@pushwoosh/rpc-gateway-ai-assistant 0.1.582 → 0.1.584

package/commonTypes.d.ts

@@ -7,12 +7,16 @@ export type JsonValue = string | number | boolean | null | JsonValue[] | {
 export type JsonObject = {
     [key: string]: JsonValue;
 };
-export type RpcMethod<Req, Res> = ((request: Req) => Promise<Res>) & {
-    safe(request: Req): Promise<[{
+/** Per-call options accepted by every method. */
+export type RpcCallOptions = {
+    signal?: AbortSignal;
+};
+export type RpcMethod<Req, Res> = ((request: Req, options?: RpcCallOptions) => Promise<Res>) & {
+    safe(request: Req, options?: RpcCallOptions): Promise<[{
         error: unknown;
     }, null] | [null, Res]>;
 };
-export type RpcMethodV3<Req, Res, Err> = ((request: Req) => Promise<Res>) & {
-    safe(request: Req): Promise<[Err, null] | [null, Res]>;
+export type RpcMethodV3<Req, Res, Err> = ((request: Req, options?: RpcCallOptions) => Promise<Res>) & {
+    safe(request: Req, options?: RpcCallOptions): Promise<[Err, null] | [null, Res]>;
     parseError(e: unknown): Err;
 };

package/data.js

@@ -6,359 +6,6 @@ export const data = {
         "key": "agentRunner",
         "methods": {}
     },
-    "pushwoosh.agentrunner.v1.TurnRequest": {
-        "type": "I",
-        "fields": {
-            "start": {
-                "type": "pushwoosh.agentrunner.v1.StartTurn"
-            },
-            "toolResult": {
-                "type": "pushwoosh.agentrunner.v1.ToolResult"
-            }
-        },
-        "oneofs": {
-            "payload": {
-                "oneof": [
-                    "start",
-                    "toolResult"
-                ]
-            }
-        }
-    },
-    "pushwoosh.agentrunner.v1.StartTurn": {
-        "type": "I",
-        "fields": {
-            "systemPrompt": {
-                "type": "pushwoosh.agentrunner.v1.SystemPrompt"
-            },
-            "history": {
-                "type": "pushwoosh.agentrunner.v1.HistoryMessage",
-                "array": true
-            },
-            "tools": {
-                "type": "pushwoosh.agentrunner.v1.ToolDef",
-                "array": true
-            },
-            "model": {
-                "type": "string"
-            },
-            "maxTurns": {
-                "type": "number"
-            },
-            "chatId": {
-                "type": "number"
-            },
-            "session": {
-                "type": "pushwoosh.agentrunner.v1.SdkSession"
-            }
-        }
-    },
-    "pushwoosh.agentrunner.v1.SdkSession": {
-        "type": "I",
-        "fields": {
-            "sessionId": {
-                "type": "string"
-            },
-            "lines": {
-                "type": "string",
-                "array": true
-            }
-        }
-    },
-    "pushwoosh.agentrunner.v1.SystemPrompt": {
-        "type": "I",
-        "fields": {
-            "static": {
-                "type": "string"
-            },
-            "dynamic": {
-                "type": "string",
-                "array": true
-            }
-        }
-    },
-    "pushwoosh.agentrunner.v1.HistoryRole": {
-        "type": "E",
-        "values": [
-            "HISTORY_ROLE_UNSPECIFIED",
-            "HISTORY_ROLE_USER",
-            "HISTORY_ROLE_ASSISTANT",
-            "HISTORY_ROLE_TOOL_REQUEST",
-            "HISTORY_ROLE_TOOL_RESULT",
-            "HISTORY_ROLE_CONTEXT_COMPACT"
-        ]
-    },
-    "pushwoosh.agentrunner.v1.HistoryMessage": {
-        "type": "I",
-        "fields": {
-            "role": {
-                "type": "pushwoosh.agentrunner.v1.HistoryRole"
-            },
-            "text": {
-                "type": "string"
-            },
-            "toolCalls": {
-                "type": "pushwoosh.agentrunner.v1.ToolCall",
-                "array": true
-            },
-            "toolResult": {
-                "type": "pushwoosh.agentrunner.v1.ToolResult"
-            }
-        }
-    },
-    "pushwoosh.agentrunner.v1.ToolDef": {
-        "type": "I",
-        "fields": {
-            "name": {
-                "type": "string"
-            },
-            "description": {
-                "type": "string"
-            },
-            "inputSchemaJson": {
-                "type": "string"
-            },
-            "deferred": {
-                "type": "boolean"
-            }
-        }
-    },
-    "pushwoosh.agentrunner.v1.ToolCall": {
-        "type": "I",
-        "fields": {
-            "id": {
-                "type": "string"
-            },
-            "name": {
-                "type": "string"
-            },
-            "argsJson": {
-                "type": "string"
-            }
-        }
-    },
-    "pushwoosh.agentrunner.v1.ToolResult": {
-        "type": "I",
-        "fields": {
-            "id": {
-                "type": "string"
-            },
-            "content": {
-                "type": "string"
-            },
-            "isError": {
-                "type": "boolean"
-            }
-        }
-    },
-    "pushwoosh.agentrunner.v1.TurnEvent": {
-        "type": "I",
-        "fields": {
-            "delta": {
-                "type": "pushwoosh.agentrunner.v1.StreamDelta"
-            },
-            "assistantMessage": {
-                "type": "pushwoosh.agentrunner.v1.AssistantMessage"
-            },
-            "toolCall": {
-                "type": "pushwoosh.agentrunner.v1.ToolCall"
-            },
-            "done": {
-                "type": "pushwoosh.agentrunner.v1.TurnDone"
-            },
-            "sessionAppend": {
-                "type": "pushwoosh.agentrunner.v1.SessionAppend"
-            },
-            "rateLimit": {
-                "type": "pushwoosh.agentrunner.v1.RateLimit"
-            },
-            "messageStarted": {
-                "type": "pushwoosh.agentrunner.v1.Usage"
-            }
-        },
-        "oneofs": {
-            "payload": {
-                "oneof": [
-                    "delta",
-                    "assistantMessage",
-                    "toolCall",
-                    "done",
-                    "sessionAppend",
-                    "rateLimit",
-                    "messageStarted"
-                ]
-            }
-        }
-    },
-    "pushwoosh.agentrunner.v1.StreamEvent": {
-        "type": "E",
-        "values": [
-            "STREAM_EVENT_UNSPECIFIED",
-            "STREAM_EVENT_MESSAGE_START",
-            "STREAM_EVENT_CONTENT_BLOCK_START",
-            "STREAM_EVENT_TEXT_DELTA",
-            "STREAM_EVENT_TOOL_USE_START",
-            "STREAM_EVENT_TOOL_USE_DELTA",
-            "STREAM_EVENT_TOOL_USE_RESULT",
-            "STREAM_EVENT_CONTENT_BLOCK_STOP",
-            "STREAM_EVENT_MESSAGE_DELTA",
-            "STREAM_EVENT_MESSAGE_STOP",
-            "STREAM_EVENT_THINKING_DELTA",
-            "STREAM_EVENT_ERROR"
-        ]
-    },
-    "pushwoosh.agentrunner.v1.StreamDelta": {
-        "type": "I",
-        "fields": {
-            "type": {
-                "type": "pushwoosh.agentrunner.v1.StreamEvent"
-            },
-            "content": {
-                "type": "string"
-            }
-        }
-    },
-    "pushwoosh.agentrunner.v1.AssistantMessage": {
-        "type": "I",
-        "fields": {
-            "text": {
-                "type": "string"
-            },
-            "toolCalls": {
-                "type": "pushwoosh.agentrunner.v1.ToolCall",
-                "array": true
-            },
-            "usage": {
-                "type": "pushwoosh.agentrunner.v1.Usage"
-            },
-            "thinking": {
-                "type": "string"
-            },
-            "messageId": {
-                "type": "string"
-            }
-        }
-    },
-    "pushwoosh.agentrunner.v1.SessionAppend": {
-        "type": "I",
-        "fields": {
-            "sessionId": {
-                "type": "string"
-            },
-            "lines": {
-                "type": "string",
-                "array": true
-            }
-        }
-    },
-    "pushwoosh.agentrunner.v1.RateLimit": {
-        "type": "I",
-        "fields": {
-            "status": {
-                "type": "string"
-            },
-            "window": {
-                "type": "string"
-            },
-            "utilization": {
-                "type": "number"
-            },
-            "resetsAt": {
-                "type": "number"
-            },
-            "overageStatus": {
-                "type": "string"
-            },
-            "usingOverage": {
-                "type": "boolean"
-            },
-            "overageDisabledReason": {
-                "type": "string"
-            }
-        }
-    },
-    "pushwoosh.agentrunner.v1.GetQuotaRequest": {
-        "type": "I",
-        "fields": {}
-    },
-    "pushwoosh.agentrunner.v1.Quota": {
-        "type": "I",
-        "fields": {
-            "windows": {
-                "type": "pushwoosh.agentrunner.v1.QuotaWindow",
-                "array": true
-            },
-            "overageStatus": {
-                "type": "string"
-            },
-            "overageDisabledReason": {
-                "type": "string"
-            },
-            "checkedAt": {
-                "type": "number"
-            },
-            "error": {
-                "type": "string"
-            }
-        }
-    },
-    "pushwoosh.agentrunner.v1.QuotaWindow": {
-        "type": "I",
-        "fields": {
-            "window": {
-                "type": "string"
-            },
-            "utilizationPercent": {
-                "type": "number"
-            },
-            "resetsAt": {
-                "type": "number"
-            },
-            "status": {
-                "type": "string"
-            }
-        }
-    },
-    "pushwoosh.agentrunner.v1.Usage": {
-        "type": "I",
-        "fields": {
-            "model": {
-                "type": "string"
-            },
-            "inputTokens": {
-                "type": "number"
-            },
-            "outputTokens": {
-                "type": "number"
-            },
-            "cacheCreationInputTokens": {
-                "type": "number"
-            },
-            "cacheReadInputTokens": {
-                "type": "number"
-            }
-        }
-    },
-    "pushwoosh.agentrunner.v1.TurnDone": {
-        "type": "I",
-        "fields": {
-            "stopReason": {
-                "type": "string"
-            },
-            "error": {
-                "type": "string"
-            },
-            "turns": {
-                "type": "number"
-            },
-            "turnCapHit": {
-                "type": "boolean"
-            },
-            "usage": {
-                "type": "pushwoosh.agentrunner.v1.Usage"
-            }
-        }
-    },
     "pushwoosh.aibuilder.v1.DocumentBlockType": {
         "type": "E",
         "values": [

package/package.json

@@ -1,6 +1,6 @@
 {
     "name": "@pushwoosh/rpc-gateway-ai-assistant",
-    "version": "0.1.582",
+    "version": "0.1.584",
     "description": "AI Assistant api gateway HTTP API Types and Data",
     "main": "index.js",
     "module": "index.js",

package/pushwoosh_accounts_assistant_v1.d.ts

@@ -1,4 +1,4 @@
-import { RpcMethod } from './commonTypes';
+import { RpcMethod, RpcCallOptions } from './commonTypes';
 /** Role represents the sender of a message */
 export type Role = 'ROLE_UNSPECIFIED' | 'ROLE_USER' | 'ROLE_ASSISTANT'
 /** Assistant message with tool calls (renamed for DB limit) */
@@ -534,19 +534,11 @@ export interface AssistantService {
      * Follow a turn still answering; ends with the turn's own status. NotFound: its events expired,
      * Aborted: its pod died, FailedPrecondition: too long to follow live — reload the chat.
      */
-    AttachTurn(request: AttachTurnRequest, options?: {
-        signal?: AbortSignal;
-    }): Promise<AsyncIterable<InferenceResponse>>;
+    AttachTurn(request: AttachTurnRequest, options?: RpcCallOptions): Promise<AsyncIterable<InferenceResponse>>;
     /** Stream AI inference responses */
-    StreamInference(request: InferenceRequest, options?: {
-        signal?: AbortSignal;
-    }): Promise<AsyncIterable<InferenceResponse>>;
-    StreamInferenceGemini(request: InferenceRequest, options?: {
-        signal?: AbortSignal;
-    }): Promise<AsyncIterable<InferenceResponse>>;
-    StreamInferenceChatGPT(request: InferenceRequest, options?: {
-        signal?: AbortSignal;
-    }): Promise<AsyncIterable<InferenceResponse>>;
+    StreamInference(request: InferenceRequest, options?: RpcCallOptions): Promise<AsyncIterable<InferenceResponse>>;
+    StreamInferenceGemini(request: InferenceRequest, options?: RpcCallOptions): Promise<AsyncIterable<InferenceResponse>>;
+    StreamInferenceChatGPT(request: InferenceRequest, options?: RpcCallOptions): Promise<AsyncIterable<InferenceResponse>>;
     /**
      * StreamInferenceLLM is the self-hosted LLM (Qwen) inference endpoint. It is
      * one of two providers that can run the onboarding agent — the frontend
@@ -554,24 +546,14 @@ export interface AssistantService {
      * (Claude) by choosing the endpoint. Every turn, including the opening one,
      * carries a real user message; an empty message is rejected. See is_onboarding.
      */
-    StreamInferenceLLM(request: InferenceRequest, options?: {
-        signal?: AbortSignal;
-    }): Promise<AsyncIterable<InferenceResponse>>;
+    StreamInferenceLLM(request: InferenceRequest, options?: RpcCallOptions): Promise<AsyncIterable<InferenceResponse>>;
     ListProposals: AssistantService_ListProposals;
     CountPendingProposals: AssistantService_CountPendingProposals;
     /** Applies decisions, executes accepted calls, writes a proposal_result row and continues the turn. */
-    DecideProposals(request: DecideProposalsRequest, options?: {
-        signal?: AbortSignal;
-    }): Promise<AsyncIterable<InferenceResponse>>;
-    DecideProposalsGemini(request: DecideProposalsRequest, options?: {
-        signal?: AbortSignal;
-    }): Promise<AsyncIterable<InferenceResponse>>;
-    DecideProposalsLLM(request: DecideProposalsRequest, options?: {
-        signal?: AbortSignal;
-    }): Promise<AsyncIterable<InferenceResponse>>;
-    DecideProposalsChatGPT(request: DecideProposalsRequest, options?: {
-        signal?: AbortSignal;
-    }): Promise<AsyncIterable<InferenceResponse>>;
+    DecideProposals(request: DecideProposalsRequest, options?: RpcCallOptions): Promise<AsyncIterable<InferenceResponse>>;
+    DecideProposalsGemini(request: DecideProposalsRequest, options?: RpcCallOptions): Promise<AsyncIterable<InferenceResponse>>;
+    DecideProposalsLLM(request: DecideProposalsRequest, options?: RpcCallOptions): Promise<AsyncIterable<InferenceResponse>>;
+    DecideProposalsChatGPT(request: DecideProposalsRequest, options?: RpcCallOptions): Promise<AsyncIterable<InferenceResponse>>;
     GetConsent: AssistantService_GetConsent;
     GiveConsent: AssistantService_GiveConsent;
     CompactifyContext: AssistantService_CompactifyContext;

package/pushwoosh_agentrunner_v1.d.ts

@@ -4,214 +4,3 @@
  */
 export interface AgentRunner {
 }
-export type TurnRequest_payload_start = {
-    type: 'start';
-    data: StartTurn;
-};
-export type TurnRequest_payload_toolResult = {
-    type: 'toolResult';
-    data: ToolResult;
-};
-export type TurnRequest_payload = TurnRequest_payload_start | TurnRequest_payload_toolResult;
-export type TurnRequest = {
-    payload: TurnRequest_payload;
-};
-export type StartTurn = {
-    systemPrompt: SystemPrompt;
-    /**
-     * History as persisted, oldest first. How the runner seeds a session from it is its
-     * own business — that decision must not need a wire change.
-     */
-    history: HistoryMessage[];
-    tools: ToolDef[];
-    model: string;
-    maxTurns: number;
-    /** Correlation only: the runner never reads the chat, it just labels its logs. */
-    chatId: number;
-    /**
-     * A prior SDK session, when Go has one. Absent means seed from `history`, which stays
-     * the fallback for every failure on this path.
-     */
-    session: SdkSession;
-};
-/**
- * SdkSession is the CLI's own transcript, verbatim and opaque: Go stores the lines and
- * hands them back, the runner is the only side that parses them.
- */
-export type SdkSession = {
-    sessionId: string;
-    /** JSONL, one entry per line, oldest first */
-    lines: string[];
-};
-/**
- * Split because the split is what makes cross-session caching possible: joined into one
- * string, the SDK treats the whole prompt as session-specific.
- */
-export type SystemPrompt = {
-    /** identical for every chat and account */
-    static: string;
-    /** per-chat / per-account blocks, never cached */
-    dynamic: string[];
-};
-export type HistoryRole = 'HISTORY_ROLE_UNSPECIFIED' | 'HISTORY_ROLE_USER' | 'HISTORY_ROLE_ASSISTANT'
-/** assistant turn that called tools */
- | 'HISTORY_ROLE_TOOL_REQUEST' | 'HISTORY_ROLE_TOOL_RESULT'
-/** supersedes everything before it */
- | 'HISTORY_ROLE_CONTEXT_COMPACT';
-export type HistoryMessage = {
-    role: HistoryRole;
-    /** USER / ASSISTANT / CONTEXT_COMPACT */
-    text: string;
-    /** TOOL_REQUEST */
-    toolCalls: ToolCall[];
-    /** TOOL_RESULT */
-    toolResult: ToolResult;
-};
-/**
- * ToolDef carries the schema verbatim as JSON: it is what the MCP server the
- * runner stands up advertises, and Go stays the only owner of the tool surface.
- */
-export type ToolDef = {
-    name: string;
-    description: string;
-    inputSchemaJson: string;
-    /**
-     * Keep the schema out of the cached prompt prefix; the model reaches it through ToolSearch.
-     * False = pinned, so a caller that never sets it gets today's behaviour.
-     */
-    deferred: boolean;
-};
-export type ToolCall = {
-    id: string;
-    name: string;
-    argsJson: string;
-};
-export type ToolResult = {
-    id: string;
-    content: string;
-    isError: boolean;
-};
-export type TurnEvent_payload_delta = {
-    type: 'delta';
-    data: StreamDelta;
-};
-export type TurnEvent_payload_assistantMessage = {
-    type: 'assistantMessage';
-    data: AssistantMessage;
-};
-export type TurnEvent_payload_toolCall = {
-    type: 'toolCall';
-    data: ToolCall;
-};
-export type TurnEvent_payload_done = {
-    type: 'done';
-    data: TurnDone;
-};
-export type TurnEvent_payload_sessionAppend = {
-    type: 'sessionAppend';
-    data: SessionAppend;
-};
-export type TurnEvent_payload_rateLimit = {
-    type: 'rateLimit';
-    data: RateLimit;
-};
-export type TurnEvent_payload_messageStarted = {
-    type: 'messageStarted';
-    data: Usage;
-};
-export type TurnEvent_payload = TurnEvent_payload_delta | TurnEvent_payload_assistantMessage | TurnEvent_payload_toolCall | TurnEvent_payload_done | TurnEvent_payload_sessionAppend | TurnEvent_payload_rateLimit | TurnEvent_payload_messageStarted;
-export type TurnEvent = {
-    payload: TurnEvent_payload;
-};
-/**
- * Projection of the client-facing StreamEventType, mapped in one place in Go
- * (TestRunnerEnumsMapExhaustively), minus the events no runner can produce.
- */
-export type StreamEvent = 'STREAM_EVENT_UNSPECIFIED' | 'STREAM_EVENT_MESSAGE_START' | 'STREAM_EVENT_CONTENT_BLOCK_START' | 'STREAM_EVENT_TEXT_DELTA' | 'STREAM_EVENT_TOOL_USE_START' | 'STREAM_EVENT_TOOL_USE_DELTA' | 'STREAM_EVENT_TOOL_USE_RESULT' | 'STREAM_EVENT_CONTENT_BLOCK_STOP' | 'STREAM_EVENT_MESSAGE_DELTA' | 'STREAM_EVENT_MESSAGE_STOP' | 'STREAM_EVENT_THINKING_DELTA' | 'STREAM_EVENT_ERROR';
-export type StreamDelta = {
-    type: StreamEvent;
-    content: string;
-};
-/**
- * AssistantMessage is one SDK assistant message: usually a single content block of a model
- * turn, the SDK splits a response that way. Go persists this instead of re-accumulating deltas.
- */
-export type AssistantMessage = {
-    text: string;
-    toolCalls: ToolCall[];
-    usage: Usage;
-    /** mirrored to Mattermost, never persisted, never streamed */
-    thinking: string;
-    /** The API response this block belongs to: blocks sharing it are one model turn. */
-    messageId: string;
-};
-/**
- * SessionAppend carries transcript lines the SDK wrote during this turn, in write order. A
- * session_id Go has not seen replaces the stored session: that is how a fresh session resets it.
- */
-export type SessionAppend = {
-    sessionId: string;
-    lines: string[];
-};
-/**
- * RateLimit is the CLI's view of the subscription's windows — the only place the remaining
- * allowance is visible to us at all, and it arrives only on a subscription profile.
- */
-export type RateLimit = {
-    /** allowed | allowed_warning | rejected */
-    status: string;
-    /** five_hour | seven_day | seven_day_opus | … */
-    window: string;
-    /** percent of the window spent */
-    utilization: number;
-    /** unix seconds; 0 when the CLI did not say */
-    resetsAt: number;
-    /** allowed | allowed_warning | rejected: an allowed one serves a rejected window */
-    overageStatus: string;
-    /** this request was served on overage */
-    usingOverage: boolean;
-    /** why overage is off, e.g. overage_not_provisioned */
-    overageDisabledReason: string;
-};
-export type GetQuotaRequest = {};
-/**
- * Quota is read off the response headers of a one-token ping: unlike RateLimit, it carries
- * every window's utilization whether or not the CLI would have warned about it.
- */
-export type Quota = {
-    windows: QuotaWindow[];
-    /** allowed | allowed_warning | rejected */
-    overageStatus: string;
-    /** e.g. out_of_credits */
-    overageDisabledReason: string;
-    /** unix seconds of the last successful ping; 0 = none yet */
-    checkedAt: number;
-    /** the last ping's failure, empty when it succeeded */
-    error: string;
-};
-export type QuotaWindow = {
-    /** five_hour | seven_day | … */
-    window: string;
-    /** 0..100 */
-    utilizationPercent: number;
-    /** unix seconds */
-    resetsAt: number;
-    /** allowed | allowed_warning | rejected */
-    status: string;
-};
-export type Usage = {
-    model: string;
-    inputTokens: number;
-    outputTokens: number;
-    cacheCreationInputTokens: number;
-    cacheReadInputTokens: number;
-};
-export type TurnDone = {
-    stopReason: string;
-    /** empty on success */
-    error: string;
-    turns: number;
-    turnCapHit: boolean;
-    /** Turn totals from result.modelUsage; AssistantMessage.usage is a message_start snapshot. */
-    usage: Usage;
-};

package/pushwoosh_aigen_v1.d.ts

@@ -1,4 +1,4 @@
-import { RpcMethod, JsonObject } from './commonTypes';
+import { RpcMethod, RpcCallOptions, JsonObject } from './commonTypes';
 import { AiBuilderDocument as pushwoosh_aibuilder_v1_AiBuilderDocument } from './pushwoosh_aibuilder_v1';
 import { DocumentSettings as pushwoosh_aibuilder_v1_DocumentSettings } from './pushwoosh_aibuilder_v1';
 import { TextVariantStyle as pushwoosh_aibuilder_v1_TextVariantStyle } from './pushwoosh_aibuilder_v1';
@@ -43,9 +43,7 @@ export type AiGenService_GenerateArticleAnnotations = RpcMethod<AiGenArticleAnno
  * schema break.
  */
 export interface AiGenService {
-    Generate(request: AiGenRequest, options?: {
-        signal?: AbortSignal;
-    }): Promise<AsyncIterable<AiGenEvent>>;
+    Generate(request: AiGenRequest, options?: RpcCallOptions): Promise<AsyncIterable<AiGenEvent>>;
     /**
      * RenderHtml renders an existing AiBuilderDocument into a finished,
      * self-contained HTML document. The model writes the whole HTML — a
@@ -63,9 +61,7 @@ export interface AiGenService {
      * corner treatment are sampled by a vision-capable provider.
      * llm_provider / llm_model are the same dev-only overrides as Generate.
      */
-    RenderHtml(request: AiGenHtmlRequest, options?: {
-        signal?: AbortSignal;
-    }): Promise<AsyncIterable<AiGenHtmlEvent>>;
+    RenderHtml(request: AiGenHtmlRequest, options?: RpcCallOptions): Promise<AsyncIterable<AiGenHtmlEvent>>;
     /**
      * ExtractStyle samples a visual style from a single image: a colour
      * palette (keyed by semantic role) plus typography hints (body /