@kuralle-agents/cf-agent 0.13.0 → 0.14.0

package/dist/KuralleAgent.js

@@ -191,7 +191,7 @@ export class KuralleAgent extends AIChatAgent {
         let text = '';
         for await (const part of handle.events) {
             if (part.type === 'text-delta')
-                text += part.delta;
+                text += part.payload.delta;
         }
         await handle;
         if (text.trim()) {
@@ -277,7 +277,7 @@ export class KuralleAgent extends AIChatAgent {
         let text = '';
         for await (const part of handle.events) {
             if (part.type === 'text-delta')
-                text += part.delta;
+                text += part.payload.delta;
         }
         await handle;
         if (text.trim()) {

package/dist/StreamAdapter.d.ts

@@ -1,7 +1,7 @@
-import type { HarnessStreamPart } from '@kuralle-agents/core';
+import type { StreamPart } from '@kuralle-agents/core';
 import type { StreamAdapterConfig } from './types.js';
 /**
- * Convert Kuralle's HarnessStreamPart generator into an SSE Response
+ * Convert Kuralle's StreamPart generator into an SSE Response
  * that CF's AIChatAgent._reply() can parse.
  *
  * CF's _streamSSEReply reads lines in the format:
@@ -19,10 +19,10 @@ type SSETextState = {
     open: boolean;
     canceledIds: Set<string>;
 };
-export declare function createSSEResponse(stream: AsyncGenerator<HarnessStreamPart>, config?: StreamAdapterConfig): Response;
+export declare function createSSEResponse(stream: AsyncGenerator<StreamPart>, config?: StreamAdapterConfig): Response;
 /**
- * Convert a single HarnessStreamPart to SSE data lines.
+ * Convert a single StreamPart to SSE data lines.
  * Returns 0+ formatted SSE lines ("data: {...}\n\n").
  */
-export declare function convertToSSELines(part: HarnessStreamPart, config: StreamAdapterConfig, textState: SSETextState): string[];
+export declare function convertToSSELines(part: StreamPart, config: StreamAdapterConfig, textState: SSETextState): string[];
 export {};

package/dist/StreamAdapter.js

@@ -30,7 +30,7 @@ export function createSSEResponse(stream, config = DEFAULT_STREAM_CONFIG) {
     });
 }
 /**
- * Convert a single HarnessStreamPart to SSE data lines.
+ * Convert a single StreamPart to SSE data lines.
  * Returns 0+ formatted SSE lines ("data: {...}\n\n").
  */
 export function convertToSSELines(part, config, textState) {
@@ -49,19 +49,19 @@ export function convertToSSELines(part, config, textState) {
             closeTextSegment();
             break;
         case 'text-cancel': {
-            textState.canceledIds.add(part.id);
+            textState.canceledIds.add(part.payload.id);
             closeTextSegment();
             break;
         }
         case 'text-delta': {
-            if (textState.canceledIds.has(part.id)) {
+            if (textState.canceledIds.has(part.payload.id)) {
                 break;
             }
             if (!textState.open) {
                 lines.push(sse({ type: 'text-start' }));
                 textState.open = true;
             }
-            lines.push(sse({ type: 'text-delta', delta: part.delta }));
+            lines.push(sse({ type: 'text-delta', delta: part.payload.delta }));
             break;
         }
         case 'tool-call': {
@@ -69,17 +69,17 @@ export function convertToSSELines(part, config, textState) {
             // CF expects tool-input-available with toolCallId, toolName, input
             lines.push(sse({
                 type: 'tool-input-available',
-                toolCallId: part.toolCallId,
-                toolName: part.toolName,
-                input: config.includeToolArgs ? part.args : undefined,
+                toolCallId: part.payload.toolCallId,
+                toolName: part.payload.toolName,
+                input: config.includeToolArgs ? part.payload.args : undefined,
             }));
             break;
         }
         case 'tool-result': {
             lines.push(sse({
                 type: 'tool-output-available',
-                toolCallId: part.toolCallId,
-                output: part.result,
+                toolCallId: part.payload.toolCallId,
+                output: part.payload.result,
             }));
             break;
         }
@@ -89,8 +89,8 @@ export function convertToSSELines(part, config, textState) {
             lines.push(sse({
                 type: 'data-handoff',
                 data: {
-                    to: part.targetAgent,
-                    reason: part.reason,
+                    to: part.payload.targetAgent,
+                    reason: part.payload.reason,
                 },
             }));
             break;
@@ -100,7 +100,7 @@ export function convertToSSELines(part, config, textState) {
                 break;
             lines.push(sse({
                 type: 'data-flow-enter',
-                data: { flow: part.flow },
+                data: { flow: part.payload.flow },
             }));
             break;
         }
@@ -109,7 +109,7 @@ export function convertToSSELines(part, config, textState) {
                 break;
             lines.push(sse({
                 type: 'data-flow-node',
-                data: { node: part.nodeName, event: 'enter' },
+                data: { node: part.payload.nodeName, event: 'enter' },
             }));
             break;
         }
@@ -118,7 +118,7 @@ export function convertToSSELines(part, config, textState) {
                 break;
             lines.push(sse({
                 type: 'data-flow-node',
-                data: { node: part.nodeName, event: 'exit' },
+                data: { node: part.payload.nodeName, event: 'exit' },
             }));
             break;
         }
@@ -127,7 +127,7 @@ export function convertToSSELines(part, config, textState) {
                 break;
             lines.push(sse({
                 type: 'data-flow-transition',
-                data: { from: part.from, to: part.to },
+                data: { from: part.payload.from, to: part.payload.to },
             }));
             break;
         }
@@ -136,14 +136,14 @@ export function convertToSSELines(part, config, textState) {
                 break;
             lines.push(sse({
                 type: 'data-flow-end',
-                data: { reason: part.reason },
+                data: { reason: part.payload.reason },
             }));
             break;
         }
         case 'error': {
             lines.push(sse({
                 type: 'data-error',
-                data: { error: part.error },
+                data: { error: part.payload.error },
             }));
             break;
         }

package/dist/inbound-runtime.js

@@ -51,8 +51,8 @@ function turnResult(parts) {
     const paused = parts.find((part) => part.type === 'paused');
     return {
         parts,
-        suspended: paused ? { signalId: paused.waitingFor } : undefined,
-        handoffToHuman: parts.some((part) => part.type === 'handoff' && part.targetAgent === 'human'),
+        suspended: paused ? { signalId: paused.payload.waitingFor } : undefined,
+        handoffToHuman: parts.some((part) => part.type === 'handoff' && part.payload.targetAgent === 'human'),
     };
 }
 export class SqlInboundLedger {

package/dist/index.d.ts

@@ -35,4 +35,4 @@ export { AgentScheduleCoalesceScheduler, QueuedTurnRunner, RuntimeTurnRunner, Sq
 export type { StreamAdapterConfig, OrchestrationState, SqlExecutor, } from './types.js';
 export type { DurableObjectInboundRuntimeOptions, ScheduleHost, } from './inbound-runtime.js';
 export { DEFAULT_STREAM_CONFIG } from './types.js';
-export type { HarnessConfig, HarnessHooks, HarnessStreamPart, Session, } from '@kuralle-agents/core';
+export type { HarnessConfig, StreamPart, Session, } from '@kuralle-agents/core';

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@kuralle-agents/cf-agent",
-  "version": "0.13.0",
+  "version": "0.14.0",
   "description": "Kuralle agent integration for Cloudflare Workers with AIChatAgent",
   "type": "module",
   "main": "./dist/index.js",
@@ -9,10 +9,6 @@
     ".": {
       "import": "./dist/index.js",
       "types": "./dist/index.d.ts"
-    },
-    "./voice": {
-      "import": "./dist/voice/index.js",
-      "types": "./dist/voice/index.d.ts"
     }
   },
   "files": [
@@ -21,22 +17,19 @@
   "dependencies": {
     "@cloudflare/ai-chat": "^0.8.4",
     "ai": "^6.0.90",
-    "@kuralle-agents/core": "0.13.0",
-    "@kuralle-agents/realtime-audio": "0.13.0",
-    "@kuralle-agents/messaging": "0.13.0"
+    "@kuralle-agents/core": "0.14.0",
+    "@kuralle-agents/messaging": "0.14.0"
   },
   "peerDependencies": {
     "agents": ">=0.14.0 <1.0.0",
-    "zod": "^4.0.0",
-    "@kuralle-agents/voice-protocol": "0.13.0"
+    "zod": "^4.0.0"
   },
   "devDependencies": {
     "@cloudflare/vitest-pool-workers": "^0.12.7",
     "@cloudflare/workers-types": "^4.0.0",
     "agents": "^0.15.0",
     "typescript": "^5.7.0",
-    "vitest": "^3.2.4",
-    "@kuralle-agents/voice-protocol": "0.13.0"
+    "vitest": "^3.2.4"
   },
   "engines": {
     "node": ">=20"
@@ -60,7 +53,7 @@
     "prebuild": "rm -rf dist",
     "build": "tsc",
     "dev": "tsc --watch",
-    "test": "bun test src/voice/__tests__ src/__tests__ && vitest run --config vitest.config.ts",
+    "test": "bun test src/__tests__ && vitest run --config vitest.config.ts",
     "test:sql-memory-workers": "vitest run --config vitest.config.ts",
     "typecheck": "tsc --noEmit",
     "clean": "rm -rf dist"

package/dist/voice/AudioConnectionManager.d.ts

@@ -1,43 +0,0 @@
-/**
- * Vendored, trimmed port of AudioConnectionManager + sendVoiceJSON from
- * `@cloudflare/voice` (cloudflare/agents, packages/voice/src/audio-pipeline.ts,
- * MIT License, © 2025 Cloudflare, Inc.).
- *
- * Original provides cascaded-pipeline state (audio buffers, transcriber
- * sessions, abort pipelines). The realtime variant needs none of that —
- * audio is pass-through, turn detection happens provider-side, there is no
- * separate STT session to manage. We keep only the connection-level
- * in-call tracking primitive.
- *
- * Upstream source:
- *   https://github.com/cloudflare/agents/blob/main/packages/voice/src/audio-pipeline.ts
- */
-/**
- * Minimal per-connection state manager for the realtime voice mixin.
- *
- * Tracks which connection IDs currently have an open realtime session.
- * Owning the Map here (rather than inline in the mixin) keeps the mixin
- * class body focused on protocol handling and matches the shape the
- * cascaded lane uses.
- */
-export declare class AudioConnectionManager {
-    #private;
-    /** Mark a connection as having an active in-call state. */
-    initConnection(connectionId: string): void;
-    /** True if this connection currently has an active realtime session. */
-    isInCall(connectionId: string): boolean;
-    /** Clear in-call state for a connection (end_call or onClose). */
-    cleanup(connectionId: string): void;
-    /** Count of active in-call connections. Used for maxConcurrentSessions gate. */
-    size(): number;
-}
-/**
- * Serialize and push a voice-protocol frame to a client connection.
- *
- * Upstream accepts a log prefix; we drop it since the realtime path has no
- * chatty log calls. Signature intentionally minimal so the mixin's send
- * sites stay readable.
- */
-export declare function sendVoiceJSON(connection: {
-    send(data: string | ArrayBuffer): void;
-}, data: unknown): void;

package/dist/voice/AudioConnectionManager.js

@@ -1,51 +0,0 @@
-/**
- * Vendored, trimmed port of AudioConnectionManager + sendVoiceJSON from
- * `@cloudflare/voice` (cloudflare/agents, packages/voice/src/audio-pipeline.ts,
- * MIT License, © 2025 Cloudflare, Inc.).
- *
- * Original provides cascaded-pipeline state (audio buffers, transcriber
- * sessions, abort pipelines). The realtime variant needs none of that —
- * audio is pass-through, turn detection happens provider-side, there is no
- * separate STT session to manage. We keep only the connection-level
- * in-call tracking primitive.
- *
- * Upstream source:
- *   https://github.com/cloudflare/agents/blob/main/packages/voice/src/audio-pipeline.ts
- */
-/**
- * Minimal per-connection state manager for the realtime voice mixin.
- *
- * Tracks which connection IDs currently have an open realtime session.
- * Owning the Map here (rather than inline in the mixin) keeps the mixin
- * class body focused on protocol handling and matches the shape the
- * cascaded lane uses.
- */
-export class AudioConnectionManager {
-    #inCall = new Set();
-    /** Mark a connection as having an active in-call state. */
-    initConnection(connectionId) {
-        this.#inCall.add(connectionId);
-    }
-    /** True if this connection currently has an active realtime session. */
-    isInCall(connectionId) {
-        return this.#inCall.has(connectionId);
-    }
-    /** Clear in-call state for a connection (end_call or onClose). */
-    cleanup(connectionId) {
-        this.#inCall.delete(connectionId);
-    }
-    /** Count of active in-call connections. Used for maxConcurrentSessions gate. */
-    size() {
-        return this.#inCall.size;
-    }
-}
-/**
- * Serialize and push a voice-protocol frame to a client connection.
- *
- * Upstream accepts a log prefix; we drop it since the realtime path has no
- * chatty log calls. Signature intentionally minimal so the mixin's send
- * sites stay readable.
- */
-export function sendVoiceJSON(connection, data) {
-    connection.send(JSON.stringify(data));
-}

package/dist/voice/RealtimeVoiceAgent.d.ts

@@ -1,48 +0,0 @@
-/**
- * KuralleRealtimeVoiceAgent — thin concrete class applying the
- * {@link withRealtimeVoice} mixin to {@link KuralleAgent}.
- *
- * Subclasses must either set `realtimeModel` to a constructed
- * {@link RealtimeAudioClient}, or override `createRealtimeModel(env)` so the
- * mixin can lazily build one (usually a `CloudflareGeminiLiveClient` with
- * `this.env.GEMINI_API_KEY`).
- *
- * @example
- * ```ts
- * import { CloudflareGeminiLiveClient } from "@kuralle-agents/realtime-audio";
- * import { KuralleRealtimeVoiceAgent } from "@kuralle-agents/cf-agent/voice";
- *
- * export class MyVoiceAgent extends KuralleRealtimeVoiceAgent<Env> {
- *   createRealtimeModel(env: Env) {
- *     return new CloudflareGeminiLiveClient({ apiKey: env.GEMINI_API_KEY });
- *   }
- *   protected getAgents() { return [{ id: "assistant", name: "Assistant", tools: {} }]; }
- *   protected getDefaultAgentId() { return "assistant"; }
- * }
- * ```
- */
-import { KuralleAgent } from "../KuralleAgent.js";
-declare const RealtimeVoiceKuralleAgentBase: typeof KuralleAgent;
-export declare abstract class KuralleRealtimeVoiceAgent<Env = unknown, State = unknown> extends RealtimeVoiceKuralleAgentBase<Env, State> {
-    /**
-     * Hibernate between calls; stay awake during active calls via `keepAlive()`.
-     *
-     * Reasoning:
-     *   - Client WS uses the agents-SDK `Connection` abstraction (hibernation-
-     *     aware), so a reconnect after eviction lands on a rehydrated DO with
-     *     `this.messages` + `this.state` restored from SQLite.
-     *   - Outbound provider WS is not hibernation-aware, but `keepAlive()`
-     *     is held for the duration of a call → no eviction during calls.
-     *   - On `end_call` / teardown we close the outbound cleanly, then release
-     *     keepAlive → DO can safely hibernate (free) between calls.
-     *
-     * Contrast: `cloudflare/agents/openai-sdk/call-my-agent` sets
-     * `hibernate: false` because its `TwilioRealtimeTransportLayer` attaches
-     * raw `.addEventListener` handlers bypassing the hibernation path — that
-     * doesn't apply to us.
-     */
-    static options: {
-        hibernate: boolean;
-    };
-}
-export {};

package/dist/voice/RealtimeVoiceAgent.js

@@ -1,46 +0,0 @@
-/**
- * KuralleRealtimeVoiceAgent — thin concrete class applying the
- * {@link withRealtimeVoice} mixin to {@link KuralleAgent}.
- *
- * Subclasses must either set `realtimeModel` to a constructed
- * {@link RealtimeAudioClient}, or override `createRealtimeModel(env)` so the
- * mixin can lazily build one (usually a `CloudflareGeminiLiveClient` with
- * `this.env.GEMINI_API_KEY`).
- *
- * @example
- * ```ts
- * import { CloudflareGeminiLiveClient } from "@kuralle-agents/realtime-audio";
- * import { KuralleRealtimeVoiceAgent } from "@kuralle-agents/cf-agent/voice";
- *
- * export class MyVoiceAgent extends KuralleRealtimeVoiceAgent<Env> {
- *   createRealtimeModel(env: Env) {
- *     return new CloudflareGeminiLiveClient({ apiKey: env.GEMINI_API_KEY });
- *   }
- *   protected getAgents() { return [{ id: "assistant", name: "Assistant", tools: {} }]; }
- *   protected getDefaultAgentId() { return "assistant"; }
- * }
- * ```
- */
-import { KuralleAgent } from "../KuralleAgent.js";
-import { withRealtimeVoice } from "./withRealtimeVoice.js";
-const RealtimeVoiceKuralleAgentBase = withRealtimeVoice(KuralleAgent);
-export class KuralleRealtimeVoiceAgent extends RealtimeVoiceKuralleAgentBase {
-    /**
-     * Hibernate between calls; stay awake during active calls via `keepAlive()`.
-     *
-     * Reasoning:
-     *   - Client WS uses the agents-SDK `Connection` abstraction (hibernation-
-     *     aware), so a reconnect after eviction lands on a rehydrated DO with
-     *     `this.messages` + `this.state` restored from SQLite.
-     *   - Outbound provider WS is not hibernation-aware, but `keepAlive()`
-     *     is held for the duration of a call → no eviction during calls.
-     *   - On `end_call` / teardown we close the outbound cleanly, then release
-     *     keepAlive → DO can safely hibernate (free) between calls.
-     *
-     * Contrast: `cloudflare/agents/openai-sdk/call-my-agent` sets
-     * `hibernate: false` because its `TwilioRealtimeTransportLayer` attaches
-     * raw `.addEventListener` handlers bypassing the hibernation path — that
-     * doesn't apply to us.
-     */
-    static options = { hibernate: true };
-}

package/dist/voice/index.d.ts

@@ -1,13 +0,0 @@
-/**
- * @kuralle-agents/cf-agent/voice — realtime voice mixin + agent base for
- * Cloudflare Durable Objects. Pairs with any
- * {@link RealtimeAudioClient} (v2) implementation — Gemini Live,
- * OpenAI Realtime, xAI Grok, etc.
- *
- * For the cascaded (STT → LLM → TTS) variant, see the sibling
- * `KuralleCascadedVoiceAgent` export.
- */
-export { withRealtimeVoice } from "./withRealtimeVoice.js";
-export type { RealtimeVoiceOptions, RealtimeVoiceMixinMembers, } from "./withRealtimeVoice.js";
-export { KuralleRealtimeVoiceAgent } from "./RealtimeVoiceAgent.js";
-export { AudioConnectionManager, sendVoiceJSON } from "./AudioConnectionManager.js";

package/dist/voice/index.js

@@ -1,12 +0,0 @@
-/**
- * @kuralle-agents/cf-agent/voice — realtime voice mixin + agent base for
- * Cloudflare Durable Objects. Pairs with any
- * {@link RealtimeAudioClient} (v2) implementation — Gemini Live,
- * OpenAI Realtime, xAI Grok, etc.
- *
- * For the cascaded (STT → LLM → TTS) variant, see the sibling
- * `KuralleCascadedVoiceAgent` export.
- */
-export { withRealtimeVoice } from "./withRealtimeVoice.js";
-export { KuralleRealtimeVoiceAgent } from "./RealtimeVoiceAgent.js";
-export { AudioConnectionManager, sendVoiceJSON } from "./AudioConnectionManager.js";

package/dist/voice/withRealtimeVoice.d.ts

@@ -1,63 +0,0 @@
-/**
- * withRealtimeVoice — functional mixin adding realtime voice to an
- * AIChatAgent-derived Durable Object class.
- *
- * Realtime voice convergence: this mixin is now pure transport glue. All
- * orchestration — tool dispatch, flow transitions, hooks, handoffs, session
- * persistence, reconnect strategy, chat_ctx replay — lives in Kuralle's
- * `Runtime` and `CloudflareRealtimeAdapter`.
- *
- * Responsibilities that stay in the mixin:
- *   - Intercept voice-protocol frames on the browser ↔ DO WebSocket.
- *   - Per-connection concurrent-session cap (`maxConcurrentSessions`).
- *   - DO-specific lifecycle (`keepAlive()` during an active call).
- *   - Expose subclass-override hooks (`beforeSessionStart`, `onSessionStart`,
- *     `onSessionEnd`, `onInterrupt`, `onUserTranscript`, `onModelTranscript`).
- *   - Build the `Runtime` per call (via `createRuntime(...)`),
- *     wire in a `BridgeSessionStore`, and hand control to the adapter.
- *
- * Responsibilities moved OUT of the mixin:
- *   - Tool execution              → the `Runtime` effect path (`ctx.tool`)
- *   - Flow reconfigure / handoff  → same, via the runtime's transition handling
- *   - Transcript persistence      → the `Runtime` session + event log +
- *                                   existing `AIChatAgent.persistMessages`
- *   - Resumption handle + chat_ctx replay → provider client (in-memory) +
- *                                           adapter (reconnect dispatch)
- *   - SQLite tables `cf_realtime_resumption` and `cf_realtime_chat_ctx`
- *     — DELETED. Storage flows through `SessionStore` (BridgeSessionStore on CF).
- *
- * Capture-and-patch lifecycle wrap mirrors `@cloudflare/voice`'s pattern
- * (voice.ts:251-334, MIT, © 2025 Cloudflare, Inc.).
- */
-import type { Connection } from "agents";
-import type { RealtimeAudioClient } from "@kuralle-agents/core/realtime";
-type Constructor<T = object> = new (...args: unknown[]) => T;
-type AbstractConstructor<T = object> = abstract new (...args: unknown[]) => T;
-type AgentLike = Constructor | AbstractConstructor;
-export interface RealtimeVoiceOptions {
-    /** Audio wire format. Fixed to PCM16 today; future provider-specific variants. */
-    audioFormat?: "pcm16";
-    /** Per-DO concurrent-session cap. 5th `start_call` rejected with an `error` frame. Default 4. */
-    maxConcurrentSessions?: number;
-    /** Max outbound-reconnect attempts under backoff before giving up. Default 3. */
-    reconnectMaxAttempts?: number;
-    /** Base delay for backoff (delay ∈ [0, min(cap, base * 2^attempt)]). Default 500ms. */
-    reconnectBaseDelayMs?: number;
-    /** Cap on the backoff delay. Default 8000ms. */
-    reconnectCapDelayMs?: number;
-    /** Bytes of PCM to ring-buffer per-connection across a reconnect window. Default 48_000. */
-    reconnectAudioBufferBytes?: number;
-}
-export interface RealtimeVoiceMixinMembers {
-    realtimeModel?: RealtimeAudioClient;
-    createRealtimeModel?(env: unknown): RealtimeAudioClient;
-    beforeSessionStart(conn: Connection): boolean | Promise<boolean>;
-    onSessionStart(conn: Connection): void | Promise<void>;
-    onSessionEnd(conn: Connection): void | Promise<void>;
-    onInterrupt(conn: Connection): void | Promise<void>;
-    onUserTranscript(text: string, isFinal: boolean, conn: Connection): void | Promise<void>;
-    onModelTranscript(text: string, isFinal: boolean, conn: Connection): void | Promise<void>;
-    forceEndSession(conn: Connection): void;
-}
-export declare function withRealtimeVoice<TBase extends AgentLike>(Base: TBase, opts?: RealtimeVoiceOptions): TBase & Constructor<RealtimeVoiceMixinMembers>;
-export {};

package/dist/voice/withRealtimeVoice.js

@@ -1,451 +0,0 @@
-/**
- * withRealtimeVoice — functional mixin adding realtime voice to an
- * AIChatAgent-derived Durable Object class.
- *
- * Realtime voice convergence: this mixin is now pure transport glue. All
- * orchestration — tool dispatch, flow transitions, hooks, handoffs, session
- * persistence, reconnect strategy, chat_ctx replay — lives in Kuralle's
- * `Runtime` and `CloudflareRealtimeAdapter`.
- *
- * Responsibilities that stay in the mixin:
- *   - Intercept voice-protocol frames on the browser ↔ DO WebSocket.
- *   - Per-connection concurrent-session cap (`maxConcurrentSessions`).
- *   - DO-specific lifecycle (`keepAlive()` during an active call).
- *   - Expose subclass-override hooks (`beforeSessionStart`, `onSessionStart`,
- *     `onSessionEnd`, `onInterrupt`, `onUserTranscript`, `onModelTranscript`).
- *   - Build the `Runtime` per call (via `createRuntime(...)`),
- *     wire in a `BridgeSessionStore`, and hand control to the adapter.
- *
- * Responsibilities moved OUT of the mixin:
- *   - Tool execution              → the `Runtime` effect path (`ctx.tool`)
- *   - Flow reconfigure / handoff  → same, via the runtime's transition handling
- *   - Transcript persistence      → the `Runtime` session + event log +
- *                                   existing `AIChatAgent.persistMessages`
- *   - Resumption handle + chat_ctx replay → provider client (in-memory) +
- *                                           adapter (reconnect dispatch)
- *   - SQLite tables `cf_realtime_resumption` and `cf_realtime_chat_ctx`
- *     — DELETED. Storage flows through `SessionStore` (BridgeSessionStore on CF).
- *
- * Capture-and-patch lifecycle wrap mirrors `@cloudflare/voice`'s pattern
- * (voice.ts:251-334, MIT, © 2025 Cloudflare, Inc.).
- */
-import { createRuntime } from "@kuralle-agents/core";
-import { VOICE_PROTOCOL_VERSION } from "@kuralle-agents/voice-protocol";
-import { CloudflareRealtimeAdapter, } from "@kuralle-agents/realtime-audio";
-import { AudioConnectionManager, sendVoiceJSON, } from "./AudioConnectionManager.js";
-import { BridgeSessionStore } from "../BridgeSessionStore.js";
-import { durableAgentSurface, } from "../durable-agent-surface.js";
-function voiceHost(self) {
-    return durableAgentSurface(self);
-}
-function resolveSql(self) {
-    const { sql } = durableAgentSurface(self);
-    if (typeof sql !== "function")
-        return undefined;
-    return sql.bind(self);
-}
-function inferModelPolicy(client) {
-    const provider = (client.provider ?? "unknown").toLowerCase();
-    const supportsInstructionUpdate = client.capabilities?.midSessionInstructionsUpdate === true ||
-        client.capabilities?.midSessionChatCtxUpdate === true;
-    if (provider.startsWith("gemini") || provider.startsWith("google")) {
-        return { provider: "google", supportsInstructionUpdate };
-    }
-    if (provider.startsWith("openai")) {
-        return { provider: "openai", supportsInstructionUpdate };
-    }
-    if (provider.startsWith("azure")) {
-        return { provider: "azure-openai", supportsInstructionUpdate };
-    }
-    if (provider.startsWith("xai")) {
-        return { provider: "xai", supportsInstructionUpdate };
-    }
-    if (provider.startsWith("phonic")) {
-        return { provider: "phonic", supportsInstructionUpdate };
-    }
-    if (provider.startsWith("workers")) {
-        return { provider: "workers-ai", supportsInstructionUpdate };
-    }
-    return { provider: "unknown", supportsInstructionUpdate };
-}
-export function withRealtimeVoice(Base, opts = {}) {
-    const maxConcurrent = opts.maxConcurrentSessions ?? 4;
-    const reconnectMaxAttempts = opts.reconnectMaxAttempts ?? 3;
-    const reconnectBaseDelayMs = opts.reconnectBaseDelayMs ?? 500;
-    const reconnectCapDelayMs = opts.reconnectCapDelayMs ?? 8000;
-    const reconnectAudioBufferBytes = opts.reconnectAudioBufferBytes ?? 48_000;
-    class RealtimeVoiceMixin extends Base {
-        realtimeModel;
-        createRealtimeModel(_env) {
-            throw new Error("Subclass must override createRealtimeModel or set realtimeModel.");
-        }
-        #cm = new AudioConnectionManager();
-        #adapters = new Map();
-        #agents = new Map();
-        #keepAlive = new Map();
-        #starting = new Set();
-        /** One-shot guard: emit `session_lost` at most once per conn. */
-        #sessionLossSignalled = new Set();
-        /** Warn once per instance when host lacks `keepAlive()` — calls will die at 70-140s idle. */
-        #keepAliveWarned = false;
-        // eslint-disable-next-line @typescript-eslint/no-explicit-any -- reason: TypeScript TS2545 mixin constructors require `any[]` rest params
-        constructor(...args) {
-            super(...args);
-            const host = voiceHost(this);
-            const _onConnect = host.onConnect?.bind(this);
-            const _onClose = host.onClose?.bind(this);
-            const _onMessage = host.onMessage?.bind(this);
-            host.onConnect = (c, ...rest) => {
-                sendVoiceJSON(c, { type: "welcome", protocol_version: VOICE_PROTOCOL_VERSION });
-                sendVoiceJSON(c, { type: "status", status: "idle" });
-                return _onConnect?.(c, ...rest);
-            };
-            host.onClose = (c, ...rest) => {
-                void this.#teardown(c);
-                return _onClose?.(c, ...rest);
-            };
-            host.onMessage = (c, m) => {
-                if (m instanceof ArrayBuffer) {
-                    const frame = new Uint8Array(m);
-                    const adapter = this.#adapters.get(c.id);
-                    if (adapter) {
-                        adapter.sendUserAudio(frame);
-                        return;
-                    }
-                    // No adapter for this conn. Three plausible causes:
-                    //   1. Mid-call DO eviction — isolate died, `#adapters` wiped, client
-                    //      WS re-hydrated by hibernation, audio now lands on a fresh
-                    //      mixin with no session to forward to.
-                    //   2. Teardown race — client sent last 20ms of audio after
-                    //      `end_call`. `@cloudflare/voice` stops streaming on end so
-                    //      this is near-zero in practice.
-                    //   3. Buggy client streaming pre-`start_call`.
-                    // In all three cases, silently dropping audio is worse than
-                    // signalling. Emit once + close so partysocket reconnects cleanly
-                    // and the user can restart.
-                    if (this.#starting.has(c.id))
-                        return;
-                    if (this.#sessionLossSignalled.has(c.id))
-                        return;
-                    this.#sessionLossSignalled.add(c.id);
-                    sendVoiceJSON(c, {
-                        type: "error",
-                        code: "session_lost",
-                        message: "Voice session lost (server restart, eviction, or no active call). " +
-                            "Please start a new call.",
-                    });
-                    try {
-                        c.close(1011, "session_lost");
-                    }
-                    catch {
-                        /* connection already closing */
-                    }
-                    return;
-                }
-                if (typeof m !== "string")
-                    return _onMessage?.(c, m);
-                let parsed;
-                try {
-                    parsed = JSON.parse(m);
-                }
-                catch {
-                    return _onMessage?.(c, m);
-                }
-                switch (parsed.type) {
-                    case "start_call":
-                        this.#handleStartCall(c).catch((err) => {
-                            console.error("[withRealtimeVoice] #handleStartCall REJECTED:", String(err), err instanceof Error ? err.stack : "");
-                            const raw = err instanceof Error ? err.message : String(err);
-                            sendVoiceJSON(c, { type: "error", message: `start_call failed: ${raw}` });
-                        });
-                        return;
-                    case "end_call":
-                        void this.#teardown(c);
-                        return;
-                    case "interrupt":
-                        void this.#handleInterrupt(c);
-                        return;
-                    case "text_message": {
-                        const text = typeof parsed.text === "string" ? parsed.text : undefined;
-                        if (text)
-                            this.#adapters.get(c.id)?.sendUserText(text);
-                        return;
-                    }
-                    case "start_of_speech":
-                    case "end_of_speech":
-                    case "hello":
-                        return; // acknowledge but no-op
-                    default:
-                        return _onMessage?.(c, m);
-                }
-            };
-        }
-        // ─── Hook defaults (overridable by subclasses) ──────────────────────────
-        async beforeSessionStart(_c) {
-            return true;
-        }
-        async onSessionStart(_c) { }
-        async onSessionEnd(_c) { }
-        async onInterrupt(_c) { }
-        async onUserTranscript(_t, _f, _c) { }
-        async onModelTranscript(_t, _f, _c) { }
-        /** Imperative end-session helper for subclasses (e.g. timeout, moderation). */
-        forceEndSession(c) {
-            void this.#teardown(c);
-        }
-        // ─── Internals ───────────────────────────────────────────────────────────
-        #resolveModel() {
-            if (this.realtimeModel)
-                return this.realtimeModel;
-            if (this.createRealtimeModel) {
-                const { env } = voiceHost(this);
-                return this.createRealtimeModel(env);
-            }
-            throw new Error("withRealtimeVoice: subclass must provide `realtimeModel` or override `createRealtimeModel(env)`.");
-        }
-        #resolveDefaultAgent() {
-            const host = voiceHost(this);
-            const agents = host.getAgents?.() ?? [];
-            const defaultId = host.getDefaultAgentId?.() ?? agents[0]?.id;
-            const agent = agents.find((a) => a.id === defaultId) ?? agents[0];
-            if (!agent) {
-                throw new Error("withRealtimeVoice: subclass must implement getAgents()/getDefaultAgentId() returning at least one agent.");
-            }
-            return agent;
-        }
-        #buildRuntime(sessionId) {
-            const host = voiceHost(this);
-            const defaultAgentId = host.getDefaultAgentId?.() ?? "";
-            const extra = host.getRuntimeConfig?.() ?? {};
-            const agents = host.getAgents?.() ?? [];
-            const sql = resolveSql(this);
-            // BridgeSessionStore is the CF-native SessionStore — messages live in
-            // AIChatAgent's `cf_ai_chat_agent_messages`, orchestration state in the
-            // `kuralle_orchestration` row. No voice-specific tables.
-            const sessionStore = sql
-                ? new BridgeSessionStore({
-                    sqlExecutor: sql,
-                    cfMessages: host.messages ?? [],
-                    sessionId,
-                    defaultAgentId,
-                })
-                : undefined;
-            return createRuntime({
-                ...extra,
-                agents,
-                defaultAgentId,
-                sessionStore,
-            });
-        }
-        /**
-         * Generate a **call-scoped** session id. Each `start_call` gets its own
-         * UUID so orchestration state (currentAgent, flow node, handoffHistory)
-         * is isolated per call — the 2nd call doesn't inherit the 1st call's
-         * flow-node position. DO-scoped state that survives across calls
-         * (Gemini resumption handle, long-term memory) belongs in `this.state`.
-         *
-         * Prior version returned `ctx.id.toString()` which collapsed every
-         * call in a DO into one orchestration row — fine for demo but wrong
-         * for any agent with flow transitions.
-         */
-        #mintCallId() {
-            return crypto.randomUUID();
-        }
-        /** DO-stable identity, distinct from per-call session id. */
-        #getDoId() {
-            const id = voiceHost(this).ctx?.id;
-            return id?.toString?.() ?? "unknown-do";
-        }
-        /**
-         * Read a previously-captured provider resumption handle from DO-scoped
-         * `this.state`. Returns undefined if missing, from a different provider,
-         * or older than the provider's TTL (Gemini: 2h). DO-scoped means
-         * `hibernate: true` + fresh wake + new call will still see it.
-         */
-        #readDoResumption(provider) {
-            const state = (voiceHost(this).state ?? {});
-            const r = state.voiceResumption;
-            if (!r || r.provider !== provider)
-                return undefined;
-            const TWO_HOURS = 2 * 60 * 60 * 1000;
-            if (Date.now() - r.issuedAt > TWO_HOURS)
-                return undefined;
-            return r.handle;
-        }
-        #writeDoResumption(handle, provider) {
-            const host = voiceHost(this);
-            if (typeof host.setState !== 'function')
-                return;
-            const prev = (host.state ?? {});
-            host.setState({
-                ...prev,
-                voiceResumption: { handle, provider, issuedAt: Date.now() },
-            });
-        }
-        async #handleStartCall(conn) {
-            if (this.#adapters.has(conn.id) || this.#starting.has(conn.id)) {
-                // Duplicate start_call on the same connection — ignore.
-                return;
-            }
-            // Count in-flight starts toward the cap — otherwise concurrent start_call
-            // frames from N distinct connections can all read size=0 before any has
-            // populated the adapter map, over-subscribing the DO.
-            if (this.#adapters.size + this.#starting.size >= maxConcurrent) {
-                sendVoiceJSON(conn, {
-                    type: "error",
-                    code: "session_cap_exceeded",
-                    message: `Max concurrent realtime sessions (${maxConcurrent}) reached.`,
-                });
-                return;
-            }
-            this.#starting.add(conn.id);
-            let client;
-            try {
-                client = this.#resolveModel();
-            }
-            catch (err) {
-                this.#starting.delete(conn.id);
-                sendVoiceJSON(conn, { type: "error", message: String(err) });
-                return;
-            }
-            const canStart = await this.beforeSessionStart(conn);
-            if (!canStart) {
-                this.#starting.delete(conn.id);
-                return;
-            }
-            const agent = this.#resolveDefaultAgent();
-            this.#agents.set(conn.id, agent);
-            // Keep the DO alive for the duration of the call (matches @cloudflare/voice).
-            const { keepAlive } = voiceHost(this);
-            if (keepAlive) {
-                try {
-                    const dispose = await keepAlive.call(this);
-                    this.#keepAlive.set(conn.id, dispose);
-                }
-                catch (err) {
-                    // Host exposes keepAlive but it threw — that's a real bug, surface it.
-                    console.error("[withRealtimeVoice] keepAlive() threw — call will not be protected " +
-                        "from idle eviction (70-140s):", err instanceof Error ? err.message : String(err));
-                }
-            }
-            else if (!this.#keepAliveWarned) {
-                // Host lacks `keepAlive()` entirely — this is an Kuralle usage error.
-                // The SDK-shipped Agent / AIChatAgent classes both provide it; a
-                // missing implementation means the mixin was applied to a custom base
-                // that didn't extend those. Without keepAlive, calls die at ~70-140s
-                // of audio-flow idle (VAD silence during a turn).
-                this.#keepAliveWarned = true;
-                console.warn("[withRealtimeVoice] host does not expose `keepAlive()` — realtime " +
-                    "calls will be killed by DO idle eviction at ~70-140s. Extend " +
-                    "`KuralleRealtimeVoiceAgent` (which inherits from `AIChatAgent`) " +
-                    "rather than applying the mixin to a custom base.");
-            }
-            this.#cm.initConnection(conn.id);
-            // Per-call session id. DO identity (`this.ctx.id`) gives scalability
-            // across users (distinct `name` on the client → distinct DO); call id
-            // gives isolation between successive calls within the same DO.
-            const sessionId = this.#mintCallId();
-            const runtime = this.#buildRuntime(sessionId);
-            // DO-scoped resumption handle (if the same user previously called and
-            // their Gemini session handle is still within the 2h TTL, pass it in
-            // so the provider resumes server-side state — audio/text context).
-            const initialResumptionHandle = this.#readDoResumption(client.provider);
-            const adapter = new CloudflareRealtimeAdapter({
-                runtime,
-                client,
-                sessionId,
-                agentId: agent.id,
-                modelPolicy: inferModelPolicy(client),
-                reconnectMaxAttempts,
-                reconnectBaseDelayMs,
-                reconnectCapDelayMs,
-                reconnectAudioBufferBytes,
-                initialResumptionHandle,
-                sendJson: (frame) => sendVoiceJSON(conn, frame),
-                sendBinary: (data) => conn.send(data),
-                onResumptionHandle: (handle, meta) => {
-                    // Persist DO-scoped (`this.state`), not call-scoped. Survives the
-                    // end of this call and is available to the next `start_call` on
-                    // the same DO.
-                    this.#writeDoResumption(handle, meta.provider);
-                },
-                onReconnecting: () => {
-                    sendVoiceJSON(conn, { type: "status", status: "connecting" });
-                },
-                onReconnected: () => {
-                    sendVoiceJSON(conn, { type: "status", status: "listening" });
-                },
-                onEnd: () => {
-                    void this.#teardown(conn);
-                },
-                onError: (err) => {
-                    console.error("[withRealtimeVoice] adapter error:", err.message);
-                },
-            });
-            this.#adapters.set(conn.id, adapter);
-            try {
-                await adapter.start();
-            }
-            catch (err) {
-                this.#adapters.delete(conn.id);
-                this.#starting.delete(conn.id);
-                throw err;
-            }
-            this.#starting.delete(conn.id);
-            sendVoiceJSON(conn, { type: "audio_config", format: "pcm16", sampleRate: 24000 });
-            sendVoiceJSON(conn, { type: "status", status: "listening" });
-            // Wrap adapter transcript frames so subclass hooks fire alongside UI events.
-            // The adapter already sent `{type:"transcript", role, text}` via sendJson;
-            // we surface to the hook here. isFinal is always true in the current
-            // RealtimeAudioClient transcript stream (every emission is cumulative).
-            const userHook = this.onUserTranscript.bind(this);
-            const modelHook = this.onModelTranscript.bind(this);
-            const clientForHooks = client;
-            clientForHooks.on("transcript", (text, role) => {
-                sendVoiceJSON(conn, { type: "transcript", role, text });
-                const fn = role === "user" ? userHook : modelHook;
-                void fn(text, true, conn);
-            });
-            await this.onSessionStart(conn);
-        }
-        async #handleInterrupt(conn) {
-            this.#adapters.get(conn.id)?.sendInterrupt();
-            sendVoiceJSON(conn, { type: "status", status: "listening" });
-            await this.onInterrupt(conn);
-        }
-        async #teardown(conn) {
-            this.#starting.delete(conn.id);
-            this.#sessionLossSignalled.delete(conn.id);
-            const adapter = this.#adapters.get(conn.id);
-            if (adapter) {
-                this.#adapters.delete(conn.id);
-                try {
-                    await adapter.stop("teardown");
-                }
-                catch {
-                    /* already closed */
-                }
-            }
-            this.#agents.delete(conn.id);
-            this.#cm.cleanup(conn.id);
-            const dispose = this.#keepAlive.get(conn.id);
-            if (dispose) {
-                this.#keepAlive.delete(conn.id);
-                try {
-                    dispose();
-                }
-                catch {
-                    /* ignore */
-                }
-            }
-            sendVoiceJSON(conn, { type: "status", status: "idle" });
-            try {
-                await this.onSessionEnd(conn);
-            }
-            catch {
-                /* ignore */
-            }
-        }
-    }
-    return RealtimeVoiceMixin;
-}