Erik Bjäreholt
Erik Bjäreholt
Since the experiment with integrating pre-commit checks wen't well, I figured I should attempt to add an autocommit feature as well. Would speed up many of my workflows, especially with...
Paths should either be stored as absolute paths, or be resolved relative to the workspace root (which should probably be stored in the logfile folder) and not the current directory....
> [!IMPORTANT] > Enhance `chats` functionality in CLI and `chats.py` with context and match limit options, improving search and read capabilities. > > - **Behavior**: > - `list_chats()` in `chats.py`...
I was testing deepseek/deepseek-reasoner with `GPTME_BREAK_ON_TOOLUSE=false` and it didn't use the right tmux session because it proposed 4 commands in one go, but it needed the output of the first...
I noticed this suspicious eval output with exact same number of tokens for markdown and xml formats: Inspected the logs and noticed it wasn't using the xml format in the...
Sometimes when using the shell tool it will give output for the last command instead of the current command, making for very confusing output to the LLM. I'm not sure...
Whenever something like `[text]` appears in stdout it corrupts content/syntax highlighting. cat'ing a pyproject.toml (or any other toml file) therefore looks trash. I've also seen cases where syntax highlighting fails...
Played with this a couple of days ago, idea was to mimic a similar tool I saw in one of the SWE-Bench submissions. Could be a more reliable stopping-condition (and...
Getting output like this when running a suite of 4 evals: ``` 2025-01-14 13:34:18,583 - WARNING - gptme.tools - Tool 'append' already loaded 2025-01-14 13:34:18,583 - WARNING - gptme.tools -...
Has some similarities to https://github.com/ErikBjare/gptme/pull/371 ---- > [!IMPORTANT] > Refactor keyboard interrupt and user command handling in `chat.py` and `commands.py` for consistent behavior. > > - **Behavior**: > - Refactor...