grepmax 0.26.81 → 0.26.82

package/.claude-plugin/marketplace.json

@@ -10,7 +10,7 @@
       "name": "grepmax",
       "source": "./plugins/grepmax",
       "description": "Semantic code search for Claude Code. Automatically indexes your project and provides intelligent search capabilities.",
-      "version": "0.26.81",
+      "version": "0.26.82",
       "author": {
         "name": "Robert Owens",
         "email": "robowens@me.com"

package/README.md

@@ -524,7 +524,7 @@ To serve bounded reads under an existing startup quarantine, run `gmax watch --d
 
 Full-table compaction and new FTS/ANN index builds are disabled in the containment release, including forced maintenance and doctor repair. Existing indexes remain readable; missing lexical indexes fall back to available retrieval. Do not repeatedly force repair, use future cutoffs or manually remove index fragments.
 
-The package includes a separate prune-only helper with pinned Lance 12 dependencies, plus an exclusively leased Node wrapper. Runtime preparation uses `uv >=0.12.18`, an existing Python interpreter and locked wheels; it does not download Python, build from source, load an embedding model or run during installation. The helper currently supports macOS and Linux. The daemon runs independent prune-only retention: at most 128 old versions per pass, keeping current, tagged and two-minute recent versions. It retries backlog every minute and checks settled history every five minutes; status polling and watched edits do not prevent cleanup. Runtime setup happens before writer suspension. Older readers and external store owners are drained under an exclusive lease; new read-only queries may use the protected current version while history is deleted. Writers resume and watchers reconcile edits after the pass. Full-table compaction admission never gates this reclamation. Normal configured host policy applies; explicit offline recovery retains its separate admission.
+The package includes a separate prune-only helper with pinned Lance 12 dependencies, plus an exclusively leased Node wrapper. Runtime preparation uses `uv >=0.12.18`, an existing Python interpreter and locked wheels; it does not download Python, build from source, load an embedding model or run during installation. The helper currently supports macOS and Linux. The daemon runs independent prune-only retention: at most 128 old versions per pass, keeping current, tagged and two-minute recent versions. It retries backlog every minute and checks settled history every five minutes; status polling and watched edits do not prevent cleanup. Runtime setup happens before writer suspension. Bulk project indexing defers retention without stopping watchers or closing reads, with a one-minute retry. Reads also remain available while watcher work quiesces; admission closes briefly to drain old native readers before pruning starts. Older readers and external store owners are drained under an exclusive lease; new read-only queries may use the protected current version while history is deleted. Writers resume and watchers reconcile edits after the pass. Full-table compaction admission never gates this reclamation. Normal configured host policy applies; explicit offline recovery retains its separate admission.
 
 `gmax recover --table /path/to/lancedb/chunks.lance --json` reads the durable recovery receipt without starting a daemon or preparing Python. `--check` samples admission without pruning. `--prune --version <current-version> --cutoff <absolute-ISO-timestamp>` explicitly requests one offline prune. A prior uncertain attempt requires `--acknowledge-uncertain <attempt-id>` after inspection. Recovery never creates/clears quarantine, stops other owners, retries automatically or enables rewriting.
 

package/dist/lib/daemon/daemon.js

@@ -163,6 +163,7 @@ class Daemon {
         this.cleanupInterval = null;
         this.cleanupPromise = null;
         this.lastCleanupAttempt = 0;
+        this.cleanupRetrySoon = false;
         this.heartbeatTick = 0;
         this.shuttingDown = false;
         this.recycling = false;
@@ -854,7 +855,8 @@ class Daemon {
             return;
         this.cleanupInterval = setInterval(() => {
             var _a;
-            const interval = ((_a = this.versionCleanupStatus()) === null || _a === void 0 ? void 0 : _a.eligibleVersionsRemaining) === 0
+            const interval = !this.cleanupRetrySoon &&
+                ((_a = this.versionCleanupStatus()) === null || _a === void 0 ? void 0 : _a.eligibleVersionsRemaining) === 0
                 ? 5 * 60000
                 : 60000;
             if (!this.ready ||
@@ -875,6 +877,7 @@ class Daemon {
         if (this.cleanupPromise)
             return this.cleanupPromise;
         this.lastCleanupAttempt = Date.now();
+        this.cleanupRetrySoon = false;
         const run = this.performVersionCleanup();
         const tracked = run.finally(() => {
             if (this.cleanupPromise === tracked)
@@ -891,9 +894,26 @@ class Daemon {
                 this.recycling ||
                 this.pausedReason !== null)
                 throw new Error("daemon resources not ready for version cleanup");
+            const bulkIndexingActive = () => this.operations
+                .activeOperationNames()
+                .some((name) => [
+                "add-project",
+                "ensure-project",
+                "index-project",
+                "index-pending",
+            ].includes(name));
+            const deferForIndexing = () => {
+                this.cleanupRetrySoon = true;
+                return (0, compaction_result_1.skippedCompaction)("project indexing is active; version cleanup deferred without closing reads");
+            };
+            if (bulkIndexingActive())
+                return deferForIndexing();
             this.assertHeavyOperationAdmission("version-cleanup");
             const deadline = AbortSignal.timeout(60000);
             const runtime = yield (0, version_cleanup_1.prepareVersionCleanup)(config_1.PATHS.lancedbDir, deadline);
+            // Runtime setup yields: a bulk index can start before exclusive intent.
+            if (bulkIndexingActive())
+                return deferForIndexing();
             let roots = [];
             try {
                 return yield this.operations.runExclusive("version-cleanup", () => __awaiter(this, void 0, void 0, function* () {
@@ -919,7 +939,11 @@ class Daemon {
                         return (0, compaction_result_1.skippedCompaction)("no indexed table");
                     (0, logger_1.log)("daemon", `Version cleanup completed: ${JSON.stringify(result)}`);
                     return result;
-                }), { queueShared: true });
+                }), {
+                    queueShared: true,
+                    pendingReads: (name) => (0, read_verbs_1.getReadVerb)(name) !== undefined ||
+                        ["search", "keyword-search", "documents", "project-stats"].includes(name),
+                });
             }
             finally {
                 if (!this.shuttingDown &&

package/dist/lib/utils/operation-coordinator.js

@@ -117,6 +117,9 @@ class OperationCoordinator {
     }
     runExclusive(name, quiesce, fn, options = {}) {
         var _a;
+        if (options.pendingReads &&
+            (name !== "version-cleanup" || !options.queueShared))
+            return Promise.reject(new Error("pending reads require queued prune-only cleanup"));
         if (this.state.kind === "closing" || this.state.kind === "closed") {
             return Promise.reject(new OperationClosedError());
         }
@@ -126,11 +129,16 @@ class OperationCoordinator {
         }
         this.state = { kind: "exclusive-pending", name };
         this.queueShared = (_a = options.queueShared) !== null && _a !== void 0 ? _a : false;
+        this.readWindow = options.pendingReads;
         const controller = new AbortController();
         this.controllers.add(controller);
         const task = (() => __awaiter(this, void 0, void 0, function* () {
             try {
                 yield quiesce();
+                // Reads remain available while writers quiesce. Close admission before
+                // taking the drain snapshot so every old native reader still settles
+                // before retention closes handles and deletes history.
+                this.readWindow = undefined;
                 yield Promise.allSettled([...this.sharedTasks]);
                 if (controller.signal.aborted) {
                     throw abortError(controller.signal.reason);
@@ -201,7 +209,7 @@ class OperationCoordinator {
         }
         if (this.state.kind === "exclusive" ||
             this.state.kind === "exclusive-pending") {
-            if (this.state.kind === "exclusive" && ((_a = this.readWindow) === null || _a === void 0 ? void 0 : _a.call(this, name)))
+            if ((_a = this.readWindow) === null || _a === void 0 ? void 0 : _a.call(this, name))
                 return;
             throw new OperationBusyError(this.state.name);
         }

package/dist/lib/workers/embeddings/mlx-client.js

@@ -102,6 +102,10 @@ function requestMlxJSON(reqPath, body) {
             hostname: MLX_HOST,
             port: MLX_PORT,
             path: reqPath,
+            // A local connection is cheap. Avoid reusing a socket that uvicorn
+            // closed at its keep-alive boundary: that reset can incorrectly mark
+            // the healthy embedding backend unavailable and force CPU fallback.
+            agent: false,
             method: payload === undefined ? "GET" : "POST",
             headers: payload === undefined
                 ? undefined

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "grepmax",
-  "version": "0.26.81",
+  "version": "0.26.82",
   "author": "Robert Owens <78518764+reowens@users.noreply.github.com>",
   "homepage": "https://github.com/reowens/grepmax",
   "bugs": {

package/plugins/grepmax/.claude-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "grepmax",
-  "version": "0.26.81",
+  "version": "0.26.82",
   "description": "Semantic code search for Claude Code. Automatically indexes your project and provides intelligent search capabilities.",
   "author": {
     "name": "Robert Owens",