/**
 * Anthropic Messages API fallback over the pooled Codex Responses transport.
 *
 * This module deliberately contains only wire-format conversion and buffered
 * SSE parsing. Account selection, OAuth, cooldowns, and quota persistence stay
 * in the native Codex proxy handler so fallback traffic follows the same pool
 * rules as a native Codex request.
 */
import type { ClaudeRequest, CodexFallbackResult, CodexFallbackStream, CodexReasoningEffort, CodexResponsesRequest } from "../types/index.js";
export declare class CodexFallbackResponseError extends Error {
    readonly status: number;
    readonly responseBody: string;
    readonly code?: string;
    readonly retryable?: boolean;
    constructor(status: number, responseBody: string);
}
/** A provider terminal event, preserved through the format adapter. */
export declare class CodexFallbackStreamError extends Error {
    readonly status: number;
    readonly code: string;
    readonly retryable?: boolean;
    readonly usage?: CodexFallbackResult["usage"];
    constructor(payload: Record<string, unknown>, eventType: string, usage?: CodexFallbackResult["usage"]);
}
/**
 * Routing key that pins a request to the cache already holding its prefix.
 *
 * Anthropic caching is explicit — the client marks breakpoints and the provider
 * honours them. OpenAI caching is automatic: a prefix is reused only when the
 * request lands on infrastructure already holding it, and `prompt_cache_key` is
 * what pins a request to one. Without it this hop takes its chances on routing,
 * which is the worst case for a fallback that arrives in bursts separated by
 * long gaps.
 *
 * The key must therefore describe the *prefix*, not the caller. Keying it by
 * session id gives every session a private namespace, so N agents that share
 * one system prompt and tool set each pay that shared prefix in full rather
 * than paying it once — the cost is linear in fan-out width and invisible,
 * because each request looks individually well-behaved. Across 42 captured
 * fallback bodies the prefixes collapse to 10 distinct values, the largest
 * shared by 12 requests belonging to different sessions.
 *
 * So derive the key from the prefix itself: the instructions and the tool
 * declarations, which are exactly the leading bytes the upstream caches and
 * the only part stable across a conversation's turns. Identical prefixes then
 * share one cache across sessions and subagents, a prefix that genuinely
 * differs still gets its own key so nothing unrelated contends, and turns
 * within one session keep sharing a key because neither input changes between
 * them.
 *
 * A request carrying neither instructions nor tools has no prefix worth
 * pinning, and falls back to the session id so its turns at least stay
 * together. With neither available the field is omitted rather than filled
 * with something shared: a key common to every request would pin unrelated
 * prefixes onto one cache instead of separating them.
 *
 * The digest covers only content already sent upstream in `instructions` and
 * `tools`; no account, device or session identifier reaches the wire.
 */
export declare function codexPromptCacheKey(body: ClaudeRequest): string | undefined;
/** Convert a Claude Messages request into the ChatGPT Codex Responses shape. */
export declare function convertClaudeRequestToCodex(body: ClaudeRequest, model: string, reasoningEffort?: CodexReasoningEffort): CodexResponsesRequest;
/**
 * Parse a complete Codex Responses SSE stream before emitting Claude output.
 *
 * A missing terminal event, malformed JSON, terminal error, or empty response
 * is rejected. That makes it safe for the caller to try the next configured
 * fallback without ever replaying output already sent to a client.
 */
export declare function parseCodexFallbackSSE(sse: string): CodexFallbackResult;
/** Consume and validate a native Codex response before producing Claude output. */
export declare function consumeCodexFallbackResponse(response: Response): Promise<CodexFallbackResult>;
/**
 * Translate a Codex stream as events arrive. A completed tool call is emitted
 * once its arguments validate; text does not wait for response.completed.
 * The caller owns error framing and must never retry after emitting output.
 */
export declare function createCodexFallbackStream(response: Response, model: string): Promise<CodexFallbackStream>;