telnyx 6.64.0 → 6.65.0

package/src/resources/ai/assistants/versions.ts

@@ -174,7 +174,7 @@ export interface UpdateAssistant {
    * `/ai/integrations/connections`. Each item references a catalog integration by
    * `integration_id`.
    */
-  integrations?: Array<AssistantsAPI.AssistantIntegration>;
+  integrations?: Array<UpdateAssistant.Integration>;
 
   /**
    * Settings for interruptions and how the assistant decides the user has finished
@@ -184,7 +184,7 @@ export interface UpdateAssistant {
    * `transcription.settings` (`eot_threshold`, `eot_timeout_ms`,
    * `eager_eot_threshold`).
    */
-  interruption_settings?: AssistantsAPI.InferenceEmbeddingInterruptionSettings;
+  interruption_settings?: UpdateAssistant.InterruptionSettings;
 
   /**
    * This is only needed when using third-party inference providers selected by
@@ -200,7 +200,7 @@ export interface UpdateAssistant {
    * MCP servers attached to the assistant. Create MCP servers with
    * `/ai/mcp_servers`, then reference them by `id` here.
    */
-  mcp_servers?: Array<AssistantsAPI.AssistantMcpServer>;
+  mcp_servers?: Array<UpdateAssistant.McpServer>;
 
   messaging_settings?: AssistantsAPI.MessagingSettings;
 
@@ -265,6 +265,125 @@ export interface UpdateAssistant {
   widget_settings?: AssistantsAPI.WidgetSettings;
 }
 
+export namespace UpdateAssistant {
+  /**
+   * Reference to a connected integration attached to an assistant. Discover
+   * available integrations with `/ai/integrations` and connected integrations with
+   * `/ai/integrations/connections`.
+   */
+  export interface Integration {
+    /**
+     * Catalog integration ID to attach. This is the `id` from the integrations catalog
+     * at `/ai/integrations` (the same value also appears as `integration_id` on
+     * entries returned by `/ai/integrations/connections`). It is **not** the
+     * connection-level `id` from `/ai/integrations/connections`.
+     */
+    integration_id: string;
+
+    /**
+     * Optional per-assistant allowlist of integration tool names. When omitted or
+     * empty, all tools allowed by the connected integration are available to the
+     * assistant.
+     */
+    allowed_list?: Array<string>;
+  }
+
+  /**
+   * Settings for interruptions and how the assistant decides the user has finished
+   * speaking. These timings are most relevant when using non turn-taking
+   * transcription models. For turn-taking models like `deepgram/flux`, end-of-turn
+   * behavior is controlled by the transcription end-of-turn settings under
+   * `transcription.settings` (`eot_threshold`, `eot_timeout_ms`,
+   * `eager_eot_threshold`).
+   */
+  export interface InterruptionSettings {
+    /**
+     * When true, disables user interruptions while the assistant greeting is playing.
+     */
+    disable_greeting_interruption?: boolean;
+
+    /**
+     * Whether users can interrupt the assistant while it is speaking.
+     */
+    enable?: boolean;
+
+    /**
+     * Controls when the assistant starts speaking after the user stops. These
+     * thresholds primarily apply to non turn-taking transcription models. For
+     * turn-taking models like `deepgram/flux`, end-of-turn detection is driven by the
+     * transcription end-of-turn settings under `transcription.settings` instead.
+     */
+    start_speaking_plan?: InterruptionSettings.StartSpeakingPlan;
+  }
+
+  export namespace InterruptionSettings {
+    /**
+     * Controls when the assistant starts speaking after the user stops. These
+     * thresholds primarily apply to non turn-taking transcription models. For
+     * turn-taking models like `deepgram/flux`, end-of-turn detection is driven by the
+     * transcription end-of-turn settings under `transcription.settings` instead.
+     */
+    export interface StartSpeakingPlan {
+      /**
+       * Endpointing thresholds used to decide when the user has finished speaking.
+       * Applies to non turn-taking transcription models. For `deepgram/flux`, use
+       * `transcription.settings.eot_threshold` / `eot_timeout_ms` /
+       * `eager_eot_threshold`.
+       */
+      transcription_endpointing_plan?: StartSpeakingPlan.TranscriptionEndpointingPlan;
+
+      /**
+       * Minimum seconds to wait before the assistant starts speaking.
+       */
+      wait_seconds?: number;
+    }
+
+    export namespace StartSpeakingPlan {
+      /**
+       * Endpointing thresholds used to decide when the user has finished speaking.
+       * Applies to non turn-taking transcription models. For `deepgram/flux`, use
+       * `transcription.settings.eot_threshold` / `eot_timeout_ms` /
+       * `eager_eot_threshold`.
+       */
+      export interface TranscriptionEndpointingPlan {
+        /**
+         * Seconds to wait after the transcript ends without punctuation.
+         */
+        on_no_punctuation_seconds?: number;
+
+        /**
+         * Seconds to wait after the transcript ends with a number.
+         */
+        on_number_seconds?: number;
+
+        /**
+         * Seconds to wait after the transcript ends with punctuation.
+         */
+        on_punctuation_seconds?: number;
+      }
+    }
+  }
+
+  /**
+   * Reference to an MCP server attached to an assistant. Create and manage MCP
+   * servers with the `/ai/mcp_servers` endpoints, then attach them to assistants by
+   * ID.
+   */
+  export interface McpServer {
+    /**
+     * ID of the MCP server to attach. This must be the `id` of an MCP server returned
+     * by the `/ai/mcp_servers` endpoints.
+     */
+    id: string;
+
+    /**
+     * Optional per-assistant allowlist of MCP tool names. When omitted, the assistant
+     * uses the MCP server's configured `allowed_tools`.
+     */
+    allowed_tools?: Array<string>;
+  }
+}
+
 export interface VersionRetrieveParams {
   /**
    * Path param
@@ -355,7 +474,7 @@ export interface VersionUpdateParams {
    * integrations are at `/ai/integrations/connections`. Each item references a
    * catalog integration by `integration_id`.
    */
-  integrations?: Array<AssistantsAPI.AssistantIntegration>;
+  integrations?: Array<VersionUpdateParams.Integration>;
 
   /**
    * Body param: Settings for interruptions and how the assistant decides the user
@@ -365,7 +484,7 @@ export interface VersionUpdateParams {
    * under `transcription.settings` (`eot_threshold`, `eot_timeout_ms`,
    * `eager_eot_threshold`).
    */
-  interruption_settings?: AssistantsAPI.InferenceEmbeddingInterruptionSettings;
+  interruption_settings?: VersionUpdateParams.InterruptionSettings;
 
   /**
    * Body param: This is only needed when using third-party inference providers
@@ -381,7 +500,7 @@ export interface VersionUpdateParams {
    * Body param: MCP servers attached to the assistant. Create MCP servers with
    * `/ai/mcp_servers`, then reference them by `id` here.
    */
-  mcp_servers?: Array<AssistantsAPI.AssistantMcpServer>;
+  mcp_servers?: Array<VersionUpdateParams.McpServer>;
 
   /**
    * Body param
@@ -468,6 +587,125 @@ export interface VersionUpdateParams {
   widget_settings?: AssistantsAPI.WidgetSettings;
 }
 
+export namespace VersionUpdateParams {
+  /**
+   * Reference to a connected integration attached to an assistant. Discover
+   * available integrations with `/ai/integrations` and connected integrations with
+   * `/ai/integrations/connections`.
+   */
+  export interface Integration {
+    /**
+     * Catalog integration ID to attach. This is the `id` from the integrations catalog
+     * at `/ai/integrations` (the same value also appears as `integration_id` on
+     * entries returned by `/ai/integrations/connections`). It is **not** the
+     * connection-level `id` from `/ai/integrations/connections`.
+     */
+    integration_id: string;
+
+    /**
+     * Optional per-assistant allowlist of integration tool names. When omitted or
+     * empty, all tools allowed by the connected integration are available to the
+     * assistant.
+     */
+    allowed_list?: Array<string>;
+  }
+
+  /**
+   * Settings for interruptions and how the assistant decides the user has finished
+   * speaking. These timings are most relevant when using non turn-taking
+   * transcription models. For turn-taking models like `deepgram/flux`, end-of-turn
+   * behavior is controlled by the transcription end-of-turn settings under
+   * `transcription.settings` (`eot_threshold`, `eot_timeout_ms`,
+   * `eager_eot_threshold`).
+   */
+  export interface InterruptionSettings {
+    /**
+     * When true, disables user interruptions while the assistant greeting is playing.
+     */
+    disable_greeting_interruption?: boolean;
+
+    /**
+     * Whether users can interrupt the assistant while it is speaking.
+     */
+    enable?: boolean;
+
+    /**
+     * Controls when the assistant starts speaking after the user stops. These
+     * thresholds primarily apply to non turn-taking transcription models. For
+     * turn-taking models like `deepgram/flux`, end-of-turn detection is driven by the
+     * transcription end-of-turn settings under `transcription.settings` instead.
+     */
+    start_speaking_plan?: InterruptionSettings.StartSpeakingPlan;
+  }
+
+  export namespace InterruptionSettings {
+    /**
+     * Controls when the assistant starts speaking after the user stops. These
+     * thresholds primarily apply to non turn-taking transcription models. For
+     * turn-taking models like `deepgram/flux`, end-of-turn detection is driven by the
+     * transcription end-of-turn settings under `transcription.settings` instead.
+     */
+    export interface StartSpeakingPlan {
+      /**
+       * Endpointing thresholds used to decide when the user has finished speaking.
+       * Applies to non turn-taking transcription models. For `deepgram/flux`, use
+       * `transcription.settings.eot_threshold` / `eot_timeout_ms` /
+       * `eager_eot_threshold`.
+       */
+      transcription_endpointing_plan?: StartSpeakingPlan.TranscriptionEndpointingPlan;
+
+      /**
+       * Minimum seconds to wait before the assistant starts speaking.
+       */
+      wait_seconds?: number;
+    }
+
+    export namespace StartSpeakingPlan {
+      /**
+       * Endpointing thresholds used to decide when the user has finished speaking.
+       * Applies to non turn-taking transcription models. For `deepgram/flux`, use
+       * `transcription.settings.eot_threshold` / `eot_timeout_ms` /
+       * `eager_eot_threshold`.
+       */
+      export interface TranscriptionEndpointingPlan {
+        /**
+         * Seconds to wait after the transcript ends without punctuation.
+         */
+        on_no_punctuation_seconds?: number;
+
+        /**
+         * Seconds to wait after the transcript ends with a number.
+         */
+        on_number_seconds?: number;
+
+        /**
+         * Seconds to wait after the transcript ends with punctuation.
+         */
+        on_punctuation_seconds?: number;
+      }
+    }
+  }
+
+  /**
+   * Reference to an MCP server attached to an assistant. Create and manage MCP
+   * servers with the `/ai/mcp_servers` endpoints, then attach them to assistants by
+   * ID.
+   */
+  export interface McpServer {
+    /**
+     * ID of the MCP server to attach. This must be the `id` of an MCP server returned
+     * by the `/ai/mcp_servers` endpoints.
+     */
+    id: string;
+
+    /**
+     * Optional per-assistant allowlist of MCP tool names. When omitted, the assistant
+     * uses the MCP server's configured `allowed_tools`.
+     */
+    allowed_tools?: Array<string>;
+  }
+}
+
 export interface VersionDeleteParams {
   assistant_id: string;
 }

package/src/resources/ai/index.ts

@@ -2,7 +2,6 @@
 
 export {
   AI,
-  type ModelMetadata,
   type AICreateResponseResponse,
   type AIRetrieveModelsResponse,
   type AISummarizeResponse,
@@ -12,8 +11,6 @@ export {
 export {
   Assistants,
   type Assistant,
-  type AssistantIntegration,
-  type AssistantMcpServer,
   type AssistantTool,
   type AssistantsList,
   type AudioVisualizerConfig,
@@ -26,7 +23,6 @@ export {
   type HangupToolParams,
   type ImportMetadata,
   type InferenceEmbedding,
-  type InferenceEmbeddingInterruptionSettings,
   type InferenceEmbeddingWebhookToolParams,
   type InsightSettings,
   type MessagingSettings,
@@ -36,9 +32,7 @@ export {
   type PostConversationSettingsReq,
   type PrivacySettings,
   type RetrievalTool,
-  type StartSpeakingPlan,
   type TelephonySettings,
-  type TranscriptionEndpointingPlan,
   type TranscriptionSettings,
   type TranscriptionSettingsConfig,
   type TransferTool,

package/src/resources/ai/openai/openai.ts

@@ -1,7 +1,6 @@
 // File generated from our OpenAPI spec by Stainless. See CONTRIBUTING.md for details.
 
 import { APIResource } from '../../../core/resource';
-import * as AIAPI from '../ai';
 import * as ChatAPI from './chat';
 import { Chat, ChatCreateCompletionParams, ChatCreateCompletionResponse } from './chat';
 import * as EmbeddingsAPI from './embeddings';
@@ -102,11 +101,141 @@ export class OpenAI extends APIResource {
 export type OpenAICreateResponseResponse = { [key: string]: unknown };
 
 export interface OpenAIListModelsResponse {
-  data: Array<AIAPI.ModelMetadata>;
+  data: Array<OpenAIListModelsResponse.Data>;
 
   object?: string;
 }
 
+export namespace OpenAIListModelsResponse {
+  /**
+   * Metadata for a model available on Telnyx Inference. Returned by
+   * `GET /v2/ai/openai/models` (and the deprecated `GET /v2/ai/models`). Open-source
+   * models live under their Hugging Face organization (e.g. `moonshotai/Kimi-K2.6`,
+   * `zai-org/GLM-5.1-FP8`, `MiniMaxAI/MiniMax-M2.7`); fine-tuned models are owned by
+   * the Telnyx organization that trained them.
+   */
+  export interface Data {
+    /**
+     * Model identifier. For open-source models, follows the
+     * `{organization}/{model_name}` convention from Hugging Face (e.g.
+     * `moonshotai/Kimi-K2.6`).
+     */
+    id: string;
+
+    /**
+     * Maximum total tokens (prompt + completion) supported by the model in a single
+     * request.
+     */
+    context_length: number;
+
+    /**
+     * Timestamp at which the model was registered on Telnyx Inference (ISO 8601).
+     */
+    created: string;
+
+    /**
+     * ISO language codes the model supports (e.g. `en`, `es`).
+     */
+    languages: Array<string>;
+
+    /**
+     * License the model is distributed under, e.g. `Apache 2.0`, `MIT`,
+     * `Llama 3 Community License`.
+     */
+    license: string;
+
+    /**
+     * Organization that originally published the model, matching the prefix of `id`
+     * for open-source models.
+     */
+    organization: string;
+
+    /**
+     * Owner of the model. `Telnyx` for Telnyx-hosted open-source models, the upstream
+     * provider name for proxied models, or the Telnyx organization id for fine-tuned
+     * models.
+     */
+    owned_by: string;
+
+    /**
+     * Total parameter count of the model.
+     */
+    parameters: number;
+
+    /**
+     * Billing tier the model belongs to. Used together with `pricing` to determine
+     * cost per 1M tokens.
+     */
+    tier: 'small' | 'medium' | 'large' | 'unlisted';
+
+    /**
+     * Base model the fine-tuned model was trained from. Only set for fine-tuned
+     * models.
+     */
+    base_model?: string | null;
+
+    /**
+     * Short, human-readable summary of what the model is best suited for.
+     */
+    description?: string | null;
+
+    /**
+     * Whether the model can be used as a base for a fine-tuning job via
+     * `POST /v2/ai/fine_tuning/jobs`.
+     */
+    is_fine_tunable?: boolean;
+
+    /**
+     * Whether the model accepts image inputs in chat completions (multimodal vision
+     * support).
+     */
+    is_vision_supported?: boolean;
+
+    /**
+     * Maximum number of completion (output) tokens the model will generate per
+     * request. `null` if unconstrained beyond `context_length`.
+     */
+    max_completion_tokens?: number | null;
+
+    /**
+     * Object type. Always `model`.
+     */
+    object?: string;
+
+    /**
+     * Human-readable parameter count, e.g. `1.0T`, `753.9B`, `8B`.
+     */
+    parameters_str?: string | null;
+
+    /**
+     * Mapping of token kind to price, as strings to preserve precision. Typical keys
+     * are `prompt`, `cached_prompt`, and `completion`. When pricing is available the
+     * block also includes `currency` (ISO 4217 code matching the account's configured
+     * billing currency) and `unit` (the denomination the prices are quoted in,
+     * currently always `1M_tokens` for token-priced models).
+     */
+    pricing?: { [key: string]: string };
+
+    /**
+     * Whether Telnyx currently recommends this model as the LLM powering a Telnyx AI
+     * Assistant.
+     */
+    recommended_for_assistants?: boolean;
+
+    /**
+     * Public region names where the model is currently deployed (e.g. `us-central-1`,
+     * `eu-central-1`).
+     */
+    regions?: Array<string>;
+
+    /**
+     * Primary task the model is intended for, e.g. `text-generation`,
+     * `audio-text-to-text`, `feature-extraction` (embeddings).
+     */
+    task?: string;
+  }
+}
+
 export interface OpenAICreateResponseParams {
   /**
    * Optional Telnyx Conversation ID from `POST /ai/conversations`. When provided,