@ruso-0/nreki 11.4.0 → 11.4.1

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 -3
  2. package/package.json +1 -1
package/CHANGELOG.md CHANGED
@@ -2,18 +2,42 @@
2
2
 
3
3
  All notable changes to NREKI will be documented in this file.
4
4
 
5
+ ## [11.4.1] - 2026-05-19
6
+
7
+ ### Added
8
+
9
+ - **Hologram-mode regression tests** (Phase 1.5 follow-up to v11.4.0). Four new tests in [`tests/kernel-pre-existing-errors-hologram.test.ts`](tests/kernel-pre-existing-errors-hologram.test.ts) that literally reproduce **TS1259** (`esModuleInterop:true` + `allowSyntheticDefaultImports:false` + `export =` source) and **TS2802** (`target:ES5` + `downlevelIteration:false` + `Set` iteration) — the exact codes the user originally reported — in hologram mode, the production MCP runtime path. The v11.4.0 suite had pinned only project mode and only triggered TS1192 (not TS1259) despite the CHANGELOG claim of "exact scenario" coverage.
10
+ - [`docs/hologram-mode-verify.md`](docs/hologram-mode-verify.md) — empirical PASS report from the hologram-mode round: differential filter works in hologram, including the JIT-specific subcase where the target file enters `rootNames` only at edit time.
11
+ - [`docs/v11.4.0-self-exploration.md`](docs/v11.4.0-self-exploration.md) — 10-finding pre-Furia self-audit that drove the Phase 1.5 round (process documentation; honest list of what I was uncomfortable with after v11.4.0).
12
+
13
+ ### Verified empirically (no code fix required)
14
+
15
+ - The differential pre-existing-error check (`count > baseline.get(fingerprint)` at [`src/kernel/backends/ts-compiler-wrapper.ts:611`](src/kernel/backends/ts-compiler-wrapper.ts#L611)) holds in hologram mode for the user's exact TS1259 and TS2802 cases. The user's original report most likely ran against the globally installed `@ruso-0/nreki@10.19.0` (the binary that was active when the report was filed); the fix had already shipped in an earlier release.
16
+
17
+ ### Internal
18
+
19
+ - v11.4.0's CHANGELOG entry overstated test coverage. v11.4.1 is the honest closure: same release intent, but with the hologram suite that actually pins the user-reported codes, and a self-exploration document that names what the original sprint missed.
20
+ - All v11.4.0 deliverables remain shipped on the `v11.4.0` tag at commit `984c781`. v11.4.1 strictly adds tests + docs on top; no runtime behavior change.
21
+
22
+ ### Tests
23
+
24
+ - 1 427 passed, 3 skipped, 0 failed (+4 vs v11.4.0).
25
+
5
26
  ## [11.4.0] - 2026-05-19
6
27
 
7
28
  ### Fixed
8
29
 
9
30
  - **EISDIR crash on `nreki_code action:"read"` / `action:"compress"` with a directory path.** Both handlers previously called `fs.readFileSync` without a stat guard; passing a directory surfaced an opaque `EISDIR: illegal operation on a directory` error from the MCP server. v11.4.0 explicitly checks `stat.isDirectory()` and returns a structured error that points to the correct tool: `nreki_navigate action:"outline"` for per-file overview, `action:"fast_grep" / "search"` for cross-tree lookup. `ENOENT` (missing file) is now also caught with a structured "Path not found" message instead of an uncaught stack trace.
10
- - **Regression guard for pre-existing TS errors in edited files.** A user reported (2026-05-19) that NREKI rolled back unrelated edits in `fillTemplate.ts` because of pre-existing TS1259/TS2802 errors elsewhere in the same file. The differential check (`count > baseline.get(fingerprint)` in `src/kernel/backends/ts-compiler-wrapper.ts:611`) was already in place and works correctly in v11.3.x — the user's binary was likely an older release (the globally installed `@ruso-0/nreki@10.19.0`). v11.4.0 closes the gap by adding four explicit regression tests in [`tests/kernel-pre-existing-errors.test.ts`](tests/kernel-pre-existing-errors.test.ts) that pin the exact same-file scenario the user described (existing kernel tests only covered separate-file pre-existing errors).
31
+ - **Regression guards for pre-existing TS errors in edited files.** A user reported (2026-05-19) that NREKI rolled back unrelated edits in `fillTemplate.ts` because of pre-existing TS1259/TS2802 errors elsewhere in the same file. The differential check (`count > baseline.get(fingerprint)` in `src/kernel/backends/ts-compiler-wrapper.ts:611`) was already in place and works correctly in v11.3.x — the user's binary was likely an older release (the globally installed `@ruso-0/nreki@10.19.0`). v11.4.0 closes the test-coverage gap with two suites: one in project mode covering same-file pre-existing TS2322 / TS1192-family, and one in hologram mode (the MCP server's default) covering the user's *exact* TS1259 and TS2802 codes plus a JIT-specific subcase where the target file enters `rootNames` only at edit time.
11
32
 
12
33
  ### Added
13
34
 
14
35
  - **`tests/read-compress-directory.test.ts`** — 4 tests covering directory + missing-file inputs to both `handleRead` and `handleCompress`. Confirms EISDIR is gone, error messages name the correct alternative tool, and the response shape is `isError: true` (not a crash).
15
- - **`tests/kernel-pre-existing-errors.test.ts`** — 4 tests pinning the same-file pre-existing-error scenario (TS2322, TS1192-family, same-symbol edit, counter-test of genuine new error).
16
- - **`docs/furia-v11.4.0-review.md`** — adversarial Q1-Q5 review documenting (Q1) why Phase 1 contributes only regression guards, (Q2) why the EISDIR fix returns an explicit error instead of auto-redirecting, (Q3) zero perf overhead, (Q4) anti-silent-failure audit, (Q5) two residual gaps logged for a future iteration (FIFO/device-file handling, push the guard into `readSource` itself).
36
+ - **`tests/kernel-pre-existing-errors.test.ts`** — 4 project-mode tests pinning the same-file pre-existing-error scenario (TS2322, TS1192-family, same-symbol edit, counter-test of genuine new error).
37
+ - **`tests/kernel-pre-existing-errors-hologram.test.ts`** — 4 hologram-mode tests covering the **exact** TS codes the user reported (TS1259 reproduced via `esModuleInterop:true` + `allowSyntheticDefaultImports:false` + `export =` source; TS2802 reproduced via `target:ES5` + `downlevelIteration:false` + Set iteration). Includes a JIT-specific case where the target file is not yet in `rootNames` at boot and only enters via the per-transaction `captureBaseline` ordering.
38
+ - **`docs/furia-v11.4.0-review.md`** — adversarial Q1-Q5 review documenting why the Phase 1 contribution is regression guards (not a code fix), why the EISDIR fix returns an explicit error instead of auto-redirecting, perf claims, anti-silent-failure audit, and two residual gaps (FIFO/device-file handling, un-guarded `readSource` helper).
39
+ - **`docs/v11.4.0-self-exploration.md`** — honest pre-Furia self-audit listing 10 findings I was uncomfortable with after the initial implementation. Drove the additional hologram-mode verification.
40
+ - **`docs/hologram-mode-verify.md`** — empirical verification result for the hologram-mode question raised by Finding D of the self-audit. All 4 hologram tests pass; no kernel scoping bug; no code change required.
17
41
 
18
42
  ### Templates
19
43
 
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@ruso-0/nreki",
3
- "version": "11.4.0",
3
+ "version": "11.4.1",
4
4
  "description": "MCP plugin that validates AI agent edits in RAM before they touch disk. Spectral clustering, architecture diffs, bridge detection, dead code oracle, and cross-file semantic checks for TypeScript (full Type Ledger), with parser-level support for JavaScript, Python (pyright), Go (gopls), Kotlin, Java, C, and C++. Zero cloud dependencies.",
5
5
  "main": "dist/index.js",
6
6
  "type": "module",