/**
 * AWS Signature Version 4, by hand.
 *
 * **Why by hand.** `@trazum/core` has zero runtime dependencies and a test that
 * fails the build if one appears. That is a security property rather than a
 * packaging preference: this library reads people's prompts, and every
 * dependency is somebody else's code reading them too. The AWS SDK is roughly
 * a hundred packages to sign one request.
 *
 * **WebCrypto, not `node:crypto`.** The browser-safe entry point cannot reach a
 * Node builtin — `apps/web` bundles this library, and a single `node:` import
 * anywhere in that graph fails the build. `crypto.subtle` exists in both, so
 * signing works in a browser and in the CLI without a second implementation.
 * Everything here is therefore async, which HMAC-SHA256 does not need to be and
 * `crypto.subtle` insists on anyway.
 *
 * **What is not asserted, and it matters.** There is no AWS-published
 * known-answer vector in the tests. This environment cannot reach the internet
 * to fetch one and cannot reach AWS to try a real call, so the tests check the
 * canonical strings this builds — which are derivable from the specification by
 * reading — plus the cryptographic properties that any correct signer has. A
 * systematically wrong canonicalisation that is wrong *consistently* would pass
 * all of them. The first real request against Bedrock is the proof, and until
 * somebody makes one this is careful code rather than verified code.
 */
/**
 * `20260810T140102Z` and `20260810`, from one instant.
 *
 * Both derive from the same `Date` rather than being formatted twice. The
 * classic SigV4 bug is a request signed at 23:59:59.9 whose date stamp comes
 * from the next day: the credential scope and the timestamp disagree by one
 * day, AWS rejects it, and it happens roughly once per thousand requests at
 * midnight UTC and never in a test.
 */
export declare function amzDates(now: Date): {
    amzDate: string;
    dateStamp: string;
};
/**
 * The signing key: four chained HMACs, each keyed by the last.
 *
 * The chain is what scopes a signature to one day, one region and one service,
 * so a leaked signature is not a leaked credential. `AWS4` prefixes the secret
 * and is part of the specification rather than decoration.
 */
export declare function signingKey(secretAccessKey: string, dateStamp: string, region: string, service: string): Promise<ArrayBuffer>;
export interface SignInput {
    method: string;
    /** Already-encoded path, e.g. `/model/anthropic.claude-v2/invoke`. */
    path: string;
    host: string;
    region: string;
    service: string;
    body: string;
    accessKeyId: string;
    secretAccessKey: string;
    /** Present for temporary credentials, and signed when it is. */
    sessionToken?: string;
    now: Date;
}
/**
 * The headers a signed request carries.
 *
 * Returned rather than mutated onto a request, so this is a pure function of its
 * input and a test can read every byte of what it produced.
 */
export interface SignedHeaders {
    authorization: string;
    'x-amz-date': string;
    'x-amz-content-sha256': string;
    'x-amz-security-token'?: string;
}
/**
 * Builds the canonical request, exactly as the specification orders it.
 *
 * Exported for the tests, because this string is where a signer goes wrong.
 * Every field is joined with a newline, headers are lowercased and sorted, and
 * the trailing newline after the header block is required — omit it and every
 * signature is wrong in a way whose only symptom is `403 SignatureDoesNotMatch`.
 */
export declare function canonicalRequest(input: SignInput): Promise<{
    canonical: string;
    signedHeaderNames: string;
    payloadHash: string;
}>;
/** The full `Authorization` header value, and the headers that go with it. */
export declare function signRequest(input: SignInput): Promise<SignedHeaders>;
//# sourceMappingURL=aws-sigv4.d.ts.map