@meetkai/mka1 0.55.24 → 0.55.25

package/src/models/components/createagentrequest.ts

@@ -5,6 +5,7 @@
 import * as z from "zod/v3";
 import { remap as remap$ } from "../../lib/primitives.js";
 import { safeParse } from "../../lib/schemas.js";
+import { ClosedEnum } from "../../types/enums.js";
 import { Result as SafeParseResult } from "../../types/fp.js";
 import { SDKValidationError } from "../errors/sdkvalidationerror.js";
 import {
@@ -26,6 +27,20 @@ import {
   ToolDefinition$outboundSchema,
 } from "./tooldefinition.js";
 
+/**
+ * What a run beyond max_concurrent_runs gets. reject (the default) refuses it with 429 and error code agent_concurrency_limit. queue stores it as waiting and starts it, oldest first, when a run finishes, with the agent configuration current at that time.
+ */
+export const CreateAgentRequestConcurrencyOverflow = {
+  Reject: "reject",
+  Queue: "queue",
+} as const;
+/**
+ * What a run beyond max_concurrent_runs gets. reject (the default) refuses it with 429 and error code agent_concurrency_limit. queue stores it as waiting and starts it, oldest first, when a run finishes, with the agent configuration current at that time.
+ */
+export type CreateAgentRequestConcurrencyOverflow = ClosedEnum<
+  typeof CreateAgentRequestConcurrencyOverflow
+>;
+
 export type CreateAgentRequest = {
   name: string;
   description?: string | null | undefined;
@@ -42,11 +57,29 @@ export type CreateAgentRequest = {
    * Ordered chain of model IDs the gateway tries when a model call fails on the model before it; forwarded unchanged to the Responses API, which validates it. Empty (the default) means the run keeps the gateway's own fallback rules for the model.
    */
   fallbackModels?: Array<string> | undefined;
+  /**
+   * Most runs of this agent that may execute at once. Null (the default) means no limit. Not part of the versioned configuration: a rollback keeps the current limit, and changing only this field or concurrency_overflow creates no new version.
+   */
+  maxConcurrentRuns?: number | null | undefined;
+  /**
+   * What a run beyond max_concurrent_runs gets. reject (the default) refuses it with 429 and error code agent_concurrency_limit. queue stores it as waiting and starts it, oldest first, when a run finishes, with the agent configuration current at that time.
+   */
+  concurrencyOverflow?: CreateAgentRequestConcurrencyOverflow | undefined;
   text?: AgentTextConfiguration | null | undefined;
   reasoning?: AgentReasoningConfiguration | null | undefined;
   metadata?: { [k: string]: string } | undefined;
 };
 
+/** @internal */
+export const CreateAgentRequestConcurrencyOverflow$inboundSchema:
+  z.ZodNativeEnum<typeof CreateAgentRequestConcurrencyOverflow> = z.nativeEnum(
+    CreateAgentRequestConcurrencyOverflow,
+  );
+/** @internal */
+export const CreateAgentRequestConcurrencyOverflow$outboundSchema:
+  z.ZodNativeEnum<typeof CreateAgentRequestConcurrencyOverflow> =
+    CreateAgentRequestConcurrencyOverflow$inboundSchema;
+
 /** @internal */
 export const CreateAgentRequest$inboundSchema: z.ZodType<
   CreateAgentRequest,
@@ -62,6 +95,9 @@ export const CreateAgentRequest$inboundSchema: z.ZodType<
   parallel_tool_calls: z.boolean().default(true),
   max_tool_calls: z.nullable(z.number().int()).optional(),
   fallback_models: z.array(z.string()).optional(),
+  max_concurrent_runs: z.nullable(z.number().int()).optional(),
+  concurrency_overflow: CreateAgentRequestConcurrencyOverflow$inboundSchema
+    .optional(),
   text: z.nullable(AgentTextConfiguration$inboundSchema).optional(),
   reasoning: z.nullable(AgentReasoningConfiguration$inboundSchema).optional(),
   metadata: z.record(z.string()).optional(),
@@ -71,6 +107,8 @@ export const CreateAgentRequest$inboundSchema: z.ZodType<
     "parallel_tool_calls": "parallelToolCalls",
     "max_tool_calls": "maxToolCalls",
     "fallback_models": "fallbackModels",
+    "max_concurrent_runs": "maxConcurrentRuns",
+    "concurrency_overflow": "concurrencyOverflow",
   });
 });
 /** @internal */
@@ -84,6 +122,8 @@ export type CreateAgentRequest$Outbound = {
   parallel_tool_calls: boolean;
   max_tool_calls?: number | null | undefined;
   fallback_models?: Array<string> | undefined;
+  max_concurrent_runs?: number | null | undefined;
+  concurrency_overflow?: string | undefined;
   text?: AgentTextConfiguration$Outbound | null | undefined;
   reasoning?: AgentReasoningConfiguration$Outbound | null | undefined;
   metadata?: { [k: string]: string } | undefined;
@@ -104,6 +144,9 @@ export const CreateAgentRequest$outboundSchema: z.ZodType<
   parallelToolCalls: z.boolean().default(true),
   maxToolCalls: z.nullable(z.number().int()).optional(),
   fallbackModels: z.array(z.string()).optional(),
+  maxConcurrentRuns: z.nullable(z.number().int()).optional(),
+  concurrencyOverflow: CreateAgentRequestConcurrencyOverflow$outboundSchema
+    .optional(),
   text: z.nullable(AgentTextConfiguration$outboundSchema).optional(),
   reasoning: z.nullable(AgentReasoningConfiguration$outboundSchema).optional(),
   metadata: z.record(z.string()).optional(),
@@ -113,6 +156,8 @@ export const CreateAgentRequest$outboundSchema: z.ZodType<
     parallelToolCalls: "parallel_tool_calls",
     maxToolCalls: "max_tool_calls",
     fallbackModels: "fallback_models",
+    maxConcurrentRuns: "max_concurrent_runs",
+    concurrencyOverflow: "concurrency_overflow",
   });
 });
 

package/src/models/components/updateagentrequest.ts

@@ -5,6 +5,7 @@
 import * as z from "zod/v3";
 import { remap as remap$ } from "../../lib/primitives.js";
 import { safeParse } from "../../lib/schemas.js";
+import { ClosedEnum } from "../../types/enums.js";
 import { Result as SafeParseResult } from "../../types/fp.js";
 import { SDKValidationError } from "../errors/sdkvalidationerror.js";
 import {
@@ -26,6 +27,20 @@ import {
   ToolDefinition$outboundSchema,
 } from "./tooldefinition.js";
 
+/**
+ * What a run beyond max_concurrent_runs gets. reject (the default) refuses it with 429 and error code agent_concurrency_limit. queue stores it as waiting and starts it, oldest first, when a run finishes, with the agent configuration current at that time.
+ */
+export const UpdateAgentRequestConcurrencyOverflow = {
+  Reject: "reject",
+  Queue: "queue",
+} as const;
+/**
+ * What a run beyond max_concurrent_runs gets. reject (the default) refuses it with 429 and error code agent_concurrency_limit. queue stores it as waiting and starts it, oldest first, when a run finishes, with the agent configuration current at that time.
+ */
+export type UpdateAgentRequestConcurrencyOverflow = ClosedEnum<
+  typeof UpdateAgentRequestConcurrencyOverflow
+>;
+
 /**
  * Updates only provided fields. Omitted fields preserve their saved values; empty updates are rejected.
  */
@@ -45,11 +60,29 @@ export type UpdateAgentRequest = {
    * Ordered chain of model IDs the gateway tries when a model call fails on the model before it; forwarded unchanged to the Responses API, which validates it. Empty (the default) means the run keeps the gateway's own fallback rules for the model.
    */
   fallbackModels?: Array<string> | undefined;
+  /**
+   * Most runs of this agent that may execute at once. Null (the default) means no limit. Not part of the versioned configuration: a rollback keeps the current limit, and changing only this field or concurrency_overflow creates no new version.
+   */
+  maxConcurrentRuns?: number | null | undefined;
+  /**
+   * What a run beyond max_concurrent_runs gets. reject (the default) refuses it with 429 and error code agent_concurrency_limit. queue stores it as waiting and starts it, oldest first, when a run finishes, with the agent configuration current at that time.
+   */
+  concurrencyOverflow?: UpdateAgentRequestConcurrencyOverflow | undefined;
   text?: AgentTextConfiguration | null | undefined;
   reasoning?: AgentReasoningConfiguration | null | undefined;
   metadata?: { [k: string]: string } | undefined;
 };
 
+/** @internal */
+export const UpdateAgentRequestConcurrencyOverflow$inboundSchema:
+  z.ZodNativeEnum<typeof UpdateAgentRequestConcurrencyOverflow> = z.nativeEnum(
+    UpdateAgentRequestConcurrencyOverflow,
+  );
+/** @internal */
+export const UpdateAgentRequestConcurrencyOverflow$outboundSchema:
+  z.ZodNativeEnum<typeof UpdateAgentRequestConcurrencyOverflow> =
+    UpdateAgentRequestConcurrencyOverflow$inboundSchema;
+
 /** @internal */
 export const UpdateAgentRequest$inboundSchema: z.ZodType<
   UpdateAgentRequest,
@@ -65,6 +98,9 @@ export const UpdateAgentRequest$inboundSchema: z.ZodType<
   parallel_tool_calls: z.boolean().optional(),
   max_tool_calls: z.nullable(z.number().int()).optional(),
   fallback_models: z.array(z.string()).optional(),
+  max_concurrent_runs: z.nullable(z.number().int()).optional(),
+  concurrency_overflow: UpdateAgentRequestConcurrencyOverflow$inboundSchema
+    .optional(),
   text: z.nullable(AgentTextConfiguration$inboundSchema).optional(),
   reasoning: z.nullable(AgentReasoningConfiguration$inboundSchema).optional(),
   metadata: z.record(z.string()).optional(),
@@ -74,6 +110,8 @@ export const UpdateAgentRequest$inboundSchema: z.ZodType<
     "parallel_tool_calls": "parallelToolCalls",
     "max_tool_calls": "maxToolCalls",
     "fallback_models": "fallbackModels",
+    "max_concurrent_runs": "maxConcurrentRuns",
+    "concurrency_overflow": "concurrencyOverflow",
   });
 });
 /** @internal */
@@ -87,6 +125,8 @@ export type UpdateAgentRequest$Outbound = {
   parallel_tool_calls?: boolean | undefined;
   max_tool_calls?: number | null | undefined;
   fallback_models?: Array<string> | undefined;
+  max_concurrent_runs?: number | null | undefined;
+  concurrency_overflow?: string | undefined;
   text?: AgentTextConfiguration$Outbound | null | undefined;
   reasoning?: AgentReasoningConfiguration$Outbound | null | undefined;
   metadata?: { [k: string]: string } | undefined;
@@ -107,6 +147,9 @@ export const UpdateAgentRequest$outboundSchema: z.ZodType<
   parallelToolCalls: z.boolean().optional(),
   maxToolCalls: z.nullable(z.number().int()).optional(),
   fallbackModels: z.array(z.string()).optional(),
+  maxConcurrentRuns: z.nullable(z.number().int()).optional(),
+  concurrencyOverflow: UpdateAgentRequestConcurrencyOverflow$outboundSchema
+    .optional(),
   text: z.nullable(AgentTextConfiguration$outboundSchema).optional(),
   reasoning: z.nullable(AgentReasoningConfiguration$outboundSchema).optional(),
   metadata: z.record(z.string()).optional(),
@@ -116,6 +159,8 @@ export const UpdateAgentRequest$outboundSchema: z.ZodType<
     parallelToolCalls: "parallel_tool_calls",
     maxToolCalls: "max_tool_calls",
     fallbackModels: "fallback_models",
+    maxConcurrentRuns: "max_concurrent_runs",
+    concurrencyOverflow: "concurrency_overflow",
   });
 });
 

package/src/react-query/agentRunsCancelAgentRun.ts

@@ -50,7 +50,7 @@ export type AgentRunsCancelAgentRunMutationError =
  * Cancel a saved-agent run
  *
  * @remarks
- * Cancels a queued run locally before dispatch, or forwards cancellation to its linked gateway response. Locally failed linked runs with an active or missing gateway snapshot are checked upstream before cancellation; genuine terminal work is never relabelled cancelled. Fresh incomplete gateway snapshots are presented as incomplete on subsequent reads while preserving the connector error and completion timestamp, even if cancel returns 409. Repeated cancellation is idempotent. Returns 409 if submission has no linked response yet or a genuine terminal outcome already won. Uses the run's organization, team and delegated-user mutation scope; operator read access does not grant cancellation access. API-key callers also need write:responses for linked cancellation and read:responses to inspect locally failed runs or reconcile a completion race.
+ * Cancels a waiting or queued run locally before dispatch, or forwards cancellation to its linked gateway response. Locally failed linked runs with an active or missing gateway snapshot are checked upstream before cancellation; genuine terminal work is never relabelled cancelled. Fresh incomplete gateway snapshots are presented as incomplete on subsequent reads while preserving the connector error and completion timestamp, even if cancel returns 409. Repeated cancellation is idempotent. Returns 409 if submission has no linked response yet or a genuine terminal outcome already won. Uses the run's organization, team and delegated-user mutation scope; operator read access does not grant cancellation access. API-key callers also need write:responses for linked cancellation and read:responses to inspect locally failed runs or reconcile a completion race.
  */
 export function useAgentRunsCancelAgentRunMutation(
   options?: MutationHookOptions<

package/src/react-query/agentRunsCreateAgentRun.ts

@@ -50,7 +50,7 @@ export type AgentRunsCreateAgentRunMutationError =
  * Start a saved agent run
  *
  * @remarks
- * Starts an asynchronous background run for a saved agent and returns the queued run record. Use the run retrieval endpoint to poll status or the events endpoint to stream progress.
+ * Starts an asynchronous background run for a saved agent and returns the queued run record. Use the run retrieval endpoint to poll status or the events endpoint to stream progress. When the agent is already running max_concurrent_runs runs, concurrency_overflow decides: reject returns 429 with error code agent_concurrency_limit and creates no run; queue returns 202 with a waiting run that starts when a slot frees.
  */
 export function useAgentRunsCreateAgentRunMutation(
   options?: MutationHookOptions<

package/src/sdk/agentruns.ts

@@ -33,7 +33,7 @@ export class AgentRuns extends ClientSDK {
    * Start a saved agent run
    *
    * @remarks
-   * Starts an asynchronous background run for a saved agent and returns the queued run record. Use the run retrieval endpoint to poll status or the events endpoint to stream progress.
+   * Starts an asynchronous background run for a saved agent and returns the queued run record. Use the run retrieval endpoint to poll status or the events endpoint to stream progress. When the agent is already running max_concurrent_runs runs, concurrency_overflow decides: reject returns 429 with error code agent_concurrency_limit and creates no run; queue returns 202 with a waiting run that starts when a slot frees.
    */
   async createAgentRun(
     request: operations.CreateAgentRunRequest,
@@ -67,7 +67,7 @@ export class AgentRuns extends ClientSDK {
    * Cancel a saved-agent run
    *
    * @remarks
-   * Cancels a queued run locally before dispatch, or forwards cancellation to its linked gateway response. Locally failed linked runs with an active or missing gateway snapshot are checked upstream before cancellation; genuine terminal work is never relabelled cancelled. Fresh incomplete gateway snapshots are presented as incomplete on subsequent reads while preserving the connector error and completion timestamp, even if cancel returns 409. Repeated cancellation is idempotent. Returns 409 if submission has no linked response yet or a genuine terminal outcome already won. Uses the run's organization, team and delegated-user mutation scope; operator read access does not grant cancellation access. API-key callers also need write:responses for linked cancellation and read:responses to inspect locally failed runs or reconcile a completion race.
+   * Cancels a waiting or queued run locally before dispatch, or forwards cancellation to its linked gateway response. Locally failed linked runs with an active or missing gateway snapshot are checked upstream before cancellation; genuine terminal work is never relabelled cancelled. Fresh incomplete gateway snapshots are presented as incomplete on subsequent reads while preserving the connector error and completion timestamp, even if cancel returns 409. Repeated cancellation is idempotent. Returns 409 if submission has no linked response yet or a genuine terminal outcome already won. Uses the run's organization, team and delegated-user mutation scope; operator read access does not grant cancellation access. API-key callers also need write:responses for linked cancellation and read:responses to inspect locally failed runs or reconcile a completion race.
    */
   async cancelAgentRun(
     request: operations.CancelAgentRunRequest,