Inquiry: How does Aider handle my code/data

[guyisra]

Issue

1. does the code gets sent directly to LLM api? or to some server on the way?
2. if not direct to LLM, what is being retained by the servers?
3. What is the general recommendations on handling private code?

Thanks

Version and model info

No response

[PACHAKUTlQ]

1. All your code are only sent to the LLM provider you configured. There is no such thing like Aider server, if that is what you mean. Aider is running purely locally, and make requests to the server of your configured LLM provider directly. The code is open-source, you can check them.
2. Aider enables analytics by default. Though no private data or code are sent to analytics server, you can easily disable analytics. I guess the analytics include the model name, how much code you generated, how many failed attempts of generating code, how much code are generated in wrong format and such things, your code are definitely not included in analytics.
3. You can disable analytics. Then only your LLM provider will get your data. If you are even not OK with that, use a local AI model, using ollama or vllm to run.

(I am an Aider user, and have nothing to do with Aider dev team)

[paul-gauthier]

Thanks for trying aider and filing this issue.

There are no aider servers. Aider runs locally and only sends your code to the LLMs you configure it to use.

Aider does NOT enable analytics by default. You must explicitly opt-in to sharing analytics. Analytics are anonymized and do not share any sensitive information like prompts, code, keys, etc.

These docs may be helpful:

https://aider.chat/docs/more/analytics.html
https://aider.chat/docs/legal/privacy.html