@theokit/sdk-memory 0.5.2 → 0.5.3

package/CHANGELOG.md

@@ -1,5 +1,57 @@
 # Changelog — @theokit/sdk-memory
 
+## 0.5.3
+
+### Patch Changes
+
+- [#570](https://github.com/usetheokit/theokit-sdk/pull/570) [`64ff1f8`](https://github.com/usetheokit/theokit-sdk/commit/64ff1f8e072725dbc4d4dbbb5f82e6f1bca6141e) Thanks [@usetheodev](https://github.com/usetheodev)! - The memory index moves with a relocated store — the note this package's consumers never got ([#562](https://github.com/usetheokit/theokit-sdk/issues/562))
+  
+  This is documentation, not behaviour. The change shipped in `@theokit/sdk@5.0.0` and is already
+  live for anyone on that version; what did not ship is any way for a `@theokit/sdk-memory` consumer
+  to find out. Its `0.5.2` section is empty — no dependency lines, no prose — so the release that
+  carried a privacy-relevant change reached its most directly affected audience with zero signal.
+  
+  **Why this package's readers are the audience.** `@theokit/sdk-memory` owns the markdown store and
+  re-exports `memoryIndexRoot` (`src/internal/store/markdown-store.ts`). Someone reading this
+  CHANGELOG to decide whether a version is safe to take is exactly the person the note was written
+  for.
+  
+  ## What changed in the behaviour
+  
+  `memory.directory` now moves the search index with the store, except into the Claude Code CLI's own
+  directory ([#554](https://github.com/usetheokit/theokit-sdk/issues/554)).
+  
+  The index is not a pointer. `chunks.text` holds the fact TEXT — FTS5/BM25 needs it to search — and
+  `files.path` holds the store's absolute path. **A copy of the index is a copy of the memory,
+  readable with `strings`.**
+  
+  Leaving it at `<cwd>/.theokit/memory/.index/` while the facts moved meant an operator who pointed
+  `directory` at one personal store had that store's contents written into *every repository the
+  agent ran in*, untracked and un-ignored. Recall worked throughout, so nothing looked wrong; what
+  was wrong was where the data landed.
+  
+  ```
+  before   /tmp/my-store/fact.md              facts move
+           /any-project/.theokit/memory/.index/memory.sqlite    ← and the content follows here
+  
+  after    /tmp/my-store/fact.md
+           /tmp/my-store/.index/memory.sqlite                   ← one store, one index
+  ```
+  
+  The Claude Code CLI's directory is unchanged, on its own argument: that CLI has no index format, so
+  a binary there is an artefact the partner does not understand inside a directory it owns
+  (`docs/memory-decisions.md` § 1). The fix narrows the behaviour to match that argument rather than
+  reversing the decision.
+  
+  ## What this does not do
+  
+  **The `0.5.2` section is not edited.** A published entry is a record of what shipped, and rewriting
+  it would make the record disagree with the tarball on npm. The note arrives in the next version
+  instead, which is later than it should have been and is the honest way to be late.
+  
+  If you are on `0.5.2` the behaviour above is already what you have, through the `@theokit/sdk` you
+  resolve — this entry only tells you so.
+
 ## 0.5.2
 
 ## 0.5.2-next.1

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@theokit/sdk-memory",
-  "version": "0.5.2",
+  "version": "0.5.3",
   "description": "Memory subsystem for @theokit/sdk — MemoryProvider impls (markdown file store today; LanceDB + embeddings + circuit breaker next iters). Consumes the kernel-facing port from sdk-core.",
   "license": "Apache-2.0",
   "homepage": "https://github.com/usetheokit/theokit-sdk#readme",
@@ -61,7 +61,7 @@
     "@types/better-sqlite3": "^7.6.13",
     "better-sqlite3": "^12.10.0",
     "sqlite-vec": "^0.1.6",
-    "@theokit/sdk": "5.0.0"
+    "@theokit/sdk": "5.0.1"
   },
   "publishConfig": {
     "provenance": true,