nexusmem 0.8.0 → 0.9.1

package/CHANGELOG.md

@@ -11,6 +11,67 @@ built from, matched by publish timestamp: `v0.1.0` → `67a4776`, `v0.1.1` → `
 
 No unreleased changes yet.
 
+## [0.9.1] — 2026-08-29
+
+Three ranking/retrieval correctness fixes, found and validated against a new 28-case retrieval-quality
+eval harness (`npm run eval`, dev-only, not part of the published package).
+
+### Fixed
+
+- `nodes_vec` (vector search) over-fetched `k` globally, then filtered by `project_id` after the join —
+  a heuristic, not a guarantee. A sparse project sharing `memory.db` with a much larger one could have
+  every true nearest neighbour fall outside the over-fetch window and get silently dropped (reproduced:
+  495 rows in one project + 5 in another, global `k=50` surfaced 0 of the 5). Schema V11 gives `nodes_vec`
+  a `PARTITION KEY` on `project_id` (`sqlite-vec` 0.1.9+), pushing the equality filter into the k-NN
+  search itself so cross-project exactness is now guaranteed, not probable. Migrates existing databases
+  automatically on next `sync`/`query`. `--as-of` date filtering still over-fetches — only the
+  `project_id` dimension was ever a correctness guarantee.
+- `MAX_PRIOR_OVERTURN` (the cap on how far a recency/signal prior may overturn relevance) raised
+  `2 → 2.4`, the highest value that both improves eval MRR (0.924→0.943) and still passes every legacy
+  regression test guarding against the original same-day-fix-commits bug this constant exists to bound.
+- A commit's `code_diff` siblings (one node per changed file, all sharing the same `ts`) could crowd a
+  packed result and bury that commit's own `git_commit` node — its answer — several ranks down. Now
+  capped per commit, matching the existing `conversation_turn`/`doc_section` family cap.
+- `package.json`/`tsconfig.json` diffs were ranking above more relevant results when a changed file's
+  path happened to echo its commit's conventional-commit scope (title is weighted 10x body, so the scope
+  word counted twice). Down-weighted as mechanical wiring, same reasoning `TEST_PATHS` already applies to
+  tests. Only affects nodes ingested from here on — existing manifest/config diffs need `sync --rebuild`
+  to get the corrected signal retroactively.
+
+## [0.9.0] — 2026-08-27
+
+Closes the three remaining mechanisms from a recurring external review (Simon Strandgaard, Agent
+Memory Atlas): trust_state, bi-temporal reads, and a dismiss verb for standing suggestions. Also
+fixes two real Windows-specific correctness bugs found dogfooding.
+
+### Added
+
+- `nodes.trust_state` (`candidate` | `verified` | `rejected`, schema V10) and `nexusmem review
+  <nodeId> --verify|--reject`: a human verdict on a node, independent of the SLM contradiction
+  checker. `--reject` down-weights ranking (harsher than the existing supersede penalty, since a
+  human said no directly); `--verify` is a label only, no ranking boost. Both render as a tag in
+  packed context. Never deletes — same demote-not-delete rule as `mark-stale`.
+- `--as-of <date>` on `nexusmem query` and the MCP `search_memory` tool: restricts both the BM25
+  and vector arms to nodes recorded at or before that instant via `created_at` (record time),
+  independent of how old the events themselves (`ts`, event time) are. Read-only — there is no
+  equivalent write, and nodes stay write-once.
+- `nexusmem stale --dismiss`: silences a contradiction suggestion the user disagrees with, without
+  fabricating a `supersedes` relationship. Previously a wrong `--check-contradictions` verdict had
+  no way to be rejected — it re-decorated every future `stale`/`sync` run forever.
+- `--prune-source`/`--prune-stale-shell` now record a `mutation_audit` row on their `--yes` path,
+  matching what `forget` already did. The dry-run preview stays a pure read.
+
+### Fixed
+
+- `git rev-parse` failures reported as "error launching git: Access is denied." (Git for Windows'
+  launcher shim failing to exec `git.exe` under handle/AV contention) were treated as git's own
+  verdict instead of a transient spawn failure, turning a one-off environment hiccup into a hard
+  failure. Now retried the same as the other known transient-spawn classes.
+- The PowerShell prompt hook read only `$LASTEXITCODE`, which cmdlets (`Remove-Item`, `Copy-Item`,
+  …) never set — every failing cmdlet was silently logged as a success, and a stale
+  `$LASTEXITCODE` from an earlier native command could misattribute to later successful cmdlet
+  runs. Now reads `$?` alongside `$LASTEXITCODE`, before anything else can overwrite `$?`.
+
 ## [0.8.0] — 2026-08-22
 
 ### Added
@@ -504,7 +565,8 @@ First public release.
   there is no local-model summarization pass, and the conversation collector has never been audited
   for the stale-node bug that was found and fixed in the docs collector.
 
-[Unreleased]: https://github.com/yaminbkk/NexusMem/compare/v0.8.0...HEAD
+[Unreleased]: https://github.com/yaminbkk/NexusMem/compare/v0.9.0...HEAD
+[0.9.0]: https://github.com/yaminbkk/NexusMem/compare/v0.8.0...v0.9.0
 [0.8.0]: https://github.com/yaminbkk/NexusMem/compare/v0.7.0...v0.8.0
 [0.7.0]: https://github.com/yaminbkk/NexusMem/compare/v0.6.0...v0.7.0
 [0.6.0]: https://github.com/yaminbkk/NexusMem/compare/v0.5.4...v0.6.0

package/README.md

@@ -248,7 +248,9 @@ One thing that is not a percentage: shell commands and conversation turns have n
 equivalent. Without something recording them, they are gone, not merely more expensive to find.
 
 For how these numbers compare to a similar tool's own claims, see
-[`docs/competitor-comparison.md`](docs/competitor-comparison.md).
+[`docs/competitor-comparison.md`](docs/competitor-comparison.md) (vs. projectmem) and
+[`docs/competitor-comparison-yesmem.md`](docs/competitor-comparison-yesmem.md) (vs. YesMem, including
+native Windows support vs. its documented WSL2 requirement).
 
 Latency on a ~530-node corpus, warm, p50 over 10 runs:
 
@@ -316,6 +318,21 @@ candidate against nodes *found by embedding similarity*; a contradiction from an
 node would never surface. Comprehensive contradiction detection (not just for the pair the vector
 search happens to surface) is still an open problem, and nothing here supersedes a node on its own.
 
+`provenance` is a separate question from `trust_state`: provenance says where a claim came from,
+never whether anyone checked it.
+
+```bash
+nexusmem review <nodeId> --verify
+nexusmem review <nodeId> --reject
+```
+
+Records your own verdict on one node, independent of the SLM contradiction checker above (which only
+ever writes a suggestion, never a verdict). `--reject` down-weights the node in ranking — same
+demote-not-delete rule as `mark-stale`, it stays queryable, just usually loses to better matches —
+and both verdicts are shown as a `[verified]`/`[rejected]` tag on every query result that returns the
+node afterward. `--verify` is a label only; it does not boost ranking. Every node starts `candidate`
+(untagged) until reviewed, and a re-sync never overwrites a verdict once one is set.
+
 ## Where it breaks
 
 - **Shell history without the hook is unscoped.** Scraped history has no directory context, so it is
@@ -381,16 +398,20 @@ search happens to surface) is still an open problem, and nothing here supersedes
 
 ## Commands
 
-`init`, `sync`, `query <text>`, `status` (add `--share` for a plain-text summary worth pasting
-somewhere), `projects`, `mcp`, `forget <value>`, `stale` (add `--check-contradictions` for a local-SLM
-content check, see above), `mark-stale <nodeId> --supersedes <newNodeId>`, and
-`hook install|remove|status`.
-
-There are also five dry-run previews (`scan-git`, `scan-diff`, `scan-shell`, `scan-docs`,
-`scan-conversation`)
-that write nothing and print the nodes ingestion *would* create along with their signal scores. That
-is the intended way to tune scoring against a real repository before committing to a change. Add
-`--json` to pipe them somewhere.
+`init`, `sync`, `query <text>` (add `--as-of <date>` for a bi-temporal read, see below), `status` (add
+`--share` for a plain-text summary worth pasting somewhere), `projects`, `mcp`, `forget <value>`,
+`stale` (add `--check-contradictions` for a local-SLM content check, see above), `mark-stale
+<nodeId> --supersedes <newNodeId>`, `review <nodeId> --verify|--reject` (record a human verdict on
+one node, see above), `precheck` (advisory — warns about staged files with an unresolved past
+failure or high recent churn; exits 0 unless `--strict`), `hook install|remove|status` (the
+PowerShell exit-code hook), and `hook git install|remove|status` (a git pre-commit hook that runs
+`precheck` before each commit).
+
+There are also seven dry-run previews (`scan-git`, `scan-diff`, `scan-shell`, `scan-docs`,
+`scan-conversation`, `scan-session`, `scan-structure`) that write nothing and print what ingestion
+*would* produce — nodes and their signal scores for the first six, import-graph edges for
+`scan-structure`. That is the intended way to tune scoring against a real repository before
+committing to a change. Add `--json` to pipe them somewhere.
 
 Every command takes `-C <path>` to target another repository. On `sync`, `--conversation` opts the
 transcript source in for one run without persisting it, `--no-embed` skips the vector pass,
@@ -423,6 +444,24 @@ scope   2 project(s): NexusMem, uploader
 - 2026-08-09 [observed] [NexusMem] fix(git): retry a transient failure to spawn git
 ```
 
+## Bi-temporal reads
+
+Every node carries two clocks: `ts`, the event's own time ("what happened then"), and `created_at`,
+the moment the store actually recorded it ("what did the store hold then") — normally the same
+question, but not when a sync runs late, a backfill lands weeks after the events it describes, or a
+teammate's clone catches up all at once. `query`/`search_memory` answer the first by default;
+`--as-of <date>` switches to the second:
+
+```bash
+nexusmem query "why does the ranker cap joint priors" --as-of 2026-08-10
+```
+
+Only nodes recorded at or before that instant are considered, even if the events they describe are
+older still. There is no equivalent write — this is a read-time filter over `created_at`, not a
+snapshot or a way to query a value that has since changed, since nodes are write-once (see
+[Staleness & provenance](#staleness--provenance) for what changes a node's *weight*, not its
+record).
+
 Databases stay per-repository — there is no shared global store, and deleting one repo's
 `.nexusmem/` still removes exactly that repo's memory. What makes the others findable is a plain
 index at `~/.nexusmem/projects.json`, written by `init` and refreshed by every `sync`. `nexusmem
@@ -458,7 +497,7 @@ Deleting `.nexusmem/` loses nothing that `sync` cannot rebuild.
 
 ## Status
 
-Ingestion, hybrid retrieval, budgeted packing and the MCP server all work and are covered by 493
+Ingestion, hybrid retrieval, budgeted packing and the MCP server all work and are covered by 701
 tests running on Linux and Windows across Node 22 and 24.
 
 ## Development