v0.6.1 Announcement

English

Agent Memory Bridge v0.6.1 is out.

This release adds the first reviewed-sample calibration layer for classifier-assisted enrichment.

What changed:

1. Classifier calibration is now explicit

• reviewed samples
• classifier-vs-keyword comparison
• exact-match and better/worse counts
• a repeatable calibration report

1. Retrieval and coordination proof stay in place

• expected_top1_accuracy
• deterministic proof
• signal lifecycle remains claim -> extend -> ack / expire / reclaim

Current calibration snapshot:

• classifier_exact_match_rate = 0.667
• fallback_exact_match_rate = 0.0
• classifier_better_count = 4
• fallback_better_count = 2
• 76 passed

This is the point of the release:

• do not widen assist mode on vibes
• compare it against reviewed examples first

GitHub:
https://github.com/zzhang82/Agent-Memory-Bridge

Release:
https://github.com/zzhang82/Agent-Memory-Bridge/releases/tag/v0.6.1

简体中文

Agent Memory Bridge v0.6.1 发布了。

这次发布补上了 classifier-assisted enrichment 的第一层 reviewed-sample calibration。

主要变化：

1. classifier calibration 现在是显式可跑的

• 有 reviewed samples
• 有 classifier-vs-keyword 对比
• 有 exact match 和 better/worse 统计
• 有可重复生成的 calibration report

1. retrieval 和 coordination proof 仍然保留

• expected_top1_accuracy
• deterministic proof
• signal lifecycle 继续是 claim -> extend -> ack / expire / reclaim

当前 calibration snapshot：

• classifier_exact_match_rate = 0.667
• fallback_exact_match_rate = 0.0
• classifier_better_count = 4
• fallback_better_count = 2
• 76 passed

这版真正重要的地方是：

• 不再凭感觉放宽 assist mode
• 而是先和 reviewed examples 做对比

GitHub:
https://github.com/zzhang82/Agent-Memory-Bridge

Release:
https://github.com/zzhang82/Agent-Memory-Bridge/releases/tag/v0.6.1