@bitkyc08/opencodex
Version:
Universal provider proxy for OpenAI Codex & Claude Code — use any LLM with Codex CLI/App/SDK and Claude Code
34 lines (31 loc) • 1.83 kB
text/typescript
/**
* Grok Build sends its conversation identity as `x-grok-conv-id` (and the same value as the body
* `prompt_cache_key`), but never the `session_id` header Codex clients send. The ChatGPT Codex
* backend only reuses a warmed prompt prefix for requests that carry `session_id`, and opencodex
* reads the same header for thread affinity and conversation logging. Without it, every Grok turn
* routed to an OpenAI model was a cold `new_bind` with near-zero cached input.
*
* Only the managed Grok surface (`x-opencodex-grok: 1`, written by `ocx` into Grok's config) is
* promoted, and an explicit caller session header always wins.
*/
const GROK_SURFACE_HEADER = "x-opencodex-grok";
const GROK_CONVERSATION_HEADER = "x-grok-conv-id";
const SESSION_HEADERS = ["session_id", "session-id", "thread-id"] as const;
// Grok emits UUIDs and `turn-summary-<uuid>`; anything else is not promoted.
const SAFE_CONVERSATION_ID = /^[A-Za-z0-9][A-Za-z0-9._:-]{0,127}$/;
export function grokConversationSessionId(headers: Headers): string | undefined {
if (headers.get(GROK_SURFACE_HEADER) !== "1") return undefined;
if (SESSION_HEADERS.some(name => headers.has(name))) return undefined;
const conversationId = headers.get(GROK_CONVERSATION_HEADER)?.trim();
if (!conversationId || !SAFE_CONVERSATION_ID.test(conversationId)) return undefined;
return conversationId;
}
/** Returns `req` unchanged unless a Grok conversation id can stand in for the missing `session_id`. */
export function withGrokSessionIdentity(req: Request): Request {
const sessionId = grokConversationSessionId(req.headers);
if (!sessionId) return req;
const headers = new Headers(req.headers);
headers.set("session_id", sessionId);
// The clone keeps the unread body stream and the caller's abort signal.
return new Request(req, { headers });
}