aider

aider copied to clipboard

Evaluate performance against SWE-Bench

1

It would be interesting to see if/how `aider` performs against the SWE-Bench benchmarks: - https://www.swebench.com/ - https://github.com/princeton-nlp/SWE-bench - > [ICLR 2024] SWE-Bench: Can Language Models Resolve Real-world Github Issues? -...

0xdevalias

Make the LLM understand when some but not all edits are applied due to a diff error

2

When asking questions or reporting issues, it is very helpful if you can include: ``` Aider v0.22.1-dev Add .aider* to .gitignore (recommended)? n Model: gpt-4-1106-preview using udiff edit format Git...

cranyy

Support typing/pasting tabs into the chat in all terminal environments

9

I see no way to copy & paste blocks containing tabs into the prompt. Multiline mode doesn't work either. `session.prompt()` is ignoring the tabs (they get lost and the context...

kripper

universal-ctags not enabled?

1

``` Aider v0.28.0 Model: gpt-4-1106-preview using udiff edit format Git repo: .git with 628 files Repo-map: using 1024 tokens ``` Following the examples in some of the video-tutorials it was...

9M6

Explore whether stack graphs may be useful in this tool

1

I'm not sure if this would add anything substantially different/better than the existing `tree-sitter` implementation, but I spent some time recently looking into stack graphs + related tooling, and I'm...

0xdevalias

Default model when invoking aider via script using GPT-4 instead of GPT-4-turbo

1

When asking questions or reporting issues, it is very helpful if you can include: - Aider version: 0.27 - Model being used: gpt-4 instead of gpt-4-turbo-preview - Other switches or...

thehqlife

Anthropic without OpenRouter middle man

3

OpenRouter is a nice solution, but since Claude is so popular, it'd be nice to have direct support within Aider so extra costs can be avoided. I found https://github.com/jtsang4/claude-to-chatgpt which...

aleclarson

Aider dumped partial context prompt

1

``` Aider v0.27.0 Model: gpt-4-1106-preview using udiff edit format Git repo: .git with 194 files Repo-map: using 2048 tokens Added bin/available-subnets to the chat. Use /help to see in-chat commands,...

harleypig

How will your benchmarks fare if you say Please and Thanks in the prompt?

7

I noticed that the prompts at https://aider.chat/docs/benchmarks.html don't start with Please and don't end in Thanks! -- I'm curious if your benchmark would improve if you did.

JPhilipp

[feature request] Allow running commands in the environment: compilation, running test cases, and so on

1

AFAICT (eg from https://aider.chat/examples/complex-change.html), aider does not currently support running commands automatically to let the agent test whether its changes have the desired effect, eg passing a test case. [SWE-Agent](https://swe-agent.com/)...

samuela