@meetkai/mka1 0.55.25 → 0.55.26

package/src/funcs/llmAuditGetRequestTrace.ts

@@ -0,0 +1,181 @@
+/*
+ * Code generated by Speakeasy (https://speakeasy.com). DO NOT EDIT.
+ */
+
+import { SDKCore } from "../core.js";
+import { encodeSimple } from "../lib/encodings.js";
+import { matchStatusCode } from "../lib/http.js";
+import * as M from "../lib/matchers.js";
+import { compactMap } from "../lib/primitives.js";
+import { safeParse } from "../lib/schemas.js";
+import { RequestOptions } from "../lib/sdks.js";
+import { extractSecurity, resolveGlobalSecurity } from "../lib/security.js";
+import { pathToFunc } from "../lib/url.js";
+import {
+  ConnectionError,
+  InvalidRequestError,
+  RequestAbortedError,
+  RequestTimeoutError,
+  UnexpectedClientError,
+} from "../models/errors/httpclienterrors.js";
+import { ResponseValidationError } from "../models/errors/responsevalidationerror.js";
+import { SDKError } from "../models/errors/sdkerror.js";
+import { SDKValidationError } from "../models/errors/sdkvalidationerror.js";
+import * as operations from "../models/operations/index.js";
+import { APICall, APIPromise } from "../types/async.js";
+import { Result } from "../types/fp.js";
+
+/**
+ * Get a request trace
+ *
+ * @remarks
+ * Returns the end-to-end trace of one API request: its timing at the API gateway, every response it created (including agents it called), and each response's model invocations and tool calls with durations. Look it up by the `X-Request-ID` returned on every response, or by the API gateway's own request ID. Org owners and admins only, scoped to their own organization. Requests audited before request IDs were recorded show their gateway timing but no responses.
+ *
+ * If set, this operation will use {@link Security.bearerAuth} from the global security.
+ */
+export function llmAuditGetRequestTrace(
+  client: SDKCore,
+  request: operations.GetRequestTraceRequest,
+  options?: RequestOptions,
+): APIPromise<
+  Result<
+    operations.GetRequestTraceResponseBody,
+    | SDKError
+    | ResponseValidationError
+    | ConnectionError
+    | RequestAbortedError
+    | RequestTimeoutError
+    | InvalidRequestError
+    | UnexpectedClientError
+    | SDKValidationError
+  >
+> {
+  return new APIPromise($do(
+    client,
+    request,
+    options,
+  ));
+}
+
+async function $do(
+  client: SDKCore,
+  request: operations.GetRequestTraceRequest,
+  options?: RequestOptions,
+): Promise<
+  [
+    Result<
+      operations.GetRequestTraceResponseBody,
+      | SDKError
+      | ResponseValidationError
+      | ConnectionError
+      | RequestAbortedError
+      | RequestTimeoutError
+      | InvalidRequestError
+      | UnexpectedClientError
+      | SDKValidationError
+    >,
+    APICall,
+  ]
+> {
+  const parsed = safeParse(
+    request,
+    (value) => operations.GetRequestTraceRequest$outboundSchema.parse(value),
+    "Input validation failed",
+  );
+  if (!parsed.ok) {
+    return [parsed, { status: "invalid" }];
+  }
+  const payload = parsed.value;
+  const body = null;
+
+  const pathParams = {
+    request_id: encodeSimple("request_id", payload.request_id, {
+      explode: false,
+      charEncoding: "percent",
+    }),
+  };
+  const path = pathToFunc("/api/v1/llm/audit/requests/{request_id}/trace")(
+    pathParams,
+  );
+
+  const headers = new Headers(compactMap({
+    Accept: "application/json",
+  }));
+
+  const secConfig = await extractSecurity(client._options.bearerAuth);
+  const securityInput = secConfig == null ? {} : { bearerAuth: secConfig };
+  const requestSecurity = resolveGlobalSecurity(securityInput, [0]);
+
+  const context = {
+    options: client._options,
+    baseURL: options?.serverURL ?? client._baseURL ?? "",
+    operationID: "getRequestTrace",
+    oAuth2Scopes: null,
+
+    resolvedSecurity: requestSecurity,
+
+    securitySource: client._options.bearerAuth,
+    retryConfig: options?.retries
+      || client._options.retryConfig
+      || {
+        strategy: "backoff",
+        backoff: {
+          initialInterval: 500,
+          maxInterval: 5000,
+          exponent: 1.5,
+          maxElapsedTime: 15000,
+        },
+        retryConnectionErrors: true,
+      }
+      || { strategy: "none" },
+    retryCodes: options?.retryCodes || ["429", "500", "502", "503", "504"],
+  };
+
+  const requestRes = client._createRequest(context, {
+    security: requestSecurity,
+    method: "GET",
+    baseURL: options?.serverURL,
+    path: path,
+    headers: headers,
+    body: body,
+    userAgent: client._options.userAgent,
+    timeoutMs: options?.timeoutMs || client._options.timeoutMs || -1,
+  }, options);
+  if (!requestRes.ok) {
+    return [requestRes, { status: "invalid" }];
+  }
+  const req = requestRes.value;
+
+  const doResult = await client._do(req, {
+    context,
+    isErrorStatusCode: (statusCode: number) =>
+      matchStatusCode({ status: statusCode } as Response, ["4XX", "5XX"]),
+    retryConfig: context.retryConfig,
+    retryCodes: context.retryCodes,
+  });
+  if (!doResult.ok) {
+    return [doResult, { status: "request-error", request: req }];
+  }
+  const response = doResult.value;
+
+  const [result] = await M.match<
+    operations.GetRequestTraceResponseBody,
+    | SDKError
+    | ResponseValidationError
+    | ConnectionError
+    | RequestAbortedError
+    | RequestTimeoutError
+    | InvalidRequestError
+    | UnexpectedClientError
+    | SDKValidationError
+  >(
+    M.json(200, operations.GetRequestTraceResponseBody$inboundSchema),
+    M.fail("4XX"),
+    M.fail("5XX"),
+  )(response, req);
+  if (!result.ok) {
+    return [result, { status: "complete", request: req, response }];
+  }
+
+  return [result, { status: "complete", request: req, response }];
+}

package/src/lib/config.ts

@@ -65,7 +65,7 @@ export function serverURLFromOptions(options: SDKOptions): URL | null {
 export const SDK_METADATA = {
   language: "typescript",
   openapiDocVersion: "1.1.0",
-  sdkVersion: "0.55.25",
+  sdkVersion: "0.55.26",
   genVersion: "2.950.2",
-  userAgent: "speakeasy-sdk/typescript 0.55.25 2.950.2 1.1.0 @meetkai/mka1",
+  userAgent: "speakeasy-sdk/typescript 0.55.26 2.950.2 1.1.0 @meetkai/mka1",
 } as const;

package/src/mcp-server/mcp-server.ts

@@ -19,7 +19,7 @@ const routes = buildRouteMap({
 export const app = buildApplication(routes, {
   name: "mcp",
   versionInfo: {
-    currentVersion: "0.55.25",
+    currentVersion: "0.55.26",
   },
 });
 

package/src/mcp-server/server.ts

@@ -220,6 +220,7 @@ import { tool$guardrailsTestGuardrails } from "./tools/guardrailsTestGuardrails.
 import { tool$guardrailsUpdateClusterDefaultGuardrails } from "./tools/guardrailsUpdateClusterDefaultGuardrails.js";
 import { tool$guardrailsUpdateClusterDefaultGuardrailsOrg } from "./tools/guardrailsUpdateClusterDefaultGuardrailsOrg.js";
 import { tool$guardrailsUpdateGuardrails } from "./tools/guardrailsUpdateGuardrails.js";
+import { tool$llmAuditGetRequestTrace } from "./tools/llmAuditGetRequestTrace.js";
 import { tool$llmBatchesCancel } from "./tools/llmBatchesCancel.js";
 import { tool$llmBatchesCreate } from "./tools/llmBatchesCreate.js";
 import { tool$llmBatchesGet } from "./tools/llmBatchesGet.js";
@@ -515,7 +516,7 @@ export function createMCPServer(deps: {
 }) {
   const server = new McpServer({
     name: "SDK",
-    version: "0.55.25",
+    version: "0.55.26",
   });
 
   const client = new SDKCore({
@@ -713,6 +714,7 @@ export function createMCPServer(deps: {
   tool(tool$permissionsLlmGrant);
   tool(tool$permissionsLlmRevoke);
   tool(tool$permissionsLlmCheck);
+  tool(tool$llmAuditGetRequestTrace);
   tool(tool$llmBatchesCreate);
   tool(tool$llmBatchesList);
   tool(tool$llmBatchesGet);

package/src/mcp-server/tools/billingCreateOrganizationBillingPortal.ts

@@ -14,7 +14,8 @@ export const tool$billingCreateOrganizationBillingPortal: ToolDefinition<
   typeof args
 > = {
   name: "billing-create-organization-billing-portal",
-  description: ``,
+  description:
+    `Create a payment-provider portal session for the bound organization. API keys require write:billing and either a human owner/admin or delegated service account. The organization is resolved from the API key.`,
   args,
   tool: async (client, args, ctx) => {
     const [result, apiCall] = await billingCreateOrganizationBillingPortal(

package/src/mcp-server/tools/billingCreateOrganizationBillingSetup.ts

@@ -14,7 +14,8 @@ export const tool$billingCreateOrganizationBillingSetup: ToolDefinition<
   typeof args
 > = {
   name: "billing-create-organization-billing-setup",
-  description: ``,
+  description:
+    `Create a payment-provider setup session for the bound organization. API keys require write:billing and either a human owner/admin or delegated service account. The organization is resolved from the API key.`,
   args,
   tool: async (client, args, ctx) => {
     const [result, apiCall] = await billingCreateOrganizationBillingSetup(

package/src/mcp-server/tools/billingGetOrganizationBilling.ts

@@ -12,7 +12,8 @@ const args = {
 
 export const tool$billingGetOrganizationBilling: ToolDefinition<typeof args> = {
   name: "billing-get-organization-billing",
-  description: ``,
+  description:
+    `Read organization billing. API keys require read:billing and either a human owner/admin or delegated service account. Keys are bound to their organization; cross-organization reads require a human cluster administrator with read:cluster.`,
   args,
   tool: async (client, args, ctx) => {
     const [result, apiCall] = await billingGetOrganizationBilling(

package/src/mcp-server/tools/billingListOrganizationBillingInvoices.ts

@@ -14,7 +14,8 @@ export const tool$billingListOrganizationBillingInvoices: ToolDefinition<
   typeof args
 > = {
   name: "billing-list-organization-billing-invoices",
-  description: ``,
+  description:
+    `List invoices for an organization. API keys require read:billing and either a human owner/admin or delegated service account. Cross-organization reads require a human cluster administrator with read:cluster.`,
   args,
   tool: async (client, args, ctx) => {
     const [result, apiCall] = await billingListOrganizationBillingInvoices(

package/src/mcp-server/tools/billingSetOrganizationBillingEnabled.ts

@@ -14,7 +14,8 @@ export const tool$billingSetOrganizationBillingEnabled: ToolDefinition<
   typeof args
 > = {
   name: "billing-set-organization-billing-enabled",
-  description: ``,
+  description:
+    `Enable or disable organization billing. Requires a human cluster administrator. API keys require both write:billing and the bootstrap-only write:cluster scope. Service-account keys cannot enable or disable billing.`,
   args,
   tool: async (client, args, ctx) => {
     const [result, apiCall] = await billingSetOrganizationBillingEnabled(

package/src/mcp-server/tools/billingUpdateOrganizationBillingProfile.ts

@@ -14,7 +14,8 @@ export const tool$billingUpdateOrganizationBillingProfile: ToolDefinition<
   typeof args
 > = {
   name: "billing-update-organization-billing-profile",
-  description: ``,
+  description:
+    `Update the bound organization's billing profile. API keys require write:billing and either a human owner/admin or delegated service account. The organization is resolved from the API key.`,
   args,
   tool: async (client, args, ctx) => {
     const [result, apiCall] = await billingUpdateOrganizationBillingProfile(

package/src/mcp-server/tools/llmAuditGetRequestTrace.ts

@@ -0,0 +1,37 @@
+/*
+ * Code generated by Speakeasy (https://speakeasy.com). DO NOT EDIT.
+ */
+
+import { llmAuditGetRequestTrace } from "../../funcs/llmAuditGetRequestTrace.js";
+import * as operations from "../../models/operations/index.js";
+import { formatResult, ToolDefinition } from "../tools.js";
+
+const args = {
+  request: operations.GetRequestTraceRequest$inboundSchema,
+};
+
+export const tool$llmAuditGetRequestTrace: ToolDefinition<typeof args> = {
+  name: "llm-audit-get-request-trace",
+  description: `Get a request trace
+
+Returns the end-to-end trace of one API request: its timing at the API gateway, every response it created (including agents it called), and each response's model invocations and tool calls with durations. Look it up by the \`X-Request-ID\` returned on every response, or by the API gateway's own request ID. Org owners and admins only, scoped to their own organization. Requests audited before request IDs were recorded show their gateway timing but no responses.`,
+  args,
+  tool: async (client, args, ctx) => {
+    const [result, apiCall] = await llmAuditGetRequestTrace(
+      client,
+      args.request,
+      { fetchOptions: { signal: ctx.mcpReq.signal } },
+    ).$inspect();
+
+    if (!result.ok) {
+      return {
+        content: [{ type: "text", text: result.error.message }],
+        isError: true,
+      };
+    }
+
+    const value = result.value;
+
+    return formatResult(value, apiCall);
+  },
+};

package/src/models/components/createvectorstorerequest.ts

@@ -87,7 +87,7 @@ export type CreateVectorStoreRequest = {
    */
   embeddingModel?: string | undefined;
   /**
-   * The number of dimensions for the embedding vectors. Only supported for models with flexible dimensions. If not specified, uses the model's default dimensions.
+   * The number of dimensions for the embedding vectors. Must equal the model's size when it is fixed, or lie within its range (`embedding_dimension_range` on `GET /models/{model_id}`). If not specified, uses the model's default dimensions. The store stays searchable only while the model's declared dimensions still include this size and the model returns embeddings of this width.
    */
   embeddingDimensions?: number | undefined;
   /**

package/src/models/components/embeddingmodellimits.ts

@@ -13,15 +13,15 @@ import { SDKValidationError } from "../errors/sdkvalidationerror.js";
  */
 export type EmbeddingModelLimits = {
   /**
-   * Maximum number of inputs allowed in a single batch request
+   * Maximum number of inputs allowed in a single batch request: the model's batch size or the gateway-wide 10,000-input limit, whichever is lower, on `POST /embeddings`. Batch `/v1/embeddings` lines are held only to the model's batch size.
    */
   maxBatchSize: number;
   /**
-   * Maximum tokens per individual input (derived from contextWindow)
+   * Maximum tokens per individual input: the model's context window. Null when the registry records no context window: no token cap applies beyond what `max_input_length` implies. For a window long enough that its estimated character length exceeds the gateway-wide character limit, `max_input_length` binds first. Advisory on `/embeddings`, which accepts only strings and enforces `max_input_length`.
    */
-  maxInputTokens: number;
+  maxInputTokens: number | null;
   /**
-   * Maximum characters per individual input (contextWindow * 2)
+   * Maximum characters per individual input: the cap derived from the model's context window or the gateway-wide 100,000-character limit, whichever is lower, on `POST /embeddings`. Batch `/v1/embeddings` lines are held only to the model's cap.
    */
   maxInputLength: number;
 };
@@ -32,9 +32,9 @@ export const EmbeddingModelLimits$inboundSchema: z.ZodType<
   z.ZodTypeDef,
   unknown
 > = z.object({
-  max_batch_size: z.number(),
-  max_input_tokens: z.number(),
-  max_input_length: z.number(),
+  max_batch_size: z.number().int(),
+  max_input_tokens: z.nullable(z.number().int()),
+  max_input_length: z.number().int(),
 }).transform((v) => {
   return remap$(v, {
     "max_batch_size": "maxBatchSize",
@@ -45,7 +45,7 @@ export const EmbeddingModelLimits$inboundSchema: z.ZodType<
 /** @internal */
 export type EmbeddingModelLimits$Outbound = {
   max_batch_size: number;
-  max_input_tokens: number;
+  max_input_tokens: number | null;
   max_input_length: number;
 };
 
@@ -55,9 +55,9 @@ export const EmbeddingModelLimits$outboundSchema: z.ZodType<
   z.ZodTypeDef,
   EmbeddingModelLimits
 > = z.object({
-  maxBatchSize: z.number(),
-  maxInputTokens: z.number(),
-  maxInputLength: z.number(),
+  maxBatchSize: z.number().int(),
+  maxInputTokens: z.nullable(z.number().int()),
+  maxInputLength: z.number().int(),
 }).transform((v) => {
   return remap$(v, {
     maxBatchSize: "max_batch_size",

package/src/models/components/embeddingsrequest.ts

@@ -39,7 +39,7 @@ export type EmbeddingsRequest = {
    */
   model: string;
   /**
-   * The number of dimensions the resulting output embeddings should have. Only supported in certain models.
+   * The number of dimensions the resulting output embeddings should have. Ignored for a model with a fixed size, which always returns that size. For a model with a range it must lie within the range (`embedding_dimension_range` on `GET /models/{model_id}`), or the request fails with 400.
    */
   dimensions?: number | undefined;
   /**

package/src/models/components/vectorstore.ts

@@ -88,7 +88,7 @@ export type VectorStore = {
    */
   embeddingModel: string | null;
   /**
-   * The number of dimensions for the embedding vectors in this vector store. Null for legacy vector stores.
+   * The number of dimensions for the embedding vectors in this vector store. Null for legacy vector stores. If the embedding model's declared dimensions stop including this size, or the model returns embeddings of another width, search fails with 412 and new files fail to ingest.
    */
   embeddingDimensions: number | null;
   /**

package/src/models/operations/getmodel.ts

@@ -160,7 +160,7 @@ export type GetModelResponseBody = {
    */
   outputModalities: Array<string>;
   /**
-   * Context window in tokens, mirroring the registry's `contextWindow`. For an LLM it is the input plus output tokens one request can span: the Responses API trims the oldest conversation items to fit it, reserving nothing for the output, unless the request sets `truncation` to `disabled`, which rejects an oversized request instead; a separate request-size limit in bytes applies the same way. Chat completions forwards requests unchanged. For an embeddings model `/embeddings` derives its per-input character cap from it, which `GET /embeddings/models` reports as `max_input_length`; a gateway-wide per-input ceiling also applies. On other model types the gateway does not read it, so a recorded value is informational only and not enforced. Null when the registry does not record one.
+   * Context window in tokens, mirroring the registry's `contextWindow`. For an LLM it is the input plus output tokens one request can span: the Responses API trims the oldest conversation items to fit it, reserving nothing for the output, unless the request sets `truncation` to `disabled`, which rejects an oversized request instead; a separate request-size limit in bytes applies the same way. Chat completions forwards requests unchanged. For an embeddings model `/embeddings` derives a per-input character cap from it, alongside a gateway-wide per-input ceiling; `GET /embeddings/models` reports whichever binds first as `max_input_length`. On other model types the gateway does not read it, so a recorded value is informational only and not enforced. Null when the registry does not record one.
    */
   contextWindow: number | null;
   /**
@@ -172,7 +172,7 @@ export type GetModelResponseBody = {
    */
   minImageResolution: MinImageResolution | null;
   /**
-   * Output vector sizes declared for this model, mirroring the registry's `embeddingLimits.dimensions`, which is normally recorded only on embeddings models. A model with a single fixed size reports it as `min`, `max` and `default` alike, so when `min` equals `max` the size is fixed and a request should omit `dimensions`. Vector stores reject a size outside the range. `/embeddings` does not check it: it drops a requested `dimensions` for a fixed-size model and otherwise forwards it to the upstream. Null when the registry does not record dimensions; a vector store using such a model must then be created with an explicit size (`embedding_dimensions` on `POST /vector_stores`). Distinct from the vector store's own `embedding_dimensions`, which is a single size.
+   * Output vector sizes declared for this model, mirroring the registry's `embeddingLimits.dimensions`, which is normally recorded only on embeddings models. When `min` equals `max` the size is fixed: `/embeddings` ignores a requested `dimensions` and returns that size, and vector stores accept only that size. Otherwise a requested size must lie within the range: `/embeddings` and vector stores reject one outside it with 400. Null when the registry does not record dimensions; a vector store using such a model must then be created with an explicit size (`embedding_dimensions` on `POST /vector_stores`). Distinct from the vector store's own `embedding_dimensions`, which is a single size.
    */
   embeddingDimensionRange: EmbeddingDimensionRange | null;
   /**