Aider
Aider connects to any OpenAI-compatible endpoint through two environment
variables and the openai/ model prefix. Step-by-step guide with live model
picks: Aider setup.
Fastest: one command
npx uberllm init --tool aider
This connects your account in the browser, saves the key to a file only you can
read and writes ~/.aider.conf.yml and the model settings file (add --project for one repo). The file is backed up first, and
npx uberllm init --undo puts it back. Details: uberllm init. The
steps to do it by hand are below.
Setup
macOS / Linux:
export OPENAI_API_BASE=https://api.martllm.com/v1
export OPENAI_API_KEY=ull_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
Windows:
setx OPENAI_API_BASE https://api.martllm.com/v1
setx OPENAI_API_KEY ull_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
(Open a new terminal after setx.)
Run
Prefix the model with openai/ so Aider routes it through the OpenAI-compatible
path. Only that first openai/ is removed; the rest is the MartLLM id:
aider --model openai/deepseek/deepseek-v4.1-flash
Edit format and repo map
Aider chooses edit settings by model name. For a name it does not know it
rewrites whole files and turns the repo map off. Tell it better settings in
.aider.model.settings.yml (home folder or repo root); name must match
--model exactly:
- name: openai/deepseek/deepseek-v4.1-flash
edit_format: diff
use_repo_map: true
Notes
- Aider prints a model warning for models it doesn't know (context size and
price). It does not stop it from working. Add the numbers in
.aider.model.metadata.json, or setshow-model-warnings: false. - To persist settings, put the two variables in a
.envfile in your repo, or use~/.aider.conf.ymlwithopenai-api-base,openai-api-keyandmodel. - Set a cheap
--weak-model(commit messages, summaries), for example--weak-model openai/deepseek/deepseek-v4.1-flash. - Each call is sent to an available third-party provider for the model (see Models & routing).