claude-mem-lite 3.38.3 → 3.38.4
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.
|
@@ -10,7 +10,7 @@
|
|
|
10
10
|
"plugins": [
|
|
11
11
|
{
|
|
12
12
|
"name": "claude-mem-lite",
|
|
13
|
-
"version": "3.38.
|
|
13
|
+
"version": "3.38.4",
|
|
14
14
|
"source": "./",
|
|
15
15
|
"description": "Persistent long-term memory for Claude Code via MCP — captures coding decisions, bugfixes, and context across sessions. Hybrid FTS5 + TF-IDF search with episode batching. Single SQLite DB, no external services. A lighter, lower-cost alternative to claude-mem (episode batching + a smaller model; cost savings are an internal estimate, not a measured benchmark)."
|
|
16
16
|
}
|
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "claude-mem-lite",
|
|
3
|
-
"version": "3.38.
|
|
3
|
+
"version": "3.38.4",
|
|
4
4
|
"description": "Persistent long-term memory for Claude Code via MCP — captures coding decisions, bugfixes, and context across sessions. Hybrid FTS5 + TF-IDF search with episode batching. Single SQLite DB, no external services. A lighter, lower-cost alternative to claude-mem (episode batching + a smaller model; cost savings are an internal estimate, not a measured benchmark).",
|
|
5
5
|
"author": {
|
|
6
6
|
"name": "sdsrss"
|
package/adopt-content.mjs
CHANGED
|
@@ -4,10 +4,13 @@
|
|
|
4
4
|
// claudemd.mjs primitives so the strings are testable without side effects.
|
|
5
5
|
//
|
|
6
6
|
// CURRENT_SENTINEL_VERSION tags the managed block as `<!-- claude-mem-lite:begin
|
|
7
|
-
// vN -->`. needsRefresh()
|
|
8
|
-
//
|
|
9
|
-
//
|
|
10
|
-
//
|
|
7
|
+
// vN -->`. needsRefresh() (claudemd.mjs) compares the version tag AND the block
|
|
8
|
+
// body AND the detail-doc content — ANY of the three differing triggers an
|
|
9
|
+
// in-place refresh (drift = intended content change, overwritten rather than
|
|
10
|
+
// treated as a user edit). So a template edit propagates on the next SessionStart
|
|
11
|
+
// even WITHOUT a version bump; the tag need not be monotonic and, in practice,
|
|
12
|
+
// need not change at all — we keep `v1` across content edits (see mem #8846).
|
|
13
|
+
// We deliberately use `v1` (not `v2`) so the version digit differs from
|
|
11
14
|
// the sibling code-graph-mcp plugin's `<!-- code-graph-mcp:begin v2 -->` block in
|
|
12
15
|
// the same CLAUDE.md; the slug already scopes the two independently (claudemd.mjs),
|
|
13
16
|
// so this is a cosmetic distinguisher, not a functional one. The pre-v3.13 legacy
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "claude-mem-lite",
|
|
3
|
-
"version": "3.38.
|
|
3
|
+
"version": "3.38.4",
|
|
4
4
|
"description": "Persistent long-term memory for Claude Code via MCP — captures coding decisions, bugfixes, and context across sessions. Hybrid FTS5 + TF-IDF search with episode batching. Single SQLite DB, no external services. A lighter, lower-cost alternative to claude-mem (episode batching + a smaller model; cost savings are an internal estimate, not a measured benchmark).",
|
|
5
5
|
"type": "module",
|
|
6
6
|
"packageManager": "npm@10.9.2",
|
package/search-scoring.mjs
CHANGED
|
@@ -32,7 +32,7 @@ const INSTRUCTIONS_BASE = [
|
|
|
32
32
|
` ${CLI_INVOKE} get 42,43 — full details by ID`,
|
|
33
33
|
` ${CLI_INVOKE} timeline --anchor 42 — chronological context`,
|
|
34
34
|
'',
|
|
35
|
-
'MCP tools: mem_search, mem_recent, mem_save, mem_get, mem_recall, mem_timeline. If already loaded, call directly (warm server, fastest path). In tool-heavy sessions these are deferred behind ToolSearch — if using one would cost a ToolSearch load first, run the Bash CLI above instead: one call, not two. Neither needs a PATH/CLI install.',
|
|
35
|
+
'MCP tools: mem_search, mem_recent, mem_save, mem_get, mem_recall, mem_timeline (plus mem_defer/mem_defer_list/mem_defer_drop for cross-session TODOs). If already loaded, call directly (warm server, fastest path). In tool-heavy sessions these are deferred behind ToolSearch — if using one would cost a ToolSearch load first, run the Bash CLI above instead: one call, not two. Neither needs a PATH/CLI install.',
|
|
36
36
|
'mem_save: Save non-obvious insights (bugfix lessons, architecture decisions).',
|
|
37
37
|
'Search tips: short keywords (2-3 words), filter with obs_type when relevant.',
|
|
38
38
|
];
|
|
@@ -46,6 +46,7 @@ const INSTRUCTIONS_VERBOSE = [
|
|
|
46
46
|
' • Starting new feature → mem_search(query="feature area") for prior art & patterns',
|
|
47
47
|
' • After fixing a tricky bug → mem_save(type="bugfix", lesson_learned="root cause & fix")',
|
|
48
48
|
' • After architecture decision → mem_save(type="decision", lesson_learned="rationale")',
|
|
49
|
+
' • Deferring work to a future session → mem_defer(title, priority, detail); when fixed, mem_save(..., closes_deferred=[N])',
|
|
49
50
|
' • Hook-injected context mentions #ID → mem_get(ids=[ID]) for full details',
|
|
50
51
|
'',
|
|
51
52
|
'Decision rules (use INSTEAD OF multi-step search):',
|