@cotal-ai/delivery 0.131.2 → 0.131.3

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.
Files changed (2) hide show
  1. package/README.md +19 -0
  2. package/package.json +3 -3
package/README.md CHANGED
@@ -34,3 +34,22 @@ not permanent loss. This mode does not verify live copies, payload integrity or
34
34
  ID tracking costs O(posts × members) memory and extra callback work. Use it as a separate
35
35
  correctness workload; do not compare its CPU or memory numbers with ordinary load runs.
36
36
  `pnpm smoke:delivery-coverage` runs the accounting controls without starting a broker.
37
+
38
+ ## Active multi-space measurement
39
+
40
+ `COTAL_MSD_SECONDS=30 COTAL_MSD_RATE=2 pnpm smoke:delivery-multi-space --measure --measure-shape host16 --sublist-cache off`
41
+ runs the built delivery CLI against one native broker with 16 separate accounts and three members
42
+ per space. It reads every account's ready delivery lease before starting members; the host-up
43
+ message alone can describe partial startup. Use `--sublist-cache on` for the cache-enabled arm. `--measure-shape single16` selects
44
+ 16 delivery processes; the default `all` runs the historical 1/4/16/shared/single comparison on
45
+ reused stores. Select one shape per invocation for fresh-state comparisons. The broker comes from
46
+ `PATH`; pin its binary there when comparing runs. Duration and rate must be positive finite
47
+ numbers whose rounded-up product is a positive safe integer; validation precedes broker setup.
48
+
49
+ `MSD_RESULT` reports broker and delivery CPU seconds and RSS in MiB, elapsed window and per-space
50
+ accepted publishes, copies and endpoint errors. Measurements include a five-second allowance and
51
+ up to 30 more seconds for receiver completion. Counts and errors affect the exit status. Add
52
+ `--verify-delivery` for the same per-recipient acknowledged-ID checks as fanout-load; its tracking
53
+ has overhead, so use a separate correctness run. Repeated copies remain allowed. Receiver
54
+ completion is not proof of empty broker queues. The test does not model hosted agents, connect
55
+ auth callouts, subject churn or message ordering, and process RSS values are not additive.
package/package.json CHANGED
@@ -1,7 +1,7 @@
1
1
  {
2
2
  "name": "@cotal-ai/delivery",
3
3
  "description": "Cotal delivery daemon: the server-side Plane-3 durable backstop (fan-out writer + trusted reader + membership/ACL authority), a scoped least-privilege NATS client co-located with the broker.",
4
- "version": "0.131.2",
4
+ "version": "0.131.3",
5
5
  "license": "Apache-2.0",
6
6
  "repository": {
7
7
  "type": "git",
@@ -18,8 +18,8 @@
18
18
  }
19
19
  },
20
20
  "dependencies": {
21
- "@cotal-ai/core": "0.131.2",
22
- "@cotal-ai/workspace": "0.131.2",
21
+ "@cotal-ai/core": "0.131.3",
22
+ "@cotal-ai/workspace": "0.131.3",
23
23
  "@nats-io/jetstream": "^3.4.0"
24
24
  },
25
25
  "files": [