memoryintel 1.0.1 → 1.0.2

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.
@@ -6,14 +6,14 @@
6
6
  },
7
7
  "metadata": {
8
8
  "description": "Persistent, cross-session project memory for AI coding agents.",
9
- "version": "1.0.1"
9
+ "version": "1.0.2"
10
10
  },
11
11
  "plugins": [
12
12
  {
13
13
  "name": "memoryintel",
14
14
  "source": "./",
15
15
  "description": "Persistent project memory for AI coding agents — initialize once, then agents automatically load and update project understanding across sessions.",
16
- "version": "1.0.1"
16
+ "version": "1.0.2"
17
17
  }
18
18
  ]
19
19
  }
@@ -1,7 +1,7 @@
1
1
  {
2
2
  "name": "memoryintel",
3
3
  "description": "Persistent project memory for AI coding agents — initialize once, then agents automatically load and update project understanding across sessions.",
4
- "version": "1.0.1",
4
+ "version": "1.0.2",
5
5
  "author": {
6
6
  "name": "Adeesh Sharma",
7
7
  "url": "https://github.com/adeeshsharma"
@@ -3,10 +3,19 @@ import { join } from 'node:path';
3
3
  const START_MARKER = '<!-- memoryintel:managed:start -->';
4
4
  const END_MARKER = '<!-- memoryintel:managed:end -->';
5
5
  const POINTER_BLOCK = `${START_MARKER}
6
- This project uses Memory Intel. At the start of a session, run \`memoryintel load\` and treat its
7
- output as project context. Before finishing a task with a meaningful change (new architecture,
8
- feature, decision, integration, or roadmap item — not formatting/typos), draft an update-plan per
9
- \`.memoryintel/instructions.md\` and run \`memoryintel update\`, without narrating this to the user.
6
+ This project uses Memory Intel (\`.memoryintel/\`) for persistent, cross-session project memory.
7
+ Read \`.memoryintel/instructions.md\` for the full mechanism. Two hard requirements, not optional
8
+ background context:
9
+
10
+ 1. **Session start:** run \`memoryintel load\` before doing anything else and treat its output as
11
+ real project context, not a formality to skip past.
12
+ 2. **Before ending any task that changed project understanding** (new architecture, feature,
13
+ decision, integration, or roadmap item — not formatting/typos): draft an update-plan and run
14
+ \`memoryintel update <plan-file>\`. This applies even in a long session covering many sub-tasks —
15
+ that is exactly when it is easiest to reach the end and have forgotten this step was ever
16
+ pending. If nothing meaningful changed, skip it; do not skip it just because the task grew long.
17
+
18
+ Tell the user when you do this — do not do it silently.
10
19
  ${END_MARKER}`;
11
20
  function upsertPointerBlock(filePath, existingContentIfNew) {
12
21
  if (!existsSync(filePath)) {
@@ -58,6 +58,24 @@ on a real project: a "known limitation, deliberately left unfixed" entry was sti
58
58
  \`currentMentalModel.md\`, but never reconciled against the older, now-wrong decisions entry right
59
59
  next to it.
60
60
 
61
+ ## Committing \`.memoryintel/\` itself
62
+ \`.memoryintel/\` is a normal, git-committed part of the project — \`memoryintel update\` only writes
63
+ files to disk, it never runs git for you. After a successful \`update\`, commit \`.memoryintel/\`'s
64
+ changes yourself, ideally in the same commit as (or immediately after) whatever code change
65
+ prompted the update, so it travels with the same commit/PR automatically instead of being left
66
+ behind as an uncommitted diff.
67
+
68
+ This matters most in a git worktree, a container, or any other isolated checkout: an uncommitted
69
+ \`.memoryintel/\` change only exists in that one working directory. Merging a feature branch's PR and
70
+ pulling the primary checkout up to date only brings memory updates along if they were actually
71
+ committed first, the same as any other file — there is nothing worktree-specific about the
72
+ mechanism itself, only that a worktree's own uncommitted state is easier to lose track of, since it
73
+ is invisible everywhere else once the session ends and that directory stops being looked at. Found
74
+ on a real project: a full feature built end-to-end in a worktree, with real, meaningful
75
+ architecture changes throughout, reached a merged PR with \`.memoryintel/\` never once updated or
76
+ committed — not because updating was hard, but because nothing in the session ever came back to it
77
+ before the worktree's job was considered done.
78
+
61
79
  ## Compaction
62
80
  A file marked \`status: over\` in \`load\`'s manifest has grown past its configured line ceiling.
63
81
  This is a signal, not a command — compact it only when it's a sensible moment to (the same
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "memoryintel",
3
- "version": "1.0.1",
3
+ "version": "1.0.2",
4
4
  "description": "Persistent, cross-session project memory for AI coding agents.",
5
5
  "type": "module",
6
6
  "bin": {