aider

aider is a CLI pair-programmer that speaks the OpenAI API. Point it at VCodeX by overriding the base URL.

Prerequisites

• aider installed (pip install aider-chat or pipx install aider-chat).
• A VCodeX API key (sk-...).

1 · Get your VCodeX key

Dashboard → Settings → External Keys → Create key.

2 · Configure aider

Export the env vars aider reads:

export OPENAI_API_BASE=https://app.vcodex.ai/v1
export OPENAI_API_KEY=sk-...

Then run aider with an explicit model id:

aider --model openai/vcodex/auto

To persist the choice, add it to ~/.aider.conf.yml:

openai-api-base: https://app.vcodex.ai/v1
model: openai/vcodex/auto

(Keep OPENAI_API_KEY in the environment — don't commit it to the config file.)

3 · Verify

aider --model openai/vcodex/auto --message "print a haiku about YAML"

Expected: a haiku streams back. Confirm in Observability → Requests with client_source: aider.

Troubleshooting

• litellm.NotFoundError — prefix the model with openai/ so LiteLLM routes it through the OpenAI-compatible driver.
• Rate limit exceeded — VCodeX keys are 120 req/min. aider can burst during repo-map indexing; slow it down with --map-tokens 512 or add a second key.
• Diffs applied but tokens double-count — aider retries on parse failures; see the correlation IDs in error logs.