Environment variables + --model flag
Native OpenAI-compatible supportAider with a custom OpenAI-compatible endpoint
Aider reads two environment variables and a model flag. Point them at the gateway and your git-aware pair programming loop runs on this catalog.
Base URL
https://uncensoredai.dev/v1
- 01
Install Aider
Aider is a Python package. Install it with pipx so it stays isolated from your project environments.
Installpipx install aider-chat - 02
Point it at the gateway
Set the base URL and key, then pass a catalog model id with the openai/ prefix so Aider uses its OpenAI client.
macOS / Linuxexport OPENAI_API_BASE="https://uncensoredai.dev/v1" export OPENAI_API_KEY="sk-ua-XXXX-XXXX-XXXX-XXXX" aider --model openai/deepseek-v4.1-flashWindows (PowerShell)$env:OPENAI_API_BASE = "https://uncensoredai.dev/v1" $env:OPENAI_API_KEY = "sk-ua-XXXX-XXXX-XXXX-XXXX" aider --model openai/deepseek-v4.1-flash - 03
Escalate for hard refactors
Aider lets you add a second model for planning. Use a reasoning tier for the architect role and a cheap tier for edits.
Architect modeaider --model openai/deepseek-v4.1-flash --architect --editor-model openai/gpt-6-sol
Commands and configuration below are shown exactly as the tool expects them, so they stay in English.
FAQ
Why the openai/ prefix on the model id?+
It tells Aider to route the request through its OpenAI-compatible client, which is what the gateway speaks.
Does Aider's repo map work?+
Yes. The repo map is built locally and sent as prompt context; it only needs a large enough context window on the model you choose.