Erik Bjäreholt

docs: gptme-suggested README improvements

1

I just asked gptme to improve it. I think it is good to including some of these key aspects in the README which we should probably mention directly without referring...

feat: add token limit warnings with improved timing

1

Adds token limit warnings that show: - Token count info every 10k tokens - Warning when reaching 50% of context limit (once per model) - Warnings appear before user input...

Open URLs to GitHub issues/PRs/repos using `gh` tool instead of browser

2

The `gh` tool supports `gh issue view --comments `, so should be easy and a decent improvement.

feat: added think tool to let agent query a reasoner

1

## TODO - [ ] Send full conversation context (different mode/tool?) - [ ] Karpathy-style consortium ---- > [!IMPORTANT] > Adds `think` tool to enhance reasoning on complex problems using...

The bot should be a lot more capable to make code changes with Claude Sonnet 3.5, and now 3.7. I should polish it and throw a few issues at it...

refactor: attempt to refactor & simplify chat.py

2

Mostly vibecoding. Some things are broken but some good ideas here too. ---- > [!IMPORTANT] > Refactor `chat.py` for improved readability, introduce `paths.py` for path handling, and enhance interrupt handling...

Retarget PR to base once #447 is merged

"Ephemeral messages" are messages which are intended to be dropped after a while, like after a few messages. To ensure effective prompt caching, we should make sure we request a...

``` User: https://gptme.org/docs/evals.html [22:37:50] ERROR Error in browser thread _browser_thread.py:76 Traceback (most recent call last): File "/home/erb/Programming/gptme/gptme/tools/_browser_thread.py", line 72, in _run result = cmd.func(browser, *cmd.args, **cmd.kwargs) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/home/erb/Programming/gptme/gptme/tools/_browser_playwright.py", line...

misc: playing with rl/verifiers

1

Didn't get it to run, was just vibecoding with gptme. ---- > [!IMPORTANT] > Adds `gptme_verifiers.py` to implement an RL environment for training LLMs using `gptme` and `verifiers`, with classes...