asked for few corrections and reached about 150k /200k session token, run /prune command, and after next message the tokens were still 120k/200k. continuing to use ahead i noticed the daily usage getting absolutely slaughtered because with every hit its passing all of the above.
    Positive with this is that its very sane acorss the whole session, but the tokens go for a toss…

except for this, I am liking it, its much more sane, easy to handle and doesnt go off on a random tangent. also, the UI is beautiful

Just adding my contribution for the Developers. :slight_smile:

image
image1550×1748 267 KB

2

I realized that the issue might not be caused by JSON data, as it also occurs when generating pure Markdown format. Therefore, I suspect that the JSON output is inherently part of the LLM model’s response, but due to unknown reasons, it causes crashes.

To address this , I now enforce incremental writing for document generation: each analyzed section is written immediately after processing, rather than all at once. This significantly mitigates the problem and helps avoid errors.