sheawinkler/ContextLattice

v3.3.23 Feature

This release adds 2 notable features for engineering teams evaluating rollout.

Published 20d AI Agents & Assistants

View tool

✓ No known CVEs patched

Read the diff → Tool health → What is this tool? →

✓ No known CVEs patched in this version

Topics

agent-framework agent-orchestration ai-infra ai-interface context-engineering context-management

+4 more

go long-horizon-agents long-horizon-intelligence rust

Summary

AI summary

Added LZ4 archive codec, mmap controls, optional .rkyv lineage cache sidecar, plus new performance crates.

Changes in this release

Type	Severity	Summary	CVE
Feature
Feature	Medium	Added five performance/storage crates to `context_storage_ops` (memmap2, lz4_flex, rkyv, ahash, crossbeam-channel). Added five performance/storage crates to `context_storage_ops` (memmap2, lz4_flex, rkyv, ahash, crossbeam-channel). Source: llm_adapter@2026-05-21 Confidence: high	—
Feature	Medium	Added archive codec/mmap controls (`--codec gzip\|lz4`, `--use-mmap true\|false`). Added archive codec/mmap controls (`--codec gzip\|lz4`, `--use-mmap true\|false`). Source: llm_adapter@2026-05-21 Confidence: high	—
Feature	Medium	Added optional `.rkyv` lineage cache sidecar for weekly lineage counts. Added optional `.rkyv` lineage cache sidecar for weekly lineage counts. Source: llm_adapter@2026-05-21 Confidence: high	—
Feature	Medium	Extended perf harness to benchmark both gzip and lz4 archive paths. Extended perf harness to benchmark both gzip and lz4 archive paths. Source: llm_adapter@2026-05-21 Confidence: high	—
Feature	Medium	New lane `archive_ndjson_lz4` added with benchmark results (trimmed mean 0.2367s, p50 0.2313s, p95 0.2834s). New lane `archive_ndjson_lz4` added with benchmark results (trimmed mean 0.2367s, p50 0.2313s, p95 0.2834s). Source: llm_adapter@2026-05-21 Confidence: high	—
Performance
Performance	Medium	Ledger performance improved by -38.08% (trimmed mean) vs v3.3.22. Ledger performance improved by -38.08% (trimmed mean) vs v3.3.22. Source: llm_adapter@2026-05-21 Confidence: high	—
Performance	Medium	Weekly lineage dry run speed increased by -62.76% (trimmed mean). Weekly lineage dry run speed increased by -62.76% (trimmed mean). Source: llm_adapter@2026-05-21 Confidence: high	—
Performance	Medium	Cold pack dry run speed improved by -53.68% (trimmed mean). Cold pack dry run speed improved by -53.68% (trimmed mean). Source: llm_adapter@2026-05-21 Confidence: high	—
Performance	Medium	Cold tier dry run speed increased by -37.01% (trimmed mean). Cold tier dry run speed increased by -37.01% (trimmed mean). Source: llm_adapter@2026-05-21 Confidence: high	—
Performance	Medium	Archive NDJSON gzip operation faster by -69.06% (trimmed mean). Archive NDJSON gzip operation faster by -69.06% (trimmed mean). Source: llm_adapter@2026-05-21 Confidence: high	—
Performance	Medium	Fanout GC dry run speed improved by -31.01% (trimmed mean). Fanout GC dry run speed improved by -31.01% (trimmed mean). Source: llm_adapter@2026-05-21 Confidence: high	—
Refactor	Medium	Switched hot set/map paths to `ahash` and cold-pack rayon collection to `crossbeam-channel`. Switched hot set/map paths to `ahash` and cold-pack rayon collection to `crossbeam-channel`. Source: llm_adapter@2026-05-21 Confidence: high	—

Full changelog

What changed

Added five performance/storage crates to context_storage_ops: memmap2, lz4_flex, rkyv, ahash, crossbeam-channel.
Added archive codec/mmap controls (--codec gzip|lz4, --use-mmap true|false).
Added optional .rkyv lineage cache sidecar for weekly lineage counts.
Switched hot set/map paths to ahash and cold-pack rayon collection to crossbeam-channel.
Extended perf harness to benchmark both gzip and lz4 archive paths.

Validation

cargo fmt --manifest-path crates/Cargo.toml --all
cargo test --manifest-path crates/Cargo.toml -p context_storage_ops
cargo build --manifest-path crates/Cargo.toml -p context_storage_ops --release

Perf snapshot (vs v3.3.22 tranche baseline)

Numbers use pooled warm-run timing from /Volumes/wd_black/contextlattice/tmp/storage_ops_perf_suite/results (trimmed mean / p50 / p95):

ledger: -38.08% / -56.11% / +285.56% (single heavy-tail transient observed; 10-run ledger check p50 ~0.76s, p95 ~3.25s)
weekly_lineage_dry: -62.76% / -70.13% / -67.35%
cold_pack_dry: -53.68% / -55.04% / -56.40%
cold_tier_dry: -37.01% / -33.71% / -56.06%
archive_ndjson_gzip: -69.06% / -70.58% / -71.82%
fanout_gc_dry: -31.01% / -40.31% / -67.14%
new lane archive_ndjson_lz4: trimmed mean 0.2367s, p50 0.2313s, p95 0.2834s

Notes

Runtime behavior remains Go/Rust-only on strict lanes; this release updates storage/maintenance transforms and benchmark coverage.

View diff on GitHub

Weekly OSS security release digest.

The CVE patches and breaking changes that affected production tools this week. One email, every Sunday.

No spam, unsubscribe anytime.

Share this release

Share on X Share on Bluesky

Track sheawinkler/ContextLattice

Get notified when new releases ship.

About sheawinkler/ContextLattice

Private-by-default memory and context layer for agents with Go/Rust runtime, staged retrieval across fused data backends, and long-horizon context continuity.

All releases →