BetterDB-inc/monitor

vsemantic-cache-v0.6.0 scope: semantic-cache Feature

This release adds 4 notable features for engineering teams evaluating rollout.

Published 1mo Monitoring & Metrics

View tool

✓ No known CVEs patched

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

✓ No known CVEs patched in this version

Topics

prometheus monitoring obsevability redis valkey

Summary

AI summary

Built-in keyword-overlap rerank factory adds configurable cosine‑similarity blending and prompt exposure.

Changes in this release

Type	Severity	Summary	CVE
Feature
Feature	Medium	Adds built-in keyword-overlap rerank factory with configurable cosine weight Adds built-in keyword-overlap rerank factory with configurable cosine weight Source: llm_adapter@2026-06-11 Confidence: high	—
Feature	Medium	Exposes stored prompt on rerank candidates via `prompt` key Exposes stored prompt on rerank candidates via `prompt` key Source: llm_adapter@2026-06-11 Confidence: high	—
Feature	Medium	Adds `cachedPrompt` field to judge context (inert by default) Adds `cachedPrompt` field to judge context (inert by default) Source: llm_adapter@2026-06-11 Confidence: high	—
Feature	Medium	Implements cost instrumentation recording `cost_saved_micros` for similarity-window hits and retroactively storing miss costs Implements cost instrumentation recording `cost_saved_micros` for similarity-window hits and retroactively storing miss costs Source: llm_adapter@2026-06-11 Confidence: high	—
Bugfix	Medium	Fixes unbounded growth of miss‑pending zset by pruning entries older than 5 minutes on each record Fixes unbounded growth of miss‑pending zset by pruning entries older than 5 minutes on each record Source: llm_adapter@2026-06-11 Confidence: high	—

Full changelog

Added

Built-in keyword-overlap rerank factory — createKeywordOverlapRerank() blends cosine similarity with word overlap. Supports compare: 'prompt' (equivalence, default) and compare: 'response' (relevance), with configurable cosineWeight.
Stored prompt exposed on rerank candidates — candidates now include a prompt key (additive).
cachedPrompt in judge context — judge context carries the stored prompt text (reserved, inert by default).
Cost instrumentation — similarity-window entries record cost_saved_micros for hits; a subsequent store() retroactively populates the cost of the preceding miss. Powers BetterDB Monitor's cost-weighted threshold recommendations.

Fixed

Miss-pending zset is pruned and flushed — entries older than the 5-minute bound are pruned on every record, so miss-only traffic cannot grow the bookkeeping zset unbounded; flush() now deletes the miss-pending key.

Full Changelog: https://github.com/BetterDB-inc/monitor/compare/semantic-cache-v0.5.1...semantic-cache-v0.6.0

Full Changelog: https://github.com/BetterDB-inc/monitor/compare/v0.23.0...semantic-cache-v0.6.0

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 BetterDB-inc/monitor

Get notified when new releases ship.

About BetterDB-inc/monitor

Valkey-first observability with Redis compatibility. Query real-time metrics, analyze slow commands, detect hot keys, and investigate performance issues directly from AI coding assistants.

All releases →

Related context

Related tools

Earlier breaking changes

vagent-cache-py-v0.7.0 Cache keys changed for tool-using requests on the LlamaIndex adapter; existing cached entries will miss once after upgrade.
vagent-cache-v0.7.0 Cache keys for tool-using requests changed on Vercel and LlamaIndex adapters; existing entries become one-time misses.
vagent-v1.5.0 TLS is now mandatory; agent exits with error if `--valkey-tls true` is not set for IAM auth mode.