@pushwoosh/rpc-gateway-ai-assistant 0.1.197 → 0.1.199

package/data.js

@@ -2039,6 +2039,108 @@ export const data = {
             }
         }
     },
+    "pushwoosh.accounts.assistant.v1.CreateRealtimeSessionRequest": {
+        "type": "I",
+        "fields": {
+            "application": {
+                "type": "string"
+            },
+            "provider": {
+                "type": "string"
+            },
+            "chatId": {
+                "type": "number"
+            }
+        }
+    },
+    "pushwoosh.accounts.assistant.v1.AppendVoiceTranscriptRequest": {
+        "type": "I",
+        "fields": {
+            "application": {
+                "type": "string"
+            },
+            "chatId": {
+                "type": "number"
+            },
+            "role": {
+                "type": "pushwoosh.accounts.assistant.v1.Role"
+            },
+            "text": {
+                "type": "string"
+            },
+            "provider": {
+                "type": "string"
+            },
+            "mattermostRootId": {
+                "type": "string"
+            }
+        }
+    },
+    "pushwoosh.accounts.assistant.v1.AppendVoiceTranscriptResponse": {
+        "type": "I",
+        "fields": {
+            "mattermostRootId": {
+                "type": "string"
+            }
+        }
+    },
+    "pushwoosh.accounts.assistant.v1.CallVoiceToolRequest": {
+        "type": "I",
+        "fields": {
+            "application": {
+                "type": "string"
+            },
+            "name": {
+                "type": "string"
+            },
+            "arguments": {
+                "type": "string"
+            }
+        }
+    },
+    "pushwoosh.accounts.assistant.v1.CallVoiceToolResponse": {
+        "type": "I",
+        "fields": {
+            "result": {
+                "type": "string"
+            }
+        }
+    },
+    "pushwoosh.accounts.assistant.v1.VoiceTool": {
+        "type": "I",
+        "fields": {
+            "name": {
+                "type": "string"
+            },
+            "description": {
+                "type": "string"
+            },
+            "parametersJsonSchema": {
+                "type": "string"
+            }
+        }
+    },
+    "pushwoosh.accounts.assistant.v1.CreateRealtimeSessionResponse": {
+        "type": "I",
+        "fields": {
+            "clientSecret": {
+                "type": "string"
+            },
+            "model": {
+                "type": "string"
+            },
+            "expiresAt": {
+                "type": "Date"
+            },
+            "provider": {
+                "type": "string"
+            },
+            "tools": {
+                "type": "pushwoosh.accounts.assistant.v1.VoiceTool",
+                "array": true
+            }
+        }
+    },
     "pushwoosh.accounts.assistant.v1.AssistantService": {
         "type": "S",
         "key": "assistant",
@@ -2142,6 +2244,24 @@ export const data = {
                 "response": "pushwoosh.accounts.assistant.v1.CompleteOnboardingHandoffResponse",
                 "method": "post",
                 "path": "/api/v1/applications/{application}/chats/{chat_id}/onboarding/handoff"
+            },
+            "CreateRealtimeSession": {
+                "request": "pushwoosh.accounts.assistant.v1.CreateRealtimeSessionRequest",
+                "response": "pushwoosh.accounts.assistant.v1.CreateRealtimeSessionResponse",
+                "method": "post",
+                "path": "/api/v1/applications/{application}/realtime/session"
+            },
+            "CallVoiceTool": {
+                "request": "pushwoosh.accounts.assistant.v1.CallVoiceToolRequest",
+                "response": "pushwoosh.accounts.assistant.v1.CallVoiceToolResponse",
+                "method": "post",
+                "path": "/api/v1/applications/{application}/realtime/tool-call"
+            },
+            "AppendVoiceTranscript": {
+                "request": "pushwoosh.accounts.assistant.v1.AppendVoiceTranscriptRequest",
+                "response": "pushwoosh.accounts.assistant.v1.AppendVoiceTranscriptResponse",
+                "method": "post",
+                "path": "/api/v1/applications/{application}/realtime/transcript"
             }
         }
     },

package/package.json

@@ -1,6 +1,6 @@
 {
     "name": "@pushwoosh/rpc-gateway-ai-assistant",
-    "version": "0.1.197",
+    "version": "0.1.199",
     "description": "AI Assistant api gateway HTTP API Types and Data",
     "main": "index.js",
     "module": "index.js",

package/pushwoosh_accounts_assistant_v1.d.ts

@@ -251,6 +251,109 @@ export type CompleteOnboardingHandoffRequest = {
 export type CompleteOnboardingHandoffResponse = {
     chat: Chat;
 };
+/**
+ * CreateRealtimeSessionRequest mints an ephemeral realtime voice secret
+ * (OpenAI Realtime over WebRTC or Gemini Live over WebSocket). The session
+ * configuration — instructions, memory, chat history, tools — is fixed
+ * server-side; the frontend only receives the short-lived secret.
+ */
+export type CreateRealtimeSessionRequest = {
+    application: string;
+    /**
+     * provider selects the realtime backend: "gemini" (default) or "openai".
+     * Unknown values are rejected; a provider not enabled in config returns
+     * Unimplemented.
+     */
+    provider: string;
+    /**
+     * chat_id, when non-zero, injects the recent history of this chat into the
+     * voice instructions so the call continues the text conversation instead of
+     * starting cold. The chat must belong to the caller.
+     */
+    chatId: number;
+};
+/**
+ * AppendVoiceTranscriptRequest persists one completed voice turn into the
+ * chat and mirrors it to Mattermost with a voice marker. The browser sends a
+ * turn as soon as its transcription completes, so a dropped call loses at
+ * most the in-flight turn.
+ */
+export type AppendVoiceTranscriptRequest = {
+    application: string;
+    /** chat the voice call is attached to */
+    chatId: number;
+    /** ROLE_USER or ROLE_ASSISTANT */
+    role: Role;
+    /** final transcript of the turn */
+    text: string;
+    /** "openai" | "gemini" — for the Mattermost voice marker */
+    provider: string;
+    /**
+     * Mattermost thread root of the current voice exchange: pass the value
+     * returned by the previous user turn so assistant replies thread under it.
+     */
+    mattermostRootId: string;
+};
+/**
+ * AppendVoiceTranscriptResponse returns the Mattermost thread root: user turns
+ * create it, assistant turns echo the request value back.
+ */
+export type AppendVoiceTranscriptResponse = {
+    mattermostRootId: string;
+};
+/**
+ * CallVoiceToolRequest relays one tool call of a Gemini voice session. The
+ * Live API has no remote-MCP support, so the browser forwards the model's
+ * toolCall messages here; the allowlist is re-checked server-side and the call
+ * is proxied to the owning upstream MCP server.
+ */
+export type CallVoiceToolRequest = {
+    application: string;
+    /** tool name exactly as declared to the model */
+    name: string;
+    /** JSON-encoded tool arguments from the model */
+    arguments: string;
+};
+/**
+ * CallVoiceToolResponse carries the raw MCP tool result as JSON; the browser
+ * passes it back to the model verbatim.
+ */
+export type CallVoiceToolResponse = {
+    /** JSON-encoded MCP toolCallResult */
+    result: string;
+};
+/**
+ * VoiceTool is a tool declaration for a voice session. Declarations are
+ * built server-side from the realtime allowlist; the browser registers them
+ * with the provider SDK and relays execution through CallVoiceTool.
+ */
+export type VoiceTool = {
+    name: string;
+    description: string;
+    /** JSON-encoded input schema */
+    parametersJsonSchema: string;
+};
+/**
+ * CreateRealtimeSessionResponse carries the ephemeral secret. The secret is
+ * single-session and expires within minutes — the frontend must request a
+ * fresh one for every call and never persist it.
+ */
+export type CreateRealtimeSessionResponse = {
+    /** OpenAI: "ek_..." / Gemini: "auth_tokens/..." */
+    clientSecret: string;
+    /** realtime model the session is pinned to */
+    model: string;
+    /** secret expiration */
+    expiresAt: Date;
+    /** provider the secret belongs to */
+    provider: string;
+    /**
+     * Tool declarations the browser must register with the provider SDK
+     * (execution relays through CallVoiceTool). Empty for Gemini: its
+     * declarations are pinned inside the ephemeral token's constraints.
+     */
+    tools: VoiceTool[];
+};
 export type AssistantService_CreateChat = RpcMethod<CreateChatRequest, CreateChatResponse>;
 export type AssistantService_GetChat = RpcMethod<GetChatRequest, GetChatResponse>;
 export type AssistantService_GetAllChats = RpcMethod<GetAllChatsRequest, GetAllChatsResponse>;
@@ -263,6 +366,9 @@ export type AssistantService_CompactifyContext = RpcMethod<CompactifyContextRequ
 export type AssistantService_CheckUsageAllowed = RpcMethod<CheckUsageAllowedRequest, CheckUsageAllowedResponse>;
 export type AssistantService_GetOnboardingStatus = RpcMethod<GetOnboardingStatusRequest, GetOnboardingStatusResponse>;
 export type AssistantService_CompleteOnboardingHandoff = RpcMethod<CompleteOnboardingHandoffRequest, CompleteOnboardingHandoffResponse>;
+export type AssistantService_CreateRealtimeSession = RpcMethod<CreateRealtimeSessionRequest, CreateRealtimeSessionResponse>;
+export type AssistantService_CallVoiceTool = RpcMethod<CallVoiceToolRequest, CallVoiceToolResponse>;
+export type AssistantService_AppendVoiceTranscript = RpcMethod<AppendVoiceTranscriptRequest, AppendVoiceTranscriptResponse>;
 export interface AssistantService {
     /** Create a new chat */
     CreateChat: AssistantService_CreateChat;
@@ -315,4 +421,22 @@ export interface AssistantService {
      * on chat reload).
      */
     CompleteOnboardingHandoff: AssistantService_CompleteOnboardingHandoff;
+    /**
+     * CreateRealtimeSession mints an ephemeral voice-session secret for the
+     * configured provider (OpenAI or Gemini). Auth, usage quota and session
+     * configuration are enforced here; provider API keys never reach the browser.
+     */
+    CreateRealtimeSession: AssistantService_CreateRealtimeSession;
+    /**
+     * CallVoiceTool relays a Gemini voice session's tool call to the MCP proxy
+     * upstreams, enforcing the realtime allowlist. Used only by the voice
+     * widget; OpenAI sessions call the MCP proxy directly instead.
+     */
+    CallVoiceTool: AssistantService_CallVoiceTool;
+    /**
+     * AppendVoiceTranscript persists one completed voice turn into the chat and
+     * mirrors it to Mattermost (marked as voice). Called by the voice widget as
+     * turns finish, so voice conversations are logged like text ones.
+     */
+    AppendVoiceTranscript: AssistantService_AppendVoiceTranscript;
 }