feat: improve audio streaming and add wait-for-agent cancellation support

This commit is contained in:
Mohamed Boudra
2025-11-16 18:02:13 +01:00
parent e1bcb97e29
commit 9feec81047
6 changed files with 249 additions and 33 deletions
+24 -1
View File
@@ -24,6 +24,10 @@ import { curateAgentActivity } from "./activity-curator.js";
import { AGENT_PROVIDER_DEFINITIONS } from "./provider-manifest.js";
import { AgentRegistry } from "./agent-registry.js";
import { createWorktree } from "../../utils/worktree.js";
import {
WaitForAgentTracker,
type WaitForAgentCanceler,
} from "./wait-for-agent-tracker.js";
export interface AgentMcpServerOptions {
agentManager: AgentManager;
@@ -103,6 +107,7 @@ async function serializeSnapshotWithMetadata(
async function waitForAgentEvent(
agentManager: AgentManager,
agentId: string,
tracker: WaitForAgentTracker,
signal?: AbortSignal
): Promise<{
status: AgentLifecycleStatus;
@@ -142,6 +147,19 @@ async function waitForAgentEvent(
resolve({ status: currentStatus, permission });
};
const unregister = tracker.register(agentId, (reason) => {
cleanup();
reject(
Object.assign(
new Error(reason ?? "wait_for_agent cancelled"),
{
name: "AbortError",
}
)
);
});
cleanupFns.push(unregister);
const unsubscribe = agentManager.subscribe(
(event) => {
if (event.type === "agent_state") {
@@ -329,7 +347,12 @@ export async function createAgentMcpServer(
},
},
async ({ agentId }, { signal }) => {
const result = await waitForAgentEvent(agentManager, agentId, signal);
const result = await waitForAgentEvent(
agentManager,
agentId,
waitTracker,
signal
);
const activity = buildActivityPayload(agentManager, agentId);
return {
@@ -0,0 +1,40 @@
import { describe, expect, it, vi } from "vitest";
import { WaitForAgentTracker } from "./wait-for-agent-tracker";
describe("WaitForAgentTracker", () => {
it("registers and cancels waiters per agent", () => {
const tracker = new WaitForAgentTracker();
const cancelA = vi.fn();
const cancelB = vi.fn();
const unregisterA = tracker.register("agent-a", cancelA);
tracker.register("agent-a", cancelB);
expect(tracker.cancel("agent-a", "test")).toBe(true);
expect(cancelA).toHaveBeenCalledWith("test");
expect(cancelB).toHaveBeenCalledWith("test");
// Subsequent cancels should no-op once cleared
expect(tracker.cancel("agent-a")).toBe(false);
// Unregister removes callback without triggering it
const cancelC = vi.fn();
const unregisterC = tracker.register("agent-b", cancelC);
unregisterC();
expect(tracker.cancel("agent-b")).toBe(false);
expect(cancelC).not.toHaveBeenCalled();
});
it("supports cancelling all waiters", () => {
const tracker = new WaitForAgentTracker();
const cancelA = vi.fn();
const cancelB = vi.fn();
tracker.register("agent-a", cancelA);
tracker.register("agent-b", cancelB);
expect(tracker.cancelAll()).toBe(2);
expect(cancelA).toHaveBeenCalled();
expect(cancelB).toHaveBeenCalled();
});
});
@@ -0,0 +1,58 @@
export type WaitForAgentCanceler = (agentId: string, reason?: string) => boolean;
/**
* Tracks long-running wait_for_agent tool calls so they can be cancelled
* explicitly (e.g., when a realtime barge-in aborts the current turn).
*/
export class WaitForAgentTracker {
private waiters = new Map<string, Set<(reason?: string) => void>>();
register(agentId: string, cancel: (reason?: string) => void): () => void {
if (!this.waiters.has(agentId)) {
this.waiters.set(agentId, new Set());
}
const waitersForAgent = this.waiters.get(agentId)!;
waitersForAgent.add(cancel);
return () => {
const current = this.waiters.get(agentId);
if (!current) {
return;
}
current.delete(cancel);
if (current.size === 0) {
this.waiters.delete(agentId);
}
};
}
cancel(agentId: string, reason?: string): boolean {
const waitersForAgent = this.waiters.get(agentId);
if (!waitersForAgent || waitersForAgent.size === 0) {
return false;
}
this.waiters.delete(agentId);
for (const cancel of waitersForAgent) {
try {
cancel(reason);
} catch (error) {
console.warn(
`[WaitForAgentTracker] Cancel callback failed for ${agentId}:`,
error
);
}
}
return true;
}
cancelAll(reason?: string): number {
let cancelled = 0;
for (const agentId of Array.from(this.waiters.keys())) {
if (this.cancel(agentId, reason)) {
cancelled += 1;
}
}
return cancelled;
}
}
+16
View File
@@ -1587,7 +1587,23 @@ export class Session {
`[Session ${this.clientId}] Buffered audio chunk (${chunkBuffer.length} bytes, chunks: ${this.audioBuffer.chunks.length}${this.audioBuffer.isPCM ? `, PCM bytes: ${this.audioBuffer.totalPCMBytes}` : ""})`
);
// In realtime mode, only process audio when the user has finished speaking (isLast = true)
// This prevents partial transcriptions from being sent to the LLM
if (this.isRealtimeMode) {
if (!msg.isLast) {
console.log(
`[Session ${this.clientId}] Realtime mode: buffering audio, waiting for speech end`
);
return;
}
console.log(
`[Session ${this.clientId}] Realtime mode: speech ended, processing complete audio`
);
}
// In non-realtime mode, use streaming threshold to process chunks
const reachedStreamingThreshold =
!this.isRealtimeMode &&
this.audioBuffer.isPCM &&
this.audioBuffer.totalPCMBytes >= MIN_STREAMING_SEGMENT_BYTES;