youknow

Coalesced memory read for a slightly faster LLM interence

3

## Background I enjoyed reading the posts related to this tweet: https://twitter.com/awnihannun/status/1776275621926375498 I'd like to contribute to the MLX side in the challenge with Llama.cpp, even if it's just a...

Support custom shortcut key commands

2

This PR introduces basic support for custom shortcut keys in Aider. Many programmers may love custom shortcuts, especially in terminal-based environments like Aider. This feature allows users to define their...

Improve SEARCH/REPLACE accuracy (fixed)

1

LLMs, including GPT-4o, often provide a very short context for the SEARCH block to match. For example: ``` > REPLACE ``` In this case, the current Aider code just uses...

Add Gemma2-27b-q8 benchmark result to leaderboard

1

Adding Gemma2-27b-q8 result to leaderboard for local LLM enthusiasts. - Tested with `Ollama 0.2.1` - `gemma2:27b-instruct-q8_0` model with hash `dab5dca674db` This LLM model struggles with diff formatting; whole format works...