memento-mcp 0.1.8 → 0.2.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (2) hide show
  1. package/CHANGELOG.md +27 -0
  2. package/package.json +1 -1
package/CHANGELOG.md CHANGED
@@ -8,6 +8,33 @@ Format follows [Keep a Changelog](https://keepachangelog.com/en/1.1.0/). Version
8
8
 
9
9
  ## [Unreleased]
10
10
 
11
+ ---
12
+
13
+ ## [0.2.0] - 2026-02-20
14
+
15
+ ### Added
16
+ - `recall_threshold` workspace setting — filter out low-confidence memories before returning results; default 0 (disabled, backward compatible). Set via `PUT /v1/settings/recall_threshold`.
17
+ - `/v1/settings` endpoint — `GET` lists all workspace settings, `PUT /:key` sets a value, `DELETE /:key` removes it (reverts to default). Exposes `recall_threshold` and `recall_alpha`.
18
+ - Structured entity tagging in `/v1/distill` — LLM now emits typed tags (`person:elena-vasquez`, `grant:2401-8827`, `specialty:photonics`, `date:2025-03-03`, `laser:532nm`, etc.) enabling tag-substring search for specific entities.
19
+ - Entity tag budget raised from 3 to 7 (+ `source:distill` = max 8 stored tags).
20
+
21
+ ### Fixed
22
+ - Stop word filtering in `scoreAndRankMemories` — common words ("is", "the", "on", "who") no longer inflate keyword scores for all memories. Short numeric identifiers ("62", "532") are preserved. Falls back to unfiltered terms for vacuous queries.
23
+ - Distill system prompt vocabulary — extracted memories now lead with searchable terms, preserve exact identifiers verbatim (`#2401-8827`, `$240,000`, `532nm`), include role vocabulary explicitly, and name project context.
24
+
25
+ ### Changed
26
+ - `/v1/memories/recall` applies `recall_threshold` filter before building response (default 0 = no filter).
27
+ - `/v1/context` applies `recall_threshold` filter to keyword results before hybrid rank fusion.
28
+
29
+ ---
30
+
31
+ ## [0.1.9] - 2026-02-20
32
+
33
+ ### Fixed
34
+ - Auto-consolidation black hole: `POST /v1/consolidate` now writes the consolidated summary back to the `memories` table (visible to recall), consistent with `POST /v1/consolidate/group`. Previously, consolidated content was written only to the `consolidations` table and was invisible to all recall endpoints.
35
+ - Misleading response message: now reads `"Consolidated N group(s) from M source memories into N new memory/memories."` instead of the ambiguous `"Consolidated N groups (M memories total)"` which counted inputs, not outputs.
36
+ - Source memories' `consolidated_into` field now correctly points to the new memory ID (in `memories` table) rather than the consolidation record ID.
37
+
11
38
  ### Added
12
39
  - `memento_item_create`, `memento_item_list`, `memento_item_update`, `memento_item_delete` — structured working memory items with categories, priorities, statuses, and next actions
13
40
  - `memento_consolidate` — merge 2+ overlapping memories into a single richer memory; originals deactivated, not deleted
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "memento-mcp",
3
- "version": "0.1.8",
3
+ "version": "0.2.0",
4
4
  "mcpName": "io.github.myrakrusemark/memento-protocol",
5
5
  "description": "The Memento Protocol — persistent memory for AI agents",
6
6
  "type": "module",