OpenCodex 1.2.0 Decouples Direct Models from External Gateway
OpenCodex version 1.2.0 has been released, introducing an update that decouples direct models like GPT from the gateway used for external models like Gemini and DeepSeek. Previously, all models routed through the gateway, meaning a gateway failure would also take down direct models. This update prevents single-point-of-failure issues, ensuring that users can still access direct models like GPT even if the gateway for external models experiences downtime. It improves the overall reliability and resilience of the local proxy tool for developers. By separating the routing paths, OpenCodex 1.2.0 ensures that local or direct API calls to OpenAI's GPT models remain functional independently of external API integrations.
## BACKGROUND
OpenCodex is a local proxy tool designed to run Codex and Claude Code by translating wire formats to various LLM providers. An LLM gateway is typically used to route, manage, and analyze API requests across multiple providers like OpenAI, Anthropic, and Google through a unified interface.