bitfab-cli 0.2.449 → 0.2.451

package/dist/index.js

@@ -28992,11 +28992,12 @@ var listGraders = {
 var saveDataset = {
   name: "save_dataset",
   title: "Save Dataset",
-  description: "Create or edit a labeled dataset for a traced function. Datasets are named buckets of traces (e.g. 'Hallucination failures', 'Edge cases for grader X') that humans review and that experiments replay against. Pass traceFunctionKey, a short human-readable name, and an optional description. Saving with the name of an existing dataset on the same trace function updates that dataset instead of creating a duplicate, so re-running a flow is safe; pass a new description to change it, or omit it to leave the current one. Returns the dataset id; use it with add_traces_to_dataset to populate. Call list_datasets first when you need to see what already exists.",
+  description: "Create or edit a labeled dataset for a traced function. Datasets are named buckets of traces (e.g. 'Hallucination failures', 'Edge cases for grader X') that humans review and that experiments replay against. To create, pass traceFunctionKey, a short human-readable name, and an optional description. Saving with the name of an existing dataset on the same trace function updates that dataset instead of creating a duplicate, so re-running a flow is safe; pass a new description to change it, or omit it to leave the current one. To rename a dataset or change its description in place, pass its datasetId with a new name and/or description; its id, traces, and graders are kept, and a name already used by another dataset on the same trace function is refused. Returns the dataset id; use it with add_traces_to_dataset to populate. Call list_datasets first when you need to see what already exists.",
   inputSchema: {
-    traceFunctionKey: external_exports.string().min(1).describe("The trace function key the dataset is scoped to (must match traces you'll add)"),
-    name: external_exports.string().min(1).describe("Short human-readable name shown in the labeling UI (e.g. 'Hallucination failures')"),
-    description: external_exports.string().optional().describe("Optional longer description of what's in the dataset")
+    datasetId: external_exports.uuid().optional().describe("Existing dataset id (from list_datasets) to rename or redescribe in place. Omit to create a dataset or save one by name."),
+    traceFunctionKey: external_exports.string().min(1).optional().describe("The trace function key the dataset is scoped to (must match traces you'll add). Required without datasetId."),
+    name: external_exports.string().min(1).optional().describe("Short human-readable name shown in the labeling UI (e.g. 'Hallucination failures'). Required without datasetId. With datasetId, the new name; omit to keep the current one."),
+    description: external_exports.string().optional().describe("Optional longer description of what's in the dataset. With datasetId, omit to keep the current one or pass an empty string to clear it.")
   }
 };
 var listDatasets = {
@@ -29111,7 +29112,7 @@ var listExperiments = {
 var listExperimentTraces = {
   name: "list_experiment_traces",
   title: "List Experiment Traces",
-  description: "Get individual trace results for an experiment, including each replay trace's verdict (fixed, regressed, still-passing, still-failing, original-unlabeled, replay-unlabeled, capture-incomplete, unpaired, skipped) by comparing against the original trace's label, plus token usage (input, output, cached, total) for the replay and the paired original so you can reason about cost and cache-read deltas. Use after list_experiments to drill into a specific experiment's results.",
+  description: "Get individual trace results for an experiment, including each replay trace's verdict (fixed, regressed, still-passing, still-failing, original-unlabeled, replay-unlabeled, capture-incomplete, unpaired, skipped) by comparing against the original trace's label, plus the replay's duration and token usage (input, output, cached, total) for the replay and the paired original so you can reason about latency, cost, and cache-read deltas. Use after list_experiments to drill into a specific experiment's results.",
   inputSchema: {
     experimentId: external_exports.uuid().describe("The experiment ID to get traces for")
   }
@@ -30025,10 +30026,10 @@ var semver3 = __toESM(require_semver2(), 1);
 
 // ../bitfab-plugin-lib/dist/bakedSdkVersions.js
 var BAKED_SDK_VERSIONS = {
-  typescript: "0.64.25",
-  python: "0.64.22",
-  ruby: "0.64.20",
-  go: "0.64.21"
+  typescript: "0.64.27",
+  python: "0.64.24",
+  ruby: "0.64.22",
+  go: "0.64.23"
 };
 
 // ../bitfab-plugin-lib/dist/installedSdk.js

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "bitfab-cli",
-  "version": "0.2.449",
+  "version": "0.2.451",
   "description": "Install and configure the Bitfab plugin in Claude Code, Codex, or Cursor, or the Bitfab skill pack in Amp (Alpha).",
   "type": "module",
   "main": "dist/index.js",