Feature: Thinking command

[pcfreak30]

Issue

Add a /thinking or /think command to toggle between basic and think/CoT mode. this should either switch models or set the budget for the thinking stage.

Version and model info

No response

[paul-gauthier]

Thanks for trying aider and filing this issue.

The fix is available in the main branch. You can get it by installing the latest version from github:

aider --install-main-branch

# or...

python -m pip install --upgrade --upgrade-strategy only-if-needed git+https://github.com/Aider-AI/aider.git

If you have a chance to try it, let me know if it works for you.

[pcfreak30]

Thanks for trying aider and filing this issue.

The fix is available in the main branch. You can get it by installing the latest version from github:

aider --install-main-branch

# or...

python -m pip install --upgrade --upgrade-strategy only-if-needed git+https://github.com/Aider-AI/aider.git

If you have a chance to try it, let me know if it works for you.

Cool however I don't like that i need to pass a number of tokens. it should have a default and be toggle so i can switch easily.

[vphantom]

The fix is available in the main branch.

Do you mean the /thinking-tokens command? Does it fully toggle thinking back off, though? When I initially launch Aider, I see:

Main model: openrouter/anthropic/claude-3.7-sonnet with ask edit format, prompt cache, infinite output

Then if I run, say, /thinking-tokens 2k followed by /thinking-tokens 0 the line is not exactly back to the same state:

Main model: openrouter/anthropic/claude-3.7-sonnet with ask edit format, 0k think tokens, prompt cache, infinite output

I don't understand reasoning models and how Aider controls them well enough to know if these two states are completely equivalent.

[github-actions[bot]]

I'm labeling this issue as stale because it has been open for 2 weeks with no activity. If there are no additional comments, I will close it in 7 days.

Note: A bot script made these updates to the issue.

[pcfreak30]

not stale

[github-actions[bot]]

I'm labeling this issue as stale because it has been open for 2 weeks with no activity. If there are no additional comments, I will close it in 7 days.

Note: A bot script made these updates to the issue.

[pcfreak30]

not stale

[p-a-x]

Do you mean the /thinking-tokens command? Does it fully toggle thinking back off, though? When I initially launch Aider, I see:

Main model: openrouter/anthropic/claude-3.7-sonnet with ask edit format, prompt cache, infinite output

Then if I run, say, /thinking-tokens 2k followed by /thinking-tokens 0 the line is not exactly back to the same state:

Main model: openrouter/anthropic/claude-3.7-sonnet with ask edit format, 0k think tokens, prompt cache, infinite output

I don't understand reasoning models and how Aider controls them well enough to know if these two states are completely equivalent.

I would like to add my experience with the thinking-tokens:

I switched them on with claude-3.7-sonnet over amazon bedrock

That worked fine until the editor got into infinite-output mode , so I guess it's just resubmitting on the end of generation and this loop broke since bedrock/sonnet expected thinking but resubmitting just want to keep generating.

So I tried to deactivate thinking again via setting /thinking-tokens to 0 but this is not deactivating the thinking but just set them to 0 , so my bedrock thinking error remained and I had to resolve via restart.

The sentences

1. infinite output is incompatible with thinking.
2. once thinking is enabled it cannot be disabled

are philosophical correct but impractical. ;-)

As I understand this Issue here is related to sentence 2 and I'm yet to find an Issue for the first sentence.