bitfab-cli 0.2.441 → 0.2.443

package/dist/index.js

@@ -28854,7 +28854,7 @@ var assertionTargetShape = external_exports.discriminatedUnion("kind", [
       external_exports.number().int().min(0)
     ]).optional().describe('Which call to check when the trace runs that span more than once, e.g. a retry loop. "first", "last" (the default), or a 0-based index. Leave it off unless the repetition matters.')
   }).describe("Check one span by name rather than the whole trace, e.g. that a particular tool call was made with the right arguments.")
-]).describe('SCOPE: what part of the trace under evaluation this assertion is checked against. OMIT IT for the whole trace, which is the common case and the right default when unsure. Pass { "kind": "output" } to check only the final output, or { "kind": "span", "name": "..." } to check one span. Scoping narrows what a judge looks at, so a wrong scope hides real failures. A target naming something the evaluated trace does not contain makes the check ERRORED, never passed, because a check that could not run must never look like a check that succeeded.');
+]).describe('SCOPE: what part of the trace under evaluation this assertion is checked against. OMIT IT for the whole trace, which is the common case and the right default when unsure. On an edit, omitting it keeps the current scope, and passing null returns the assertion to the whole trace. Pass { "kind": "output" } to check only the final output, or { "kind": "span", "name": "..." } to check one span. Scoping narrows what a judge looks at, so a wrong scope hides real failures. A target naming something the evaluated trace does not contain makes the check ERRORED, never passed, because a check that could not run must never look like a check that succeeded.');
 var justificationShape = external_exports.preprocess(parseJsonString, external_exports.array(external_exports.object({
   spanId: external_exports.uuid().describe("The span this reason comes from, taken from the `[ID: ...]` shown for that span in get_traces output."),
   text: external_exports.string().trim().min(1).max(1e4).describe("What that span shows, in one or two sentences.")
@@ -28919,9 +28919,9 @@ var saveTraceAssertions = {
       humanNote: external_exports.string().max(1e4).optional().describe("People-only context attached to this assertion. It is returned by reads but must never be used as evidence when assessing a replay. On an edit, omit to preserve it and pass an empty string to clear it."),
       passCriteria: external_exports.string().optional().describe("How a judge should recognise a pass, when the assertion alone leaves room to argue, e.g. 'arrival timestamp is strictly before 09:00 in the destination timezone'. Optional, and only worth writing when it removes real ambiguity. Same field save_grader takes, so an assertion that proves out across many traces is promoted into a grader by copying it. On an edit, omit to keep the current value and pass an empty string to clear it."),
       failCriteria: external_exports.string().optional().describe("How a judge should recognise a failure, for cases the pass criteria do not obviously exclude, e.g. 'any leg departing after 09:00, including connections'. Optional. On an edit, omit to keep the current value and pass an empty string to clear it."),
-      targetOnEvaluatedTrace: assertionTargetShape.optional(),
-      justification: justificationShape.optional().describe("Why this assertion is right for THIS trace, cited span by span: the spans that show what the run actually did, each with its own reason. Omit on an edit to preserve it. It is the evidence behind `approvalState`, so an assertion with no justification is one a reviewer has to reconstruct from scratch. Cite spans on this trace only."),
-      categoryJustification: justificationShape.optional().describe("Why `category_assertion_id` is the right category for this assertion, cited span by span. Omit on an edit to preserve it. Changing the category clears it, since it justified the old one. Cite spans on this trace only."),
+      targetOnEvaluatedTrace: assertionTargetShape.nullish(),
+      justification: justificationShape.nullish().describe("Why this assertion is right for THIS trace, cited span by span: the spans that show what the run actually did, each with its own reason. Omit on an edit to preserve it, or pass null to remove it. It is the evidence behind `approvalState`, so an assertion with no justification is one a reviewer has to reconstruct from scratch. Cite spans on this trace only."),
+      categoryJustification: justificationShape.nullish().describe("Why `category_assertion_id` is the right category for this assertion, cited span by span. Omit on an edit to preserve it, or pass null to remove it. Changing the category clears it, since it justified the old one. Cite spans on this trace only."),
       approvalState: optionalApprovalState(),
       assigneeEmail: external_exports.email().max(MAX_ASSIGNEE_EMAIL_LENGTH).nullish().describe("Email of the organization member who should review this assertion, which routes it into their queue in Bitfab. Assigning decides nothing, so it leaves the approval state alone. The email must belong to a member of this organization, or the call fails naming it. Omit on an edit to preserve the current assignee, and pass null to leave it to nobody.")
     })).min(1).max(MAX_ASSERTIONS_PER_REQUEST)).describe(`One entry per assertion (1-${MAX_ASSERTIONS_PER_REQUEST})`)
@@ -29275,9 +29275,9 @@ var saveScenario = {
 var generateScenarioAssertions = {
   name: "generate_scenario_assertions",
   title: "Generate Scenario Assertions",
-  description: "Turn a person's label on the scenario built from one trace into up to four assertions, saved as pending agent assertions on that trace for review. The note says what the run got right or wrong; the verdict says whether the run was right overall. Each assertion cites the spans that settle its values, from the scenario's trigger, history, and records the run saw; records the run never saw are context only. A note that gives nothing a reviewer could check against a replay yields no assertions. A trace that is a replay, or a trace with no scenario, is refused. Returns the saved assertions.",
+  description: "Turn a person's label on one scenario, found by its id or by a trace it was built from, into up to four assertions, saved as pending agent assertions for review on the trace the scenario was built from. Create the scenario first with save_scenario and check its trigger, data, and records before calling this. The note says what the run got right or wrong; the verdict says whether the run was right overall. Each assertion cites the spans that settle its values, from the scenario's trigger, history, and records the run saw, and is saved on the trace those spans are on; records the run never saw are context only. A note that gives nothing a reviewer could check against a replay yields no assertions. A trace with no scenario, a scenario id that does not exist, or a scenario built from a replay is refused. Returns the saved assertions.",
   inputSchema: {
-    traceId: external_exports.uuid().describe("ID of the trace whose run the person labeled. Its scenario supplies the context, and the assertions are saved on this trace."),
+    ...scenarioTargetSchema,
     verdict: external_exports.preprocess(parseJsonString, external_exports.boolean()).optional().describe("true when the person said the run was right, false when wrong. Leave it out when they gave no verdict."),
     note: external_exports.string().min(1).describe("What the person said was right or wrong about the run, in their words.")
   }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "bitfab-cli",
-  "version": "0.2.441",
+  "version": "0.2.443",
   "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",