gptme
gptme copied to clipboard
ErikBjare
A CLI and web UI to interact with LLMs in a Chat-style interface, with code execution capabilities and other tools.
This might require significant rewrite of how tokens are output, turning the request into a generator that emits tokens instead of a full message. The tokens can then be collected...
Having the agent speak to you could be useful so you can drop your attention from the terminal. Highly synergistic with voice input.
When you know you are going to produce a bunch of stdout that you do not want Assistant to analyze, make an optional way to suppress the stdout, so we...
This could allow for running/testing GUI applications, and more E2E multimodal behavior. Xvfb in itself might be a bad idea (although good for running in headless/CI), could just take screenshots...
Right now it sometimes tries to install Python packages, but fails since it doesn't install them in the Python environment it is running in (I assume). Could probably set some...
We've had some CI issue where OpenAI sometimes spuriously closes the connection. In these cases we should retry the request at least once. I worked around the CI thing by...
To show the user what you can do, fill in basic info about them to better serve their requests. Could depend on #35
- Add web UI screenshots - Record demos with WizardCoder/Mistral - Multiple demos would be nice, to show off breadth of functionality/usecases Good demo-material exists in `./tests/test-integration.sh`
Right now, it tends to output large chunks of the original file when writing patches.
We can already give conversations a name automatically with `/rename auto` but this should be improved: - [x] Name conversation automatically with `/rename auto` - [ ] Don't include date...