From 5f0a976d6daa93d8e5a4d5470becd00ce11d5442 Mon Sep 17 00:00:00 2001 From: Aaron Queen Date: Fri, 9 Oct 2026 21:06:44 -0600 Subject: [PATCH] docs: fold the unreleased notes into 0.5.2 and align the benchmark remeasure line 0.5.2 is bumped on main but was never tagged or published (tags stop at v0.5.1, npm has no 0.5.2), so the three entries under Unreleased belong in its section. docs/benchmarks.md said the results had not been remeasured for 0.4.0 only, while the README lists 0.4.0 through 0.5.2; both now list the same versions. --- CHANGELOG.md | 22 ++++++++++------------ docs/benchmarks.md | 2 +- 2 files changed, 11 insertions(+), 13 deletions(-) diff --git a/CHANGELOG.md b/CHANGELOG.md index bbacb39..8d2caef 100644 --- a/CHANGELOG.md +++ b/CHANGELOG.md @@ -2,18 +2,6 @@ Notable changes to Swarmail. Versions follow [semantic versioning](https://semver.org). -## Unreleased - -- `swarmail who` asks git about each running T3 thread directory once instead of once per thread. With about 800 - running threads in 149 directories it took 0.3 s instead of 1.6 s. - -- `fetch_inbox` and `fetch_session_inbox` no longer read each message body from storage when `include_bodies` is off, - which is the default. A page of 20 messages took 48 µs instead of 57 µs with 2 KB bodies and 52 µs instead of 82 µs - with 17 KB bodies. - -- The `swarmail-mcp` README lists the memory each launch command costs: about 180 MiB for `npx -y swarmail-mcp`, 82 MiB - for `bunx swarmail-mcp` and 46 MiB for `bunx --bun swarmail-mcp`, which runs the same relay under Bun. - ## 0.5.2 - 2026-10-09 - Require agents to record whole-session closeouts and retire their own project @@ -26,6 +14,16 @@ Notable changes to Swarmail. Versions follow [semantic versioning](https://semve - The release check fails for a package that has no row in the listings section of `CONTRIBUTING.md`, so a new package cannot ship without a listing decision. +- `swarmail who` asks git about each running T3 thread directory once instead of once per thread. With about 800 + running threads in 149 directories it took 0.3 s instead of 1.6 s. + +- `fetch_inbox` and `fetch_session_inbox` no longer read each message body from storage when `include_bodies` is off, + which is the default. A page of 20 messages took 48 µs instead of 57 µs with 2 KB bodies and 52 µs instead of 82 µs + with 17 KB bodies. + +- The `swarmail-mcp` README lists the memory each launch command costs: about 180 MiB for `npx -y swarmail-mcp`, 82 MiB + for `bunx swarmail-mcp` and 46 MiB for `bunx --bun swarmail-mcp`, which runs the same relay under Bun. + ## 0.5.1 - 2026-10-09 - A session whose usual name is already held by another session in a project diff --git a/docs/benchmarks.md b/docs/benchmarks.md index 7fc10af..b93f442 100644 --- a/docs/benchmarks.md +++ b/docs/benchmarks.md @@ -10,7 +10,7 @@ The results cover this workload only. They say nothing about durability after a crash, large archives or production load. Swarmail was measured at [`67e187a`](https://github.com/bompus/swarmail/tree/67e187a6b4beb7319c6ab30466448742297e2a01). -These results have not been remeasured for 0.4.0. +These results have not been remeasured for 0.4.0, 0.5.0, 0.5.1 or 0.5.2. ## Summary