@dalmia/calibrate-mcp 0.0.57 → 0.0.59

package/bin/mcp-server.js

@@ -455,9 +455,6 @@ function members(proto, table) {
     else
       defineBound(proto, key, desc.value);
   }
-  for (const sym of Object.getOwnPropertySymbols(table)) {
-    defineBound(proto, sym, table[sym]);
-  }
 }
 function own(inst, key, value, enumerable = true) {
   Object.defineProperty(inst, key, { configurable: true, writable: true, enumerable, value });
@@ -632,8 +629,7 @@ function $constructor(name, initializer, proto, params) {
       } finally {
         _zodDesc.value = undefined;
       }
-    }
-    if (inst._zod.traits.has(name)) {
+    } else if (inst._zod.traits.has(name)) {
       return;
     }
     inst._zod.traits.add(name);
@@ -1435,7 +1431,7 @@ var init_versions = __esm(() => {
   version2 = {
     major: 4,
     minor: 6,
-    patch: 2
+    patch: 5
   };
 });
 
@@ -1459,11 +1455,32 @@ function standardProps(inst) {
     version: 1
   };
 }
+function canParseURL(input) {
+  try {
+    if (typeof URL !== "undefined" && typeof URL.canParse === "function")
+      return URL.canParse(input);
+    new URL(input);
+    return true;
+  } catch {
+    return false;
+  }
+}
+function validateURL(trimmed, def) {
+  if (!("normalize" in def) && !("hostname" in def) && !("protocol" in def)) {
+    return canParseURL(trimmed) || URL_UNPARSEABLE;
+  }
+  return parseURLObject(trimmed, def);
+}
 function parseURLObject(trimmed, def) {
   if (!def.normalize && def.protocol?.source === httpProtocol.source && !/^https?:\/\//i.test(trimmed)) {
     return URL_BAD_FORMAT;
   }
   try {
+    if (typeof URL !== "undefined") {
+      const URLStatic = URL;
+      if (typeof URLStatic.parse === "function")
+        return URLStatic.parse(trimmed) ?? URL_UNPARSEABLE;
+    }
     return new URL(trimmed);
   } catch {
     return URL_UNPARSEABLE;
@@ -1483,12 +1500,7 @@ function urlProtocolOk(url, protocol) {
 function isValidIPv6(value) {
   if (!ipv6Alphabet.test(value))
     return false;
-  try {
-    new URL(`http://[${value}]`);
-    return true;
-  } catch {
-    return false;
-  }
+  return canParseURL(`http://[${value}]`);
 }
 function isValidCIDRv6(value) {
   const parts = value.split("/");
@@ -1850,12 +1862,7 @@ function handleRefineResult(result, payload, input, inst) {
     payload.issues.push(issue(_iss));
   }
 }
-function handlePropertiesResult(result, payload, key) {
-  if (result.issues.length) {
-    payload.issues.push(...prefixIssues(key, result.issues));
-  }
-}
-var $ZodType, toStandardResult = (r, ctx) => r.issues.length ? { issues: r.issues.map((iss) => finalizeIssue(iss, ctx, config())) } : { value: r.value }, $ZodString, $ZodStringFormat, $ZodGUID, $ZodUUID, $ZodEmail, URL_BAD_FORMAT = 1, URL_UNPARSEABLE = 2, asciiTabOrNewline, $ZodURL, $ZodEmoji, $ZodNanoID, $ZodCUID, $ZodCUID2, $ZodULID, $ZodXID, $ZodKSUID, $ZodISODateTime, $ZodISODate, $ZodISOTime, $ZodISODuration, $ZodIPv4, ipv6Alphabet, $ZodIPv6, $ZodCIDRv4, $ZodCIDRv6, base64Charset, $ZodBase64, base64urlCharset, $ZodBase64URL, $ZodE164, $ZodJWT, $ZodNumber, $ZodNumberFormat, $ZodBoolean, $ZodNull, $ZodAny, $ZodUnknown, $ZodNever, $ZodArray, NO_SYMBOL_KEYS, $ZodObject, $ZodObjectJIT, $ZodUnion, $ZodDiscriminatedUnion, $ZodIntersection, $ZodRecord, $ZodEnum, $ZodLiteral, $ZodTransform, $ZodOptional, $ZodExactOptional, $ZodNullable, $ZodDefault, $ZodPrefault, $ZodNonOptional, $ZodCatch, $ZodPipe, $ZodPreprocess, $ZodReadonly, $ZodCustom, $ZodProperties;
+var $ZodType, toStandardResult = (r, ctx) => r.issues.length ? { issues: r.issues.map((iss) => finalizeIssue(iss, ctx, config())) } : { value: r.value }, $ZodString, $ZodStringFormat, $ZodGUID, $ZodUUID, $ZodEmail, URL_BAD_FORMAT = 1, URL_UNPARSEABLE = 2, asciiTabOrNewline, $ZodURL, $ZodEmoji, $ZodNanoID, $ZodCUID, $ZodCUID2, $ZodULID, $ZodXID, $ZodKSUID, $ZodISODateTime, $ZodISODate, $ZodISOTime, $ZodISODuration, $ZodIPv4, ipv6Alphabet, $ZodIPv6, $ZodCIDRv4, $ZodCIDRv6, base64Charset, $ZodBase64, base64urlCharset, $ZodBase64URL, $ZodE164, $ZodJWT, $ZodNumber, $ZodNumberFormat, $ZodBoolean, $ZodNull, $ZodAny, $ZodUnknown, $ZodNever, $ZodArray, NO_SYMBOL_KEYS, $ZodObject, $ZodObjectJIT, $ZodUnion, $ZodDiscriminatedUnion, $ZodIntersection, $ZodRecord, $ZodEnum, $ZodLiteral, $ZodTransform, $ZodOptional, $ZodExactOptional, $ZodNullable, $ZodDefault, $ZodPrefault, $ZodNonOptional, $ZodCatch, $ZodPipe, $ZodPreprocess, $ZodReadonly, $ZodCustom;
 var init_schemas = __esm(() => {
   init_checks();
   init_core();
@@ -2028,7 +2035,7 @@ var init_schemas = __esm(() => {
     inst._zod.check = (payload) => {
       try {
         const trimmed = payload.value.trim();
-        const url = parseURLObject(trimmed, def);
+        const url = validateURL(trimmed, def);
         if (url === URL_BAD_FORMAT) {
           payload.issues.push({
             code: "invalid_format",
@@ -2050,6 +2057,10 @@ var init_schemas = __esm(() => {
           });
           return;
         }
+        if (url === true) {
+          payload.value = stripTabAndNewline(trimmed);
+          return;
+        }
         if (def.hostname && !urlHostnameOk(url, def.hostname)) {
           payload.issues.push({
             code: "invalid_format",
@@ -3065,61 +3076,11 @@ var init_schemas = __esm(() => {
       return;
     };
   });
-  $ZodProperties = /* @__PURE__ */ $constructor("$ZodProperties", (inst, def) => {
-    $ZodType.init(inst, def);
-    $ZodCheck.init(inst, def);
-    const memo = globalConfig.memoizer;
-    memo?.attach(inst);
-    let entries;
-    const runShape = (payload, ctx) => {
-      entries ?? (entries = Reflect.ownKeys(def.shape).map((key) => [key, def.shape[key]]));
-      const input = payload.value;
-      let proms;
-      for (const [key, schema] of entries) {
-        const result = schema._zod.run({ value: input[key], issues: [] }, ctx);
-        if (result instanceof Promise) {
-          proms ?? (proms = []);
-          proms.push(result.then((result) => handlePropertiesResult(result, payload, key)));
-        } else {
-          handlePropertiesResult(result, payload, key);
-        }
-      }
-      if (proms)
-        return Promise.all(proms).then(() => {
-          return;
-        });
-      return;
-    };
-    inst._zod.parse = (payload, ctx) => {
-      const input = payload.value;
-      if (input === null || typeof input !== "object" && typeof input !== "function") {
-        payload.issues.push({ expected: "object", code: "invalid_type", input, inst });
-        return payload;
-      }
-      if (ctx.direction === "backward")
-        ctx = { ...ctx, direction: "forward" };
-      if (memo)
-        memo.alloc(inst, payload, input, ctx);
-      const result = runShape(payload, ctx);
-      return result instanceof Promise ? result.then(() => payload) : payload;
-    };
-    inst._zod.check = (payload) => {
-      if (payload.value == null) {
-        payload.issues.push({ expected: "object", code: "invalid_type", input: payload.value, inst });
-        return;
-      }
-      return runShape(payload, {});
-    };
-  }, {
-    *[Symbol.iterator]() {
-      yield this;
-    }
-  });
 });
 
 // node_modules/zod/v4/core/memoizer.js
 function isRef(value) {
-  return value !== null && (typeof value === "object" || typeof value === "function");
+  return value !== null && typeof value === "object";
 }
 function cloneIssues(issues) {
   return issues.map((iss) => iss.path ? { ...iss, path: iss.path.slice() } : { ...iss });
@@ -3164,9 +3125,6 @@ function isRecursive(inst, stack, resolve) {
       check(def.catchall);
       break;
     }
-    case "properties":
-      merge(shape(def.shape, false));
-      break;
     case "array":
       check(def.element);
       break;
@@ -3434,6 +3392,7 @@ var error = () => {
     base64url: "base64url-encoded string",
     json_string: "JSON string",
     e164: "E.164 number",
+    currency_code: "currency code",
     credit_card: "credit card number",
     iban: "IBAN",
     jwt: "JWT",
@@ -3609,11 +3568,13 @@ var init_compile = __esm(() => {
 });
 
 // node_modules/zod/v4/core/api.js
+function snapshotChecks(def) {
+  if (def.checks)
+    def.checks = [...def.checks];
+  return def;
+}
 function _string(Class, params) {
-  return new Class({
-    type: "string",
-    ...normalizeParams(params)
-  });
+  return new Class(snapshotChecks({ type: "string", ...normalizeParams(params) }));
 }
 function _email(Class, params) {
   return new Class({
@@ -3853,19 +3814,10 @@ function _isoDuration(Class, params) {
   });
 }
 function _number(Class, params) {
-  return new Class({
-    type: "number",
-    checks: [],
-    ...normalizeParams(params)
-  });
+  return new Class(snapshotChecks({ type: "number", checks: [], ...normalizeParams(params) }));
 }
 function _coercedNumber(Class, params) {
-  return new Class({
-    type: "number",
-    coerce: true,
-    checks: [],
-    ...normalizeParams(params)
-  });
+  return new Class(snapshotChecks({ type: "number", coerce: true, checks: [], ...normalizeParams(params) }));
 }
 function _int(Class, params) {
   return new Class({
@@ -5014,17 +4966,15 @@ var narrowMin = (agg, key, value) => {
       path: [...params.path, "properties", key]
     }));
   }
-  const allKeys = new Set(Object.keys(shape));
-  const requiredKeys = new Set([...allKeys].filter((key) => {
+  const requiredKeys = [];
+  for (const key of Object.keys(shape)) {
     const field = def.shape[key];
-    if (ctx.io === "input") {
-      return inputOptin(field) === undefined;
-    } else {
-      return field._zod.optout === undefined;
+    if (ctx.io === "input" ? inputOptin(field) === undefined : field._zod.optout === undefined) {
+      requiredKeys.push(key);
     }
-  }));
-  if (requiredKeys.size > 0) {
-    json.required = Array.from(requiredKeys);
+  }
+  if (requiredKeys.length > 0) {
+    json.required = requiredKeys;
   }
   if (def.catchall?._zod.def.type === "never") {
     json.additionalProperties = false;
@@ -5037,30 +4987,6 @@ var narrowMin = (agg, key, value) => {
       path: [...params.path, "additionalProperties"]
     });
   }
-}, propertiesProcessor = (schema, ctx, _json, params) => {
-  const json = _json;
-  const def = schema._zod.def;
-  if (Object.getOwnPropertySymbols(def.shape).length && handleUnrepresentable(schema, ctx, json, params, "Symbol keys cannot be represented in JSON Schema")) {
-    return;
-  }
-  if (ctx.io === "output") {
-    for (const key in def.shape) {
-      if (isTransforming(def.shape[key]) && handleUnrepresentable(schema, ctx, json, params, `z.properties() returns its input, so the output of a transforming schema at key "${key}" cannot be represented in JSON Schema`)) {
-        return;
-      }
-    }
-  }
-  json.type = "object";
-  json.properties = {};
-  for (const key in def.shape) {
-    assignProp(json.properties, key, processSchema(def.shape[key], ctx, {
-      ...params,
-      path: [...params.path, "properties", key]
-    }));
-  }
-  const required = Object.keys(def.shape).filter((key) => inputOptin(def.shape[key]) === undefined);
-  if (required.length > 0)
-    json.required = required;
 }, unionProcessor = (schema, ctx, json, params) => {
   const def = schema._zod.def;
   const isExclusive = def.inclusive === false;
@@ -5331,7 +5257,6 @@ var init_json_schema_processors = __esm(() => {
     file: fileProcessor,
     success: successProcessor,
     custom: customProcessor,
-    properties: propertiesProcessor,
     function: functionProcessor,
     transform: transformProcessor,
     map: mapProcessor,
@@ -28204,6 +28129,9 @@ var require_proxy_addr = __commonJS(function(exports, module) {
       if (!isip(addr))
         return false;
       var ip = parseip(addr);
+      if (ip.kind() === "ipv6" && ip.isIPv4MappedAddress()) {
+        ip = ip.toIPv4Address();
+      }
       var ipconv;
       var kind = ip.kind();
       for (var i = 0;i < subnets.length; i++) {
@@ -28216,10 +28144,15 @@ var require_proxy_addr = __commonJS(function(exports, module) {
           if (subnetkind === "ipv4" && !ip.isIPv4MappedAddress()) {
             continue;
           }
+          if (subnetkind !== "ipv4" && !(subnetrange >= 96 && subnetip.isIPv4MappedAddress())) {
+            continue;
+          }
           if (!ipconv) {
             ipconv = subnetkind === "ipv4" ? ip.toIPv4Address() : ip.toIPv4MappedAddress();
           }
           trusted = ipconv;
+        } else if (kind === "ipv6" && subnetip.isIPv4MappedAddress()) {
+          continue;
         }
         if (trusted.match(subnetip, subnetrange)) {
           return true;
@@ -28237,12 +28170,20 @@ var require_proxy_addr = __commonJS(function(exports, module) {
       if (!isip(addr))
         return false;
       var ip = parseip(addr);
+      if (ip.kind() === "ipv6" && ip.isIPv4MappedAddress()) {
+        ip = ip.toIPv4Address();
+      }
       var kind = ip.kind();
       if (kind !== subnetkind) {
         if (subnetisipv4 && !ip.isIPv4MappedAddress()) {
           return false;
         }
+        if (!subnetisipv4 && !(subnetrange >= 96 && subnetip.isIPv4MappedAddress())) {
+          return false;
+        }
         ip = subnetisipv4 ? ip.toIPv4Address() : ip.toIPv4MappedAddress();
+      } else if (kind === "ipv6" && subnetip.isIPv4MappedAddress()) {
+        return false;
       }
       return ip.match(subnetip, subnetrange);
     };
@@ -42594,25 +42535,26 @@ var require_schemes = __commonJS(function(exports, module) {
           mailtoComponent.error = mailtoComponent.error || "URI mailto has malformed header fields.";
           continue;
         }
-        const name = eqIdx === -1 ? token : token.slice(0, eqIdx);
+        const name = decodeHex(eqIdx === -1 ? token : token.slice(0, eqIdx));
+        const normalizedName = name.toLowerCase();
         const value = eqIdx === -1 ? "" : token.slice(eqIdx + 1);
-        if (name === "to") {
+        if (normalizedName === "to") {
           const addrs = value.split(",");
           for (let j = 0;j < addrs.length; j++)
             to.push(addrs[j]);
           continue;
         }
-        if (name === "subject") {
+        if (normalizedName === "subject") {
           mailtoComponent.subject = decodeHex(value);
           continue;
         }
-        if (name === "body") {
+        if (normalizedName === "body") {
           mailtoComponent.body = decodeHex(value);
           continue;
         }
         if (headers === null)
           headers = Object.create(null);
-        headers[decodeHex(name)] = decodeHex(value);
+        headers[name] = decodeHex(value);
       }
       if (headers !== null)
         mailtoComponent.headers = headers;
@@ -42634,9 +42576,31 @@ var require_schemes = __commonJS(function(exports, module) {
       mailtoComponent.to = to;
     return mailtoComponent;
   }
+  function mailtoEncodeHeaderName(name) {
+    return encodeWithAllow(name.replace(/%/gu, "%25"), HFNAME);
+  }
   function mailtoSerialize(component, options) {
     const mailtoComponent = component;
     const to = Array.isArray(mailtoComponent.to) ? mailtoComponent.to.slice() : [];
+    const sourceHeaders = mailtoComponent.headers && typeof mailtoComponent.headers === "object" ? Object.assign(Object.create(null), mailtoComponent.headers) : Object.create(null);
+    const headers = Object.create(null);
+    for (const name in sourceHeaders) {
+      const normalizedName = name.toLowerCase();
+      if (normalizedName === "to") {
+        const addrs = String(sourceHeaders[name]).split(",");
+        for (let i = 0;i < addrs.length; i++)
+          to.push(addrs[i]);
+      } else if (normalizedName === "subject" || normalizedName === "body") {
+        headers[normalizedName] = sourceHeaders[name];
+      } else {
+        headers[name] = sourceHeaders[name];
+      }
+    }
+    if (mailtoComponent.subject !== undefined)
+      headers.subject = mailtoComponent.subject;
+    if (mailtoComponent.body !== undefined)
+      headers.body = mailtoComponent.body;
+    mailtoComponent.headers = headers;
     if (to.length) {
       for (let i = 0;i < to.length; i++) {
         const addr = String(to[i]);
@@ -42653,18 +42617,12 @@ var require_schemes = __commonJS(function(exports, module) {
     } else {
       mailtoComponent.path = undefined;
     }
-    const headers = mailtoComponent.headers && typeof mailtoComponent.headers === "object" ? Object.assign(Object.create(null), mailtoComponent.headers) : Object.create(null);
-    if (mailtoComponent.subject)
-      headers.subject = mailtoComponent.subject;
-    if (mailtoComponent.body)
-      headers.body = mailtoComponent.body;
-    mailtoComponent.headers = headers;
     let query = "";
     let count = 0;
     for (const name in headers) {
       if (count++ !== 0)
         query += "&";
-      query += encodeWithAllow(name, HFNAME) + "=" + encodeWithAllow(String(headers[name]), HFNAME);
+      query += mailtoEncodeHeaderName(name) + "=" + encodeWithAllow(String(headers[name]), HFNAME);
     }
     if (count !== 0) {
       mailtoComponent.query = query;
@@ -43038,13 +42996,14 @@ var require_fast_uri = __commonJS(function(exports, module) {
       if (!malformedIPLiteral) {
         malformedHost = canonicalizeHost(parsed, options, schemeHandler, isIP);
       }
-      if (!schemeHandler || schemeHandler && !schemeHandler.skipNormalize) {
-        if (uri.indexOf("%") !== -1) {
-          if (parsed.host !== undefined && !malformedIPLiteral) {
-            const host = isIP ? parsed.host : normalizePercentEncoding(parsed.host, true);
-            parsed.host = reescapeHostDelimiters(host, isIP);
-          }
+      if (uri.indexOf("%") !== -1 && parsed.host !== undefined && !malformedIPLiteral) {
+        let host = isIP ? parsed.host : normalizePercentEncoding(parsed.host, true);
+        if (!isIP) {
+          host = normalizePercentEncoding(host.toLowerCase());
         }
+        parsed.host = reescapeHostDelimiters(host, isIP);
+      }
+      if (!schemeHandler || schemeHandler && !schemeHandler.skipNormalize) {
         if (parsed.path) {
           parsed.path = normalizePathEncoding(parsed.path);
         }
@@ -52802,9 +52761,9 @@ var init_config = __esm(() => {
   SDK_METADATA = {
     language: "typescript",
     openapiDocVersion: "0.1.0",
-    sdkVersion: "0.0.57",
+    sdkVersion: "0.0.59",
     genVersion: "2.915.1",
-    userAgent: "speakeasy-sdk/mcp-typescript 0.0.57 2.915.1 0.1.0 @dalmia/calibrate-mcp"
+    userAgent: "speakeasy-sdk/mcp-typescript 0.0.59 2.915.1 0.1.0 @dalmia/calibrate-mcp"
   };
 });
 
@@ -55654,6 +55613,7 @@ var init_benchmarkrequest = __esm(() => {
   init_zod();
   BenchmarkRequest$zodSchema = object({
     models: array(string2()).describe("Model names to benchmark"),
+    parallel_models: boolean2().default(true).describe("Whether to run the models at the same time. Set false to run them one after another"),
     test_uuids: array(string2()).nullable().optional().describe("A subset of the agent's linked tests to benchmark. Each ID must be linked to the agent. Omit to run all linked tests")
   });
 });
@@ -56027,10 +55987,12 @@ var init_modelresult = __esm(() => {
     message: string2().describe("Status or result message for this model"),
     model: string2().describe("Model name these results are for"),
     passed: int().nullable().optional().describe("Number of test cases that passed"),
+    stopped_early: boolean2().default(false).describe("Whether this model's run stopped before starting every test case, after too many failed in a row"),
     success: boolean2().nullable().optional().describe("Whether this model's run succeeded"),
     test_results: array(TestCaseResult$zodSchema).nullable().optional().describe("Results for each test case for this model"),
     total_tests: int().nullable().optional().describe("Total test cases for this model"),
-    total_tokens: record(string2(), any()).nullable().optional().describe("Aggregated token usage as `{mean, min, max, count}`")
+    total_tokens: record(string2(), any()).nullable().optional().describe("Aggregated token usage as `{mean, min, max, count}`"),
+    unanswered_tests: int().nullable().optional().describe("Number of test cases that produced no answer because the agent or the judge could not be reached, which makes the pass rate an unfair measure of the agent")
   });
 });
 
@@ -56076,7 +56038,7 @@ var init_benchmarkstatusresponse = __esm(() => {
   init_testrunevaluator();
   BenchmarkStatusResponse$zodSchema = object({
     aborted: boolean2().default(false).describe("Whether a user stopped this run before it finished. The results collected up to that point are kept, and test cases that never ran are counted neither as passed nor as failed"),
-    error: boolean2().default(false).describe("True if the run failed"),
+    error: string2().nullable().optional().describe("Why the run could not be carried out, when it failed before producing any result"),
     evaluators: array(TestRunEvaluator$zodSchema).nullable().optional().describe("The evaluators used in this run. Each verdict in `judge_results` links to one of these by `evaluator_uuid`"),
     is_public: boolean2().default(false).describe("Whether the run is shared publicly"),
     leaderboard_summary: array(record(string2(), any())).nullable().optional().describe("Leaderboard comparing the models, one row per model. Columns vary by benchmark: a `model` column plus pass/fail counts, latency, cost, and one score column per evaluator, keyed by evaluator name"),
@@ -56084,6 +56046,7 @@ var init_benchmarkstatusresponse = __esm(() => {
     name: string2().describe("Name of the run. A run nobody has renamed shows its number instead, such as `Run 1` for a test run or `Benchmark 1` for a benchmark"),
     share_token: string2().nullable().optional().describe("Token for building the public share URL"),
     status: TaskStatus$zodSchema,
+    stopped_early: boolean2().default(false).describe("Whether any model's run stopped before starting every test case, after too many failed in a row"),
     task_id: string2().describe("Benchmark run job ID"),
     test_uuids: array(string2()).nullable().optional().describe("IDs of the tests this benchmark executed, in run order")
   });
@@ -56245,7 +56208,7 @@ var init_testrunstatusresponse = __esm(() => {
   TestRunStatusResponse$zodSchema = object({
     aborted: boolean2().default(false).describe("Whether a user stopped this run before it finished. The results collected up to that point are kept, and test cases that never ran are counted neither as passed nor as failed"),
     cost: record(string2(), any()).nullable().optional().describe("Aggregated cost as `{mean, min, max, count}` (USD)"),
-    error: boolean2().default(false).describe("True if the run failed"),
+    error: string2().nullable().optional().describe("Why the run could not be carried out, when it failed before producing any result"),
     evaluator_summary: array(record(string2(), any())).nullable().optional().describe("Totals for each evaluator over the whole run, matching the shape a benchmark reports for each model. Only evaluators that returned a verdict appear"),
     evaluators: array(TestRunEvaluator$zodSchema).nullable().optional().describe("The evaluators used in this run. Each verdict in `judge_results` links to one of these by `evaluator_uuid`"),
     failed: int().nullable().optional().describe("Number of test cases that failed"),
@@ -56971,6 +56934,7 @@ var init_agenttestrunlistitem = __esm(() => {
     results: array(TestRunCaseSummary$zodSchema).nullable().optional().describe("Flat pass/fail summary for each test case (fetch the run detail for full results)"),
     share_token: string2().nullable().optional().describe("Token for building the public share URL"),
     status: TaskStatus$zodSchema,
+    stopped_early: boolean2().default(false).describe("Whether the run stopped before starting every test case, after too many failed in a row"),
     total_tests: int().nullable().optional().describe("Total number of test cases"),
     total_tokens: record(string2(), any()).nullable().optional().describe("Aggregated token usage as `{mean, min, max, count}`"),
     type: AgentTestRunListItemType$zodSchema.describe("What kind of run this is:\n- `llm-unit-test`: a single run of the agent's tests\n- `llm-benchmark`: a multi-model comparison"),
@@ -61599,7 +61563,7 @@ hits its trace limit.
 function createMCPServer(deps) {
   const server = new McpServer({
     name: "CalibrateMcp",
-    version: "0.0.57"
+    version: "0.0.59"
   });
   const getClient = deps.getSDK || (() => new CalibrateMcpCore({
     security: deps.security,
@@ -62888,7 +62852,7 @@ 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/v0.0.57/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/v0.0.59/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>
@@ -65772,7 +65736,7 @@ var routes = buildRouteMap({
 var app = buildApplication(routes, {
   name: "mcp",
   versionInfo: {
-    currentVersion: "0.0.57"
+    currentVersion: "0.0.59"
   }
 });
 run(app, process3.argv.slice(2), buildContext(process3));
@@ -65780,5 +65744,5 @@ export {
   app
 };
 
-//# debugId=4CA7B963874437F364756E2164756E21
+//# debugId=FE1CD44B360D062964756E2164756E21
 //# sourceMappingURL=mcp-server.js.map