Step 1
Add the upstream after setup
Wait for manual delivery, then add the unlimitedcodex base URL, API key, and model IDs to your proxy or routing config.
Integration
Point LiteLLM-style proxy workflows at a delivered upstream base URL.
For teams that already use proxy or router layers, unlimitedcodex can be evaluated as a packaged OpenAI-compatible upstream during build and test windows.
- OpenAI-compatible upstream pattern
- Model-list and endpoint smoke testing
- Manual setup with delivered base URL and key
- 4 concurrent connections for controlled routing tests
Setup steps
Step 2
Validate routing and endpoint coverage
Test /models, chat/completions, embeddings, and image generation where supported before routing real workloads.
Step 3
Keep the router honest
Make sure the proxy exposes provider errors, retry counts, and the 4 concurrent connection boundary so failed loops do not hide inside the router.
Example configuration
# Example LiteLLM-style upstream values api_base: https://your-workspace.unlimitedcodex.com/v1 api_key: $UNLIMITEDCODEX_API_KEY model: codex-5.5 notes: keep concurrency at or below 4 connections
FAQ
Is unlimitedcodex a replacement for LiteLLM?
No. LiteLLM is proxy infrastructure. unlimitedcodex is an independent API access package that can be evaluated as an upstream provider.
What should proxy users monitor?
Monitor model ID mapping, retry count, provider errors, endpoint support, and concurrent requests so the proxy does not hide provider behavior.