@bitkyc08/opencodex
Version:
Universal provider proxy for OpenAI Codex & Claude Code — use any LLM with Codex CLI/App/SDK and Claude Code
71 lines (63 loc) • 3.59 kB
text/typescript
import { CODEX_TEXT_GUARDED_BUDGET_POLICY, createRequestExecutionBudget, type RequestExecutionBudget } from "../../lib/request-execution-budget";
import { TRANSIENT_RETRY_MAX_ATTEMPTS, type TransientSendBudget } from "../../lib/upstream-retry";
import { GENERIC_OAUTH_MAX_ACCOUNTS_PER_REQUEST } from "../../oauth/generic-account-failover";
import type { RequestLogContext } from "../request-log";
import { attachRequestSpendTracker } from "../responses/request-spend";
import type { SpendAdmissionPolicy } from "../../lib/spend-reservation-ledger";
// Only ingress owns an expandable policy. Exact caller budgets and combo-derived scopes
// never enter this map, even when their policy happens to equal the default profile.
const oauthExpanded = new WeakSet<TransientSendBudget>();
const ingressPolicies = new WeakMap<TransientSendBudget, {
maxTotalModelSends: number;
baseSendAllowance: number;
}>();
/**
* The one construction of an ingress-owned send budget: the default guarded policy, no logical
* request id, and this request's spend tracker as the send observer. Attaching the tracker
* parks it on `logCtx`, so callers that inherit a holder must not call this at all.
*/
export function createInferenceSendBudget(
req: Pick<Request, "headers">,
logCtx: RequestLogContext,
capturedPolicy?: SpendAdmissionPolicy,
): RequestExecutionBudget {
const policy = { ...CODEX_TEXT_GUARDED_BUDGET_POLICY };
const budget = createRequestExecutionBudget(policy, undefined, attachRequestSpendTracker(req, logCtx, undefined, capturedPolicy));
ingressPolicies.set(budget, policy);
return budget;
}
/**
* Fund each account's normal transient ladder once, before the first physical send. The account
* count is clamped here as well as at the caller, so no caller can raise the ceiling past
* GENERIC_OAUTH_MAX_ACCOUNTS_PER_REQUEST × TRANSIENT_RETRY_MAX_ATTEMPTS.
*/
export function expandInferenceOAuthSendBudget(budget: TransientSendBudget | undefined, accounts: number): void {
const fundedAccounts = Number.isSafeInteger(accounts) && accounts >= 2 ? Math.min(accounts, GENERIC_OAUTH_MAX_ACCOUNTS_PER_REQUEST) : 0;
expandIngressSendBudget(budget, fundedAccounts * TRANSIENT_RETRY_MAX_ATTEMPTS);
}
/** Fund an explicitly configured standalone ChatGPT retry ladder before any send. */
export function fundChatGptRetryAttempts(budget: TransientSendBudget | undefined, attempts: number): void {
if (Number.isInteger(attempts) && attempts > TRANSIENT_RETRY_MAX_ATTEMPTS && attempts <= 10) {
expandIngressSendBudget(budget, attempts);
}
}
function expandIngressSendBudget(budget: TransientSendBudget | undefined, sends: number): void {
if (!budget) {
return;
}
const policy = ingressPolicies.get(budget);
if (!policy || oauthExpanded.has(budget) || budget.used !== 0 || ("physicalStarted" in budget && Number(budget.physicalStarted) > 0)) {
return;
}
// One ingress expansion per request; the policy stays for the hosted-search allowance.
oauthExpanded.add(budget);
policy.baseSendAllowance = Math.max(policy.baseSendAllowance, sends);
policy.maxTotalModelSends = Math.max(policy.maxTotalModelSends, sends);
}
/** Fund the hosted loop's bounded model sends before its first dispatch. */
export function expandInferenceSearchSendBudget(budget: TransientSendBudget | undefined, baseSends: number): void {
const policy = budget && ingressPolicies.get(budget);
if (!policy || baseSends <= policy.baseSendAllowance) return;
policy.maxTotalModelSends += baseSends - policy.baseSendAllowance;
policy.baseSendAllowance = baseSends;
}