@dalmia/calibrate-mcp 0.0.89 → 0.0.91

package/bin/mcp-server.js

@@ -6570,7 +6570,15 @@ function writeFromReadableStream(stream, writable) {
   }
   return writeFromReadableStreamDefaultReader(stream.getReader(), writable);
 }
-var RequestError, toRequestError = (e) => {
+var RequestError, validateRequestTarget = (target) => {
+  const backslash = target.indexOf("\\");
+  if (backslash !== -1) {
+    const query = target.indexOf("?");
+    if (query === -1 || backslash < query) {
+      throw new RequestError("Backslash in request path");
+    }
+  }
+}, toRequestError = (e) => {
   if (e instanceof RequestError) {
     return e;
   }
@@ -6640,6 +6648,7 @@ var RequestError, toRequestError = (e) => {
       throw new RequestError("Absolute URL for :path is not allowed in HTTP/2");
     }
     try {
+      validateRequestTarget(incomingUrl);
       const url2 = new URL(incomingUrl);
       req[urlKey] = url2.href;
     } catch (e) {
@@ -6660,6 +6669,7 @@ var RequestError, toRequestError = (e) => {
   } else {
     scheme = incoming.socket && incoming.socket.encrypted ? "https" : "http";
   }
+  validateRequestTarget(incomingUrl);
   const url = new URL(`${scheme}://${host}${incomingUrl}`);
   if (url.hostname.length !== host.length && url.hostname !== host.replace(/:\d+$/, "")) {
     throw new RequestError("Invalid host header");
@@ -53298,9 +53308,9 @@ var init_config = __esm(() => {
   SDK_METADATA = {
     language: "typescript",
     openapiDocVersion: "0.1.0",
-    sdkVersion: "0.0.89",
+    sdkVersion: "0.0.91",
     genVersion: "2.915.1",
-    userAgent: "speakeasy-sdk/mcp-typescript 0.0.89 2.915.1 0.1.0 @dalmia/calibrate-mcp"
+    userAgent: "speakeasy-sdk/mcp-typescript 0.0.91 2.915.1 0.1.0 @dalmia/calibrate-mcp"
   };
 });
 
@@ -59105,12 +59115,19 @@ var init_taskagreementresponse = __esm(() => {
 });
 
 // src/models/taskagreementannotationtaskstaskuuidagreementgetop.ts
-var TaskAgreementAnnotationTasksTaskUuidAgreementGetRequest$zodSchema, TaskAgreementAnnotationTasksTaskUuidAgreementGetResponse$zodSchema;
+var TaskAgreementAnnotationTasksTaskUuidAgreementGetDateField$zodSchema, TaskAgreementAnnotationTasksTaskUuidAgreementGetRequest$zodSchema, TaskAgreementAnnotationTasksTaskUuidAgreementGetResponse$zodSchema;
 var init_taskagreementannotationtaskstaskuuidagreementgetop = __esm(() => {
   init_zod();
   init_httpvalidationerror();
   init_taskagreementresponse();
+  TaskAgreementAnnotationTasksTaskUuidAgreementGetDateField$zodSchema = _enum([
+    "row_added",
+    "annotator_answered"
+  ]).describe("Which date `start` and `end` apply to:\n\n- `row_added`: when the row was added to the task\n- `annotator_answered`: when the annotator first gave the answer in their latest labelling job. Editing does not change this date\n\nOmit to use `row_added`");
   TaskAgreementAnnotationTasksTaskUuidAgreementGetRequest$zodSchema = object({
+    date_field: TaskAgreementAnnotationTasksTaskUuidAgreementGetDateField$zodSchema.default("row_added").describe("Which date `start` and `end` apply to:\n\n- `row_added`: when the row was added to the task\n- `annotator_answered`: when the annotator first gave the answer in their latest labelling job. Editing does not change this date\n\nOmit to use `row_added`"),
+    end: datetime2({ offset: true }).describe("Return only rows or answers dated before this time (ISO 8601). A time with no timezone is read as UTC. Omit for no latest date").nullable().optional(),
+    start: datetime2({ offset: true }).describe("Return only rows or answers dated at or after this time (ISO 8601). A time with no timezone is read as UTC. Omit for no earliest date").nullable().optional(),
     task_uuid: string2().describe("Annotation task to act on"),
     xAPIKey: string2().nullable().optional()
   });
@@ -59138,6 +59155,11 @@ async function $do24(client$, request, options) {
     })
   };
   const path$ = pathToFunc("/annotation-tasks/{task_uuid}/agreement")(pathParams$);
+  const query$ = encodeFormQuery({
+    date_field: payload$.date_field,
+    end: payload$.end,
+    start: payload$.start
+  });
   const headers$ = new Headers(compactMap({
     Accept: "application/json",
     "X-API-Key": encodeSimple("X-API-Key", payload$.xAPIKey, {
@@ -59169,6 +59191,7 @@ async function $do24(client$, request, options) {
     baseURL: options?.serverURL,
     path: path$,
     headers: headers$,
+    query: query$,
     body: body$,
     userAgent: client$._options.userAgent,
     timeoutMs: options?.timeoutMs || client$._options.timeoutMs || -1
@@ -59450,7 +59473,7 @@ var init_tasksummaryresponse = __esm(() => {
     evaluators: array(record(string2(), any())).describe("The task's evaluators, stable across pages, each with its live rubric"),
     item_comments: record(string2(), any()).nullable().optional().describe("Annotator comments, keyed by item ID"),
     pagination: PaginationMeta$zodSchema,
-    rows: array(record(string2(), any())).describe("One row per item, evaluator, and version on this page, with the evaluator's value and each annotator's label"),
+    rows: array(record(string2(), any())).describe("One row per item, evaluator, and version on this page, with the evaluator's value and each annotator's label. `item_added_at` is when the row was added (ISO 8601 UTC). Each label's `answered_at` is when the annotator first gave it in their latest labelling job, and editing does not change it"),
     task_id: string2().describe("ID of the task"),
     task_type: TaskType$zodSchema.describe("Type of the task")
   }).catchall(any()), "AdditionalProperties", true);
@@ -59473,18 +59496,24 @@ var init_tasksummaryresponse = __esm(() => {
 });
 
 // src/models/tasksummaryannotationtaskstaskuuidsummarygetop.ts
-var Order$zodSchema, TaskSummaryAnnotationTasksTaskUuidSummaryGetRequest$zodSchema, TaskSummaryAnnotationTasksTaskUuidSummaryGetResponse$zodSchema;
+var TaskSummaryAnnotationTasksTaskUuidSummaryGetDateField$zodSchema, Order$zodSchema, TaskSummaryAnnotationTasksTaskUuidSummaryGetRequest$zodSchema, TaskSummaryAnnotationTasksTaskUuidSummaryGetResponse$zodSchema;
 var init_tasksummaryannotationtaskstaskuuidsummarygetop = __esm(() => {
   init_zod();
   init_httpvalidationerror();
   init_tasksummaryresponse();
+  TaskSummaryAnnotationTasksTaskUuidSummaryGetDateField$zodSchema = _enum([
+    "row_added",
+    "annotator_answered"
+  ]).describe("Which date `start` and `end` apply to:\n\n- `row_added`: when the row was added to the task\n- `annotator_answered`: when the annotator first gave the answer in their latest labelling job. Editing does not change this date\n\nOmit to use `row_added`");
   Order$zodSchema = _enum([
     "asc",
     "desc"
   ]).describe("Sort direction");
   TaskSummaryAnnotationTasksTaskUuidSummaryGetRequest$zodSchema = object({
     compact: boolean2().default(false).describe("Return a compact response that omits heavy detail fields (`rows.payload`, `rows.evaluator_reasoning`, `rows.annotations.reasoning`, `evaluators.versions.system_prompt`, `evaluators.versions.output_config`, `evaluators.versions.variables`, `item_comments`), keeping only the lightweight decision fields. Omit for full detail"),
+    date_field: TaskSummaryAnnotationTasksTaskUuidSummaryGetDateField$zodSchema.default("row_added").describe("Which date `start` and `end` apply to:\n\n- `row_added`: when the row was added to the task\n- `annotator_answered`: when the annotator first gave the answer in their latest labelling job. Editing does not change this date\n\nOmit to use `row_added`"),
     disagreement_only: boolean2().default(false).describe("When true, keep only rows where the evaluator disagreed with at least one annotator"),
+    end: datetime2({ offset: true }).describe("Return only rows or answers dated before this time (ISO 8601). A time with no timezone is read as UTC. Omit for no latest date").nullable().optional(),
     item_id: string2().describe("Filter rows to a single item. The full task-wide annotator union is still returned in `annotators`").nullable().optional(),
     limit: int().default(50).describe("Maximum number of items to return"),
     live_only: boolean2().default(false).describe("When true, emit only one row for each (item, evaluator) pair using the evaluator's live version. Versions other than the live one that have runs are excluded"),
@@ -59492,6 +59521,7 @@ var init_tasksummaryannotationtaskstaskuuidsummarygetop = __esm(() => {
     order: Order$zodSchema.default("desc").describe("Sort direction"),
     q: string2().describe("Case-insensitive substring search on `payload.name`. Blank is a no-op").nullable().optional(),
     sort_by: string2().default("created_at").describe("Sort key for the results"),
+    start: datetime2({ offset: true }).describe("Return only rows or answers dated at or after this time (ISO 8601). A time with no timezone is read as UTC. Omit for no earliest date").nullable().optional(),
     task_uuid: string2().describe("Annotation task to act on"),
     xAPIKey: string2().nullable().optional()
   });
@@ -59521,14 +59551,17 @@ async function $do26(client$, request, options) {
   const path$ = pathToFunc("/annotation-tasks/{task_uuid}/summary")(pathParams$);
   const query$ = encodeFormQuery({
     compact: payload$.compact,
+    date_field: payload$.date_field,
     disagreement_only: payload$.disagreement_only,
+    end: payload$.end,
     item_id: payload$.item_id,
     limit: payload$.limit,
     live_only: payload$.live_only,
     offset: payload$.offset,
     order: payload$.order,
     q: payload$.q,
-    sort_by: payload$.sort_by
+    sort_by: payload$.sort_by,
+    start: payload$.start
   });
   const headers$ = new Headers(compactMap({
     Accept: "application/json",
@@ -63155,7 +63188,7 @@ var init_tracesummary = __esm(() => {
     metadata_count: int().describe("Number of metadata entries stored with the trace"),
     output: TraceOutput$zodSchema.nullable().optional().describe("What the agent produced for this turn. Sent only with `include_content`"),
     response_preview: string2().nullable().optional().describe("The agent reply, truncated for display"),
-    results: array(TraceScoreResult$zodSchema).optional().describe("Results for each evaluator on the latest scoring run"),
+    results: array(TraceScoreResult$zodSchema).optional().describe("Results on the latest scoring run from the evaluators linked to the agent"),
     tool_call_count: int().describe("Number of tool calls the agent issued for this turn"),
     tool_calls: array(TraceToolCall$zodSchema).describe("Tools the agent issued on this turn, with the arguments it passed"),
     tool_names: array(string2()).describe("Names of the tools the agent issued on this turn, in order"),
@@ -63174,7 +63207,7 @@ var init_tracelistresponse = __esm(() => {
     items: array(TraceSummary$zodSchema).describe("The page of results"),
     limit: int().nullable().optional().describe("The applied `limit` (null when unbounded)"),
     offset: int().default(0).describe("The applied `offset`"),
-    score_averages: array(TraceScoreAverage$zodSchema).nullable().optional().describe("One entry per evaluator that scored at least one matching trace, averaged over every matching trace rather than this page. Ask for it with `include_score_averages`"),
+    score_averages: array(TraceScoreAverage$zodSchema).nullable().optional().describe("One average for each evaluator linked to the agent that scored at least one trace matching the filters, covering all the matching traces it scored rather than this page. Ask for it with `include_score_averages`"),
     total: int().describe("Total number of items matching the query, before pagination")
   }).describe("The page, plus the averages when `include_score_averages` asked for them.");
 });
@@ -63196,8 +63229,8 @@ var init_listtracesendpointtracesgetop = __esm(() => {
   ListTracesEndpointTracesGetRequest$zodSchema = object({
     agent_id: string2().describe("Return only traces from this agent").nullable().optional(),
     conversation_id: string2().describe("Return only the turns of this conversation. Requires `agent_id`").nullable().optional(),
-    created_after: datetime2({ offset: true }).describe("Return only traces created at or after this time (ISO 8601). A time with no timezone is read as UTC").nullable().optional(),
-    created_before: datetime2({ offset: true }).describe("Return only traces created before this time (ISO 8601). A time with no timezone is read as UTC").nullable().optional(),
+    created_after: datetime2({ offset: true }).describe("Return only traces created at or after this time (ISO 8601). A time with no timezone is read as UTC. Omit for no earliest date").nullable().optional(),
+    created_before: datetime2({ offset: true }).describe("Return only traces created before this time (ISO 8601). A time with no timezone is read as UTC. Omit for no latest date").nullable().optional(),
     include_content: boolean2().default(false).describe("Also return each trace's full input, output and metadata. Use `limit` and `offset` to page through them"),
     include_score_averages: boolean2().default(false).describe("Also return each evaluator's mean score over every matching trace, not just this page. Costs a pass over those traces, so ask for it when the filters change rather than on every page"),
     labels: array(string2()).describe("Return only traces carrying at least one of these labels. Repeat the parameter for each label").nullable().optional(),
@@ -63206,7 +63239,7 @@ var init_listtracesendpointtracesgetop = __esm(() => {
     output_type: ListTracesEndpointTracesGetOutputType$zodSchema.nullable().optional().describe("Return only traces whose output is of this kind. `response` covers every trace carrying a reply, including one that also issued tool calls. `tool_call` covers traces that only issued tool calls"),
     q: string2().describe("Return only traces containing this text in their message ID, conversation ID, conversation history, output, or metadata").nullable().optional(),
     score: array(string2()).describe("Return only traces whose latest scoring run matches this condition. Each is written as `<evaluator ID>:<condition>`, where the condition is `passed`, `failed`, or one of `=`, `!=`, `<`, `<=`, `>`, `>=` followed by a number, such as `>=4`. Repeat the parameter to require every condition at once").nullable().optional(),
-    sort_by_evaluator: string2().describe("Order traces by this evaluator's score on each trace's latest scoring run. Traces it has not scored come last whichever direction you pick").nullable().optional(),
+    sort_by_evaluator: string2().describe("Order traces by this evaluator's score on each trace's latest scoring run. Traces this evaluator has not scored come last whichever direction you pick, and an evaluator not linked to the agent counts as having scored none").nullable().optional(),
     sort_order: SortOrder$zodSchema.default("desc").describe("Whether `sort_by_evaluator` puts the highest scores first or the lowest. Ignored without it"),
     xAPIKey: string2().nullable().optional()
   });
@@ -63354,7 +63387,7 @@ and metadata.
 function createMCPServer(deps) {
   const server = new McpServer({
     name: "CalibrateMcp",
-    version: "0.0.89"
+    version: "0.0.91"
   });
   const getClient = deps.getSDK || (() => new CalibrateMcpCore({
     security: deps.security,
@@ -64692,7 +64725,7 @@ http_headers = { "api-key-auth" = "YOUR_API_KEY_AUTH" }`;
         <h1>Instructions</h1>
         <p>One-click installation for Claude Desktop users</p>
         <div class="instruction-item">
-          <a href="https://github.com/dalmia/calibrate-mcp/releases/download/v0.0.89/mcp-server.mcpb" download="mcp-server.mcpb" class="action-button header-action" style="display: inline-flex; margin-bottom: 16px;">
+          <a href="https://github.com/dalmia/calibrate-mcp/releases/download/v0.0.91/mcp-server.mcpb" download="mcp-server.mcpb" class="action-button header-action" style="display: inline-flex; margin-bottom: 16px;">
             \uD83D\uDCE5 Download MCP Bundle
           </a>
         </div>
@@ -67646,7 +67679,7 @@ var routes = buildRouteMap({
 var app = buildApplication(routes, {
   name: "mcp",
   versionInfo: {
-    currentVersion: "0.0.89"
+    currentVersion: "0.0.91"
   }
 });
 run(app, process3.argv.slice(2), buildContext(process3));
@@ -67654,5 +67687,5 @@ export {
   app
 };
 
-//# debugId=5A80FFDF0EAFC95564756E2164756E21
+//# debugId=0AC59BD4C954B3B164756E2164756E21
 //# sourceMappingURL=mcp-server.js.map