First of all, thank you for your great work on this project.
I noticed that when new entries are appended to the output (or context window), the view automatically scrolls to the bottom. This is useful in many cases, but it makes it difficult to keep reading at my current position if I’m reviewing earlier content while new data is being added.
Feature request:
It would be great to have a behaviour where:
- If I am at the bottom, auto-scroll continues as normal.
- If I have scrolled up to read previous content, my reading position is preserved, and auto-scroll is paused until I manually scroll back to the bottom.
As a general note: I tend to read only the "thinking..." part of Claude's output. I would appreciate a UI with much more focus on what the LLM thinks is going on, as I want to monitor the LLM (not the code). A "compressed view" with much less vertical empty screenspace, token usage moved to the side, and filtering for system messages (the cog wheel comment), edits, and user messages would be appriciated.
Thank you for considering this!
First of all, thank you for your great work on this project.
I noticed that when new entries are appended to the output (or context window), the view automatically scrolls to the bottom. This is useful in many cases, but it makes it difficult to keep reading at my current position if I’m reviewing earlier content while new data is being added.
Feature request:
It would be great to have a behaviour where:
As a general note: I tend to read only the "thinking..." part of Claude's output. I would appreciate a UI with much more focus on what the LLM thinks is going on, as I want to monitor the LLM (not the code). A "compressed view" with much less vertical empty screenspace, token usage moved to the side, and filtering for system messages (the cog wheel comment), edits, and user messages would be appriciated.
Thank you for considering this!