@absolutejs/rag 0.22.0 → 0.22.1

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/CHANGELOG.md CHANGED
@@ -6,6 +6,12 @@ This file is generated by `absolute-changelog` from the entries in
6
6
  `changelog/`. Edit an entry, not this file — and add new ones under
7
7
  `changelog/unreleased/`.
8
8
 
9
+ ## 0.22.1 — 2026-09-24
10
+
11
+ ### Fixed
12
+
13
+ - **Retry owned-index initialization after a transient database outage instead of retaining a rejected initialization promise** (`createWebIndex`)
14
+
9
15
  ## 0.22.0 — 2026-09-24
10
16
 
11
17
  ### Added
package/WEB_INDEX.md CHANGED
@@ -104,3 +104,5 @@ Schedule retention of obsolete generations in the host database/storage lifecycl
104
104
  `WEB_INDEX_TEST_DATABASE_URL=... bun test tests/webIndex.test.ts tests/webIndexFlows.test.ts` uses an explicit disposable PostgreSQL database with pgvector. Tests clear the owned-index tables in that database. They cover restart, native retrieval filters, worker fencing, tenant isolation, bounded admission, robots failure, conditional requests, deletion, model-generation isolation, authorization and evidence attribution.
105
105
 
106
106
  `benchmarks/web-index/live.ts --live --output=NEW_PATH` declares four public pages and compares native lexical retrieval with Brave using identical extraction/review settings. It caps the run at eight research searches and sixteen model calls, preserves raw journals and reports cold/first and warm latency samples separately. It requires its own disposable database. Those pages were selected in advance; results are not web-wide or held-out discovery claims. No Exa parity is implied.
107
+
108
+ The final controlled 100-page lifecycle run measured 11.31 ms warm query p50 / 13.36 ms p95, 88.50 ms update-to-query visibility and 37.94 ms takedown-to-confirmation. It uses PostgreSQL 15, simulated public-page responses and deterministic vectors, not external embedding APIs. Run `benchmarks/web-index/lifecycle.ts` with a disposable database and `--output=NEW_PATH` to reproduce; do not generalize those numbers to public network/model latency.
package/changelog.json CHANGED
@@ -2,6 +2,19 @@
2
2
  "contract": 1,
3
3
  "name": "@absolutejs/rag",
4
4
  "releases": [
5
+ {
6
+ "changes": [
7
+ {
8
+ "kind": "fixed",
9
+ "summary": "Retry owned-index initialization after a transient database outage instead of retaining a rejected initialization promise",
10
+ "symbols": [
11
+ "createWebIndex"
12
+ ]
13
+ }
14
+ ],
15
+ "date": "2026-09-24",
16
+ "version": "0.22.1"
17
+ },
5
18
  {
6
19
  "changes": [
7
20
  {
@@ -118551,7 +118551,10 @@ var createWebIndex = (options) => {
118551
118551
  origins: [...origins].sort(),
118552
118552
  representation: 1
118553
118553
  })));
118554
- })();
118554
+ })().catch((error) => {
118555
+ initialization = undefined;
118556
+ throw error;
118557
+ });
118555
118558
  const resolve2 = async (id) => {
118556
118559
  await ready();
118557
118560
  const g = generations.get(id ?? await store.active(scope) ?? "");
@@ -125951,5 +125954,5 @@ export {
125951
125954
  webIndexPostgresSchemaSql
125952
125955
  };
125953
125956
 
125954
- //# debugId=F061712CE7DD66E064756E2164756E21
125957
+ //# debugId=A13572243F640B2064756E2164756E21
125955
125958
  //# sourceMappingURL=index.js.map