sheawinkler/ContextLattice

v3.3.22 Breaking

This release includes breaking changes for platform teams planning a safe upgrade.

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

Storage operation performance improvements across multiple workloads

Changes in this release

Type	Severity	Summary	CVE
Feature	Medium	JSON parse fast-path added via simd-json with serde fallback. JSON parse fast-path added via simd-json with serde fallback. Source: llm_adapter@2026-05-21 Confidence: high	—
Feature	Medium	Added repo-agnostic benchmark harness bench/context_storage_ops_perf_tranche.py. Added repo-agnostic benchmark harness bench/context_storage_ops_perf_tranche.py. Source: llm_adapter@2026-05-21 Confidence: high	—
Dependency	Medium	Content hashing migrated from SHA-256 to blake3. Content hashing migrated from SHA-256 to blake3. Source: llm_adapter@2026-05-21 Confidence: high	—
Performance
Performance	Medium	Async HTTP runtime client for storage ops using tokio, reqwest-middleware, and reqwest-retry implemented. Async HTTP runtime client for storage ops using tokio, reqwest-middleware, and reqwest-retry implemented. Source: llm_adapter@2026-05-21 Confidence: high	—
Performance	Medium	Snapshot processing parallelized with rayon bounded worker pool. Snapshot processing parallelized with rayon bounded worker pool. Source: llm_adapter@2026-05-21 Confidence: high	—
Performance	Medium	Ledger operation performance improved by 87.08% (p50 from 2.1514s to 0.2781s). Ledger operation performance improved by 87.08% (p50 from 2.1514s to 0.2781s). Source: llm_adapter@2026-05-21 Confidence: high	—
Performance	Medium	Weekly lineage dry run performance improved by 59.20% (p50 from 1.1010s to 0.4492s). Weekly lineage dry run performance improved by 59.20% (p50 from 1.1010s to 0.4492s). Source: llm_adapter@2026-05-21 Confidence: high	—
Performance	Medium	Cold pack dry operation performance improved by 65.62% (p50 from 1.1135s to 0.3829s). Cold pack dry operation performance improved by 65.62% (p50 from 1.1135s to 0.3829s). Source: llm_adapter@2026-05-21 Confidence: high	—
Performance	Medium	Cold tier dry operation performance improved by 47.13% (p50 from 0.5472s to 0.2893s). Cold tier dry operation performance improved by 47.13% (p50 from 0.5472s to 0.2893s). Source: llm_adapter@2026-05-21 Confidence: high	—
Performance	Medium	Archive NDJSON operation performance improved by 36.84% (p50 from 1.1530s to 0.7283s). Archive NDJSON operation performance improved by 36.84% (p50 from 1.1530s to 0.7283s). Source: llm_adapter@2026-05-21 Confidence: high	—
Performance	Medium	Fanout GC dry run performance improved by 30.62% (p50 from 0.4539s to 0.3149s). Fanout GC dry run performance improved by 30.62% (p50 from 0.4539s to 0.3149s). Source: llm_adapter@2026-05-21 Confidence: high	—
Refactor	Medium	Filesystem traversal migrated from walkdir to jwalk. Filesystem traversal migrated from walkdir to jwalk. Source: llm_adapter@2026-05-21 Confidence: high	—

Full changelog

ContextLattice v3.3.22 — storage-ops perf tranche (async/retry + parallel IO)

This release lands the first focused Rust performance tranche on context_storage_ops (following v3.3.21 runtime cutover).

Included changes

Async HTTP runtime client for storage ops using tokio + reqwest-middleware + reqwest-retry
Filesystem traversal migrated from walkdir to jwalk
Snapshot processing parallelized with rayon (bounded worker pool)
JSON parse fast-path via simd-json with serde fallback
Content hashing migrated from SHA-256 to blake3
Added a repo-agnostic benchmark harness: bench/context_storage_ops_perf_tranche.py

Validation

cargo test --manifest-path crates/Cargo.toml -p context_storage_ops
cargo build --manifest-path crates/Cargo.toml -p context_storage_ops --release
Operational command suite via benchmark harness over synthetic storage dataset

Perf matrix (p50 seconds; baseline vs post-warm)

| operation | baseline p50 | post-warm p50 | delta |
|---|---:|---:|---:|
| ledger | 2.1514s | 0.2781s | -87.08% |
| weekly_lineage_dry | 1.1010s | 0.4492s | -59.20% |
| cold_pack_dry | 1.1135s | 0.3829s | -65.62% |
| cold_tier_dry | 0.5472s | 0.2893s | -47.13% |
| archive_ndjson | 1.1530s | 0.7283s | -36.84% |
| fanout_gc_dry | 0.4539s | 0.3149s | -30.62% |

Notes

Cold-start run also showed major p50 improvements, with one transient first-call outlier removed in warm sampling.
This tranche is additive to the Rust storage-ops cutover and keeps behavior compatible with existing scripts.

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 →