/**
 * Compute a deterministic content hash for a Lambda@Edge asset bundle.
 *
 * Why this exists: `cloudfront.experimental.EdgeFunction` publishes a new
 * `AWS::Lambda::Version` whenever the asset hash changes. Passing a stable
 * `assetHash` with `assetHashType: AssetHashType.CUSTOM` pins the SOURCE
 * identity so unchanged inputs yield an unchanged hash and no new version.
 *
 * NOTE: a CUSTOM assetHash alone is not sufficient for determinism — CDK also
 * folds `JSON.stringify(bundling)` into the final hash (see aws-cdk-lib
 * `AssetStaging.calculateHash`), so any non-deterministic value inside the
 * `bundling` options (e.g. a random `volumes[].hostPath`) still churns the
 * hash. See {@link deterministicConfigDir}, which keeps that path stable.
 *
 * This helper produces a hash from the actual INPUTS — the source directory
 * contents, the generated config, and the optional pre-bundled deps directory —
 * so an unchanged input yields an unchanged hash and no new version. Pass the
 * result as `assetHash` with `assetHashType: AssetHashType.CUSTOM`.
 *
 * Files are hashed in sorted order for determinism. `requirements.txt` is
 * excluded because it is not shipped in the bundle (the bundler skips it).
 */
export declare function computeEdgeAssetHash(sourceDir: string, generatedConfig: string, bundledDepsDir?: string): string;
/**
 * Create a DETERMINISTIC temp directory holding the generated `config_generated.py`,
 * and return its path. Used as the `hostPath` of the Docker-fallback bundling
 * volume for a Lambda@Edge function.
 *
 * Why this must be deterministic: CDK derives an asset's final hash even under
 * `assetHashType: CUSTOM` by folding `JSON.stringify(bundling)` into the custom
 * hash (see aws-cdk-lib `AssetStaging.calculateHash`). `bundling.volumes[].hostPath`
 * is therefore part of the hash. A random `fs.mkdtempSync(...)` path changes on
 * every synth, so the asset hash — and thus the published S3 key — changes every
 * synth for byte-identical code, publishing a fresh `AWS::Lambda::Version` and
 * forcing a global CloudFront distribution update on EVERY deploy (and orphaning
 * the prior replicated edge version, which cannot be deleted for hours).
 *
 * Keying the directory on `key` (a per-function id) plus the config content keeps
 * it stable for unchanged inputs while staying unique per function.
 */
export declare function deterministicConfigDir(key: string, configPy: string): string;