Full routing test build 7

Preserve native OpenAI provider identity so existing Codex chats remain visible, while disabling WebSocket transport.

Full routing test build 6

Automatically upgrades older Codex integrations and disables Responses WebSockets; fixes ws://127.0.0.1:8787/v1/responses reconnects.

Full routing test build 5

Disable Codex Responses WebSocket transport for the local HTTP/SSE proxy; fixes ws://127.0.0.1:8787/v1/responses 404 reconnects.

Full routing test build 4

Fix HTTP/1.1 streaming disconnects and make TUI Ctrl+E stop match opencodex stop.

Full routing test build 3

Fixes Codex model casing normalization and updates the official catalog to the GPT-5.6 family.

Full routing test build 2

Fixes model-list routing: only explicit openai/, Provider/model, and intentional defaults are advertised. Prevents Codex reconnecting on unroutable catalog entries.

Full routing test build

Test build with native openai/ Responses passthrough and custom Provider/model routing. Existing TUI and provider/model features are preserved.

Codex app compatibility test

Test build from codex/codex-app-compat. This keeps the existing TUI and provider/model features while using the current Codex native-provider loopback integration.

OpenCodeX Windows executable

Standalone Windows executable. Download opencodex.exe, run it once from its folder with .\opencodex.exe install, then reopen your terminal to use opencodex commands from any folder.

Windows executable test build

Fixes standalone executable Codex integration. Ctrl+S now updates Codex config, refreshes the full model catalog, and restarts Codex. The executable now bundles the required catalog template.