scavio 0.23.0 → 0.23.1

package/README.md

@@ -126,7 +126,7 @@ map.data.truncated;   // true when more URLs exist than `limit`
 // Crawl: start a job, then poll it (polling and cancelling are free).
 const job = await client.extract.crawl(
   { url: "https://example.com/docs", max_pages: 50, max_depth: 3, format: "markdown" },
-  { idempotencyKey: "docs-crawl-2026-10-12" }, // optional, sent as Idempotency-Key
+  { idempotencyKey: "docs-crawl-2026-10-12" }, // optional; one is generated when omitted
 );
 job.data.id;          // "crawl_..."
 job.data.status_url;
@@ -185,16 +185,21 @@ call. Results are kept for 3 days after the job finishes (`expires_at`), up to
 **`wait: true`.** `crawl()` and `batch()` poll the status endpoint every
 `pollInterval` ms (default 2,000) until the job is completed, failed or
 cancelled, then page through every cursor and return
-`{ data: { ...summary, results: [all], next_cursor: null } }`. If the job is
+`{ data: { ...summary, results: [all], next_cursor: null }, created: {...} }`,
+where `created` is the 202 job handle (including `webhook_secret`, which is
+only returned on creation), as in the Python SDK (since 0.23.1). If the job is
 still running after `waitTimeout` ms (default 10 minutes) they throw
 `ScavioJobTimeoutError`; the job keeps running and the error carries
 `jobId`, `jobType` and `statusUrl` so you can keep polling or cancel it. For
 big crawls, prefer a webhook or your own polling loop.
 
-**Retries and idempotency.** The SDK retries 429/5xx and network errors. Pass
-`idempotencyKey` when starting a job: repeating a POST with the same key within
-24 hours returns the original job, so a retried request never starts a second
-one.
+**Retries and idempotency.** The SDK retries 429/5xx and network errors.
+`crawl()` and `batch()` always send an `Idempotency-Key`: yours when you pass
+`idempotencyKey`, otherwise a random UUID generated for that call and reused on
+every retry of it (since 0.23.1). Repeating a POST with the same key within 24
+hours returns the original job, so a retried start never begins a second,
+billed one. Pass your own key to dedupe across calls or processes. Plain
+`extract()` and `map()` are not jobs and send no key.
 
 #### Completion webhooks
 

package/dist/index.cjs

@@ -232,7 +232,7 @@ function parseRetryAfter(header) {
 var BASE_URL = "https://api.scavio.dev";
 var DEFAULT_TIMEOUT = 3e4;
 var DEFAULT_MAX_RETRIES = 2;
-var SDK_VERSION = true ? "0.23.0" : "0.0.0";
+var SDK_VERSION = true ? "0.23.1" : "0.0.0";
 var USER_AGENT = `scavio-js/${SDK_VERSION}`;
 function isServerRuntime() {
   const g = globalThis;
@@ -2999,6 +2999,12 @@ var COLLECT_PAGE_SIZE = 100;
 function sleep2(ms) {
   return new Promise((resolve) => setTimeout(resolve, ms));
 }
+async function randomUUID() {
+  const g = globalThis;
+  if (typeof g.crypto?.randomUUID === "function") return g.crypto.randomUUID();
+  const nodeCrypto = await import("crypto");
+  return nodeCrypto.randomUUID();
+}
 var ExtractNamespace = class {
   constructor(client) {
     this.client = client;
@@ -3038,14 +3044,15 @@ var ExtractNamespace = class {
     return this.cancel("batch", id);
   }
   async start(type, body, request2) {
-    const headers = request2.idempotencyKey ? { "Idempotency-Key": request2.idempotencyKey } : void 0;
+    const headers = { "Idempotency-Key": request2.idempotencyKey ?? await randomUUID() };
     const created = await this.client._post(
       `/api/v1/extract/${type}`,
       body,
       headers
     );
     if (!request2.wait) return created;
-    return this.waitFor(type, created.data.id, created.data.status_url, request2);
+    const done = await this.waitFor(type, created.data.id, created.data.status_url, request2);
+    return { ...done, created: created.data };
   }
   async status(type, id, query = {}) {
     return await this.client._get(`/api/v1/extract/${type}/${encodeURIComponent(id)}`, {