This release fixes issues for SREs watching stability and regressions.
✓ No known CVEs patched in this version
Summary
AI summaryFixed compilation lifetime issues and eliminated KV cache contamination across embedding passes for Qwen3Model.
Full changelog
🐛 Bug Fixes
- Embedding Engine (Definitive KV Cache Fix): The previous workaround for the
Qwen3ModelKV cache accumulation (usingVarBuilder::clone()) caused compilation lifetimes issues. This release introduces the ultimate and correct fix. We added a local patch forcandle-transformersv0.9.2 via[patch.crates-io]to exposeclear_kv_cache()as apub fndirectly on the baseQwen3Model. - The engine now elegantly and explicitly calls
model_mut.clear_kv_cache()before each independent forward pass (both in single and batched embedding generation), completely eliminating theshape mismatch in broadcast_addpanic and preventing cross-contamination between texts.
Full Changelog: https://github.com/pomazanbohdan/memory-mcp-1file/compare/v0.4.4...v0.4.5
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
About pomazanbohdan/memory-mcp-1file
A self-contained Memory server with single-binary architecture (embedded DB & models, no dependencies). Provides persistent semantic and graph-based memory for AI agents.
Related context
Beta — feedback welcome: [email protected]