Integration guides
Aider
An open-source AI pair-programming tool in your terminal that edits code in your local git repo and commits automatically. Via LiteLLM it supports OpenAI-compatible endpoints — set the env vars and prefix the model name with openai/ to use XylemNode.
- 01
Install
The official recommendation is aider-install for a one-step setup — it installs aider in an isolated environment without polluting system Python. Print the version afterward to confirm.
# Install (all platforms) python -m pip install aider-install aider-install # Verify aider --version - 02
Configure the endpoint and key
Hand aider the XylemNode endpoint and key via environment variables; the address needs /v1. On Windows, run setx and reopen the terminal for it to take effect.
# Windows (PowerShell / CMD) setx OPENAI_API_BASE "https://api.xylemnode.com/v1" setx OPENAI_API_KEY "sk-your-key" # reopen the terminal after setx # macOS / Linux export OPENAI_API_BASE="https://api.xylemnode.com/v1" export OPENAI_API_KEY="sk-your-key" - 03
Enter the project and pick a model
Switch to your git project directory and launch with the model name prefixed by openai/. That prefix tells aider (via LiteLLM) to request in OpenAI-compatible mode; the model name after the prefix must exactly match one in the console's /v1/models.
cd /path/to/your/project aider --model openai/gpt-5.6-sol
Two notes
1) For unfamiliar model names, aider shows model warnings (missing context length, pricing and other metadata) — these are just reminders and don't affect usage, so continue. 2) To avoid setting env vars every time, put OPENAI_API_BASE and OPENAI_API_KEY in a .env at your project root, or fix them in ~/.aider.conf.yml.
Common questions
aider warns at startup that the model is unknown, with no context length or pricing
That is an expected notice, not an error. When the model name isn't in aider's built-in table it prints Unknown context window size and costs, using sane defaults and then treats the model as having an unlimited context window and costing nothing; the official warnings page says this one can usually be ignored. The cost is that it no longer manages context or estimates spend for you, so keep an eye on long sessions — and if you want the notice gone, supply the metadata through aider's advanced model settings. Two common misconfigurations while you're here: the model name must carry the openai/ prefix, which is what tells the underlying LiteLLM to speak OpenAI-compatible, and on Windows variables written with setx only apply after you reopen the terminal, which the same page calls out. If it answers with Did you mean…, it has read your model name as a typo — check the full identifier in the console.