This release fixes issues for SREs watching stability and regressions.
✓ No known CVEs patched in this version
ReleasePort's take
Light signalv3.2.1 fixes suggest_metadata to require concrete chunk evidence before proposing tag candidates, preventing loose text matches from returning tags without supporting chunks.
Why it matters: Severity 50 fix. Previously loose text matches produced tag candidates without evidence chunks. Update if suggest_metadata tag suggestions influence your workflows.
Summary
AI summaryFixed require concrete chunk evidence before suggesting tag candidates in suggest_metadata.
Changes in this release
| Type | Severity | Summary | CVE |
|---|---|---|---|
| Bugfix | Medium |
Require concrete chunk evidence before returning vocabulary-based tag_candidates from suggest_metadata, preventing loose text matches from proposing tags without a supporting chunk/snippet. Require concrete chunk evidence before returning vocabulary-based tag_candidates from suggest_metadata, preventing loose text matches from proposing tags without a supporting chunk/snippet. Source: llm_adapter@2026-05-21 Confidence: high |
— |
Full changelog
Patch release after live tag/chunk audit.\n\nFixed:\n- Require concrete chunk evidence before returning vocabulary-based tag_candidates from suggest_metadata, preventing loose text matches from proposing tags without a supporting chunk/snippet.\n\nVerified:\n- Live Obsidian DB rehearsal and live indexing with index_all(with_embeddings=true).\n- python -m pytest -q\n- python test_server.py
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 KVANTRA-dev/NOUZ-MCP
Semantic knowledge graph for Obsidian. Three modes (pure graph / semantic classification / strict hierarchy)
Related context
Beta — feedback welcome: [email protected]