@gajae-code/coding-agent 0.7.0 → 0.7.2

package/src/notifications/index.ts

@@ -24,11 +24,13 @@ import * as fs from "node:fs";
 import * as os from "node:os";
 import * as path from "node:path";
 import { promisify } from "node:util";
+import type { ImageContent, TextContent } from "@gajae-code/ai";
 import { NotificationServer } from "@gajae-code/natives";
 import { logger } from "@gajae-code/utils";
 import { Settings } from "../config/settings";
 import type { ExtensionCommandContext, ExtensionContext, ExtensionFactory } from "../extensibility/extensions";
 import { registerAskAnswerSource } from "../tools/ask-answer-registry";
+import { registerTelegramFileSink } from "./attachment-registry";
 import {
 	getNotificationConfig,
 	isGloballyConfigured,
@@ -136,12 +138,19 @@ interface SessionRuntime {
 	pendingInteractive: Map<string, PendingInteractiveAsk>;
 	/** Deregisters this session's ask answer source. */
 	disposeAnswerSource: () => void;
+	/** Deregisters this session's Telegram file sink. */
+	disposeFileSink: () => void;
 	redact: boolean;
 	sessionTag: string;
 	/** Whether the agent loop is currently running (drives the typing indicator). */
 	busy: boolean;
 	/** Inbound Telegram update ids injected but not yet consumed by a turn. */
 	pendingInbound: Set<number>;
+	/** Latest assistant text of the in-flight turn (from message_update). */
+	currentTurnText?: string;
+	/** Assistant text already flushed before an ask this turn (turn-scoped dedupe
+	 * so turn_end does not re-emit the pre-ask lead-in). Reset each turn. */
+	preAskFlushedText?: string;
 }
 
 interface ResolvedSettings {
@@ -152,6 +161,16 @@ interface ResolvedSettings {
 
 const defaultConfig: NotificationConfig = {
 	enabled: false,
+	botToken: undefined,
+	chatId: undefined,
+	discord: {
+		botToken: undefined,
+		channelId: undefined,
+	},
+	slack: {
+		botToken: undefined,
+		channelId: undefined,
+	},
 	redact: false,
 	verbosity: "lean",
 	idleTimeoutMs: 60_000,
@@ -223,6 +242,56 @@ function mapAnswerToGate(
 	return { selected: [] };
 }
 
+/** Register the interactive `ask` answer source for a session (the ask tool
+ * races the local UI against a remote reply). Returns the deregister disposer. */
+function registerInteractiveAnswerSource(
+	id: string,
+	server: NotificationServer,
+	pendingInteractive: Map<string, PendingInteractiveAsk>,
+	redact: boolean,
+	tag: string,
+): () => void {
+	return registerAskAnswerSource(id, {
+		awaitAnswer(question, options, signal) {
+			if (signal?.aborted) return Promise.resolve(undefined);
+			const askId = `ask:${crypto.randomUUID()}`;
+			try {
+				server.registerAsk(
+					JSON.stringify(
+						notificationActionPayload(
+							{ id: askId, kind: "ask", sessionId: id, question, options },
+							{ redact, sessionTag: tag },
+						),
+					),
+					true,
+				);
+			} catch (e) {
+				logger.warn(`notifications: registerAsk failed: ${String(e)}`);
+				return Promise.resolve(undefined);
+			}
+			return new Promise<string | undefined>(resolve => {
+				pendingInteractive.set(askId, { resolve, options });
+				signal?.addEventListener("abort", () => {
+					if (!pendingInteractive.delete(askId)) return;
+					// Local UI answered: mark the remote action resolved-locally.
+					try {
+						server.resolveLocal(askId, undefined);
+					} catch {}
+					resolve(undefined);
+				});
+			});
+		},
+	});
+}
+
+/** Extract the session id from a `<timestamp>_<uuid>.jsonl` session file path. */
+function sessionIdFromFile(file: string | undefined): string | undefined {
+	if (!file) return undefined;
+	const base = path.basename(file).replace(/\.jsonl$/, "");
+	const underscore = base.indexOf("_");
+	return underscore >= 0 ? base.slice(underscore + 1) : undefined;
+}
+
 export const createNotificationsExtension: ExtensionFactory = api => {
 	const runtimes = new Map<string, SessionRuntime>();
 	const disabledSessions = new Set<string>();
@@ -234,6 +303,7 @@ export const createNotificationsExtension: ExtensionFactory = api => {
 		runtimes.delete(id);
 		try {
 			rt.disposeAnswerSource();
+			rt.disposeFileSink();
 		} catch {}
 		// Resolve any still-pending interactive asks so the ask tool is not left hanging.
 		for (const pending of rt.pendingInteractive.values()) pending.resolve(undefined);
@@ -266,6 +336,7 @@ export const createNotificationsExtension: ExtensionFactory = api => {
 		const pendingInteractive = new Map<string, PendingInteractiveAsk>();
 		const tag = sessionTag(id);
 		const redact = cfg.redact;
+		let runtime: SessionRuntime | undefined;
 
 		// The SDK can always answer now (interactive via the answer source, or the
 		// unattended gate), so the endpoint advertises a resolver.
@@ -312,23 +383,34 @@ export const createNotificationsExtension: ExtensionFactory = api => {
 		// thread (forwarded by the daemon over the WS, fail-closed at the daemon).
 		server.onInbound((err, inbound) => {
 			if (err || !inbound) return;
-			if (inbound.kind === "user_message" && inbound.text) {
+			if (inbound.kind === "user_message") {
 				// Inject as a user turn (steers/continues the agent; the resulting
 				// turn streams back via the turn_end handler even when not idle).
 				// Record the update id so it can be acked as "consumed" on the next
 				// turn_start, and steer (vs start a fresh turn) when already busy.
-				const rt = runtimes.get(id);
-				if (rt && typeof inbound.updateId === "number") rt.pendingInbound.add(inbound.updateId);
+				const text = inbound.text ?? "";
+				const images = inbound.images ?? [];
+				if (!text && images.length === 0) return;
+				if (runtime && typeof inbound.updateId === "number") runtime.pendingInbound.add(inbound.updateId);
+				const content: string | (TextContent | ImageContent)[] =
+					images.length > 0
+						? [
+								...(text ? [{ type: "text", text } as TextContent] : []),
+								...images.map(
+									img =>
+										({ type: "image", data: img.data, mimeType: img.mime ?? "image/jpeg" }) as ImageContent,
+								),
+							]
+						: text;
 				try {
-					api.sendUserMessage(inbound.text, rt?.busy ? { deliverAs: "steer" } : undefined);
+					api.sendUserMessage(content, runtime?.busy ? { deliverAs: "steer" } : undefined);
 				} catch (e) {
 					logger.warn(`notifications: sendUserMessage failed: ${String(e)}`);
 				}
 				return;
 			}
 			if (inbound.kind === "config_command") {
-				const rt = runtimes.get(id);
-				if (rt && typeof inbound.redact === "boolean") rt.redact = inbound.redact;
+				if (runtime && typeof inbound.redact === "boolean") runtime.redact = inbound.redact;
 			}
 		});
 
@@ -336,48 +418,37 @@ export const createNotificationsExtension: ExtensionFactory = api => {
 			const endpoint = await server.start();
 
 			// Interactive answer source: the ask tool races the local UI against this.
-			const disposeAnswerSource = registerAskAnswerSource(id, {
-				awaitAnswer(question, options, signal) {
-					if (signal?.aborted) return Promise.resolve(undefined);
-					const askId = `ask:${crypto.randomUUID()}`;
-					try {
-						server.registerAsk(
-							JSON.stringify(
-								notificationActionPayload(
-									{ id: askId, kind: "ask", sessionId: id, question, options },
-									{ redact, sessionTag: tag },
-								),
-							),
-							true,
-						);
-					} catch (e) {
-						logger.warn(`notifications: registerAsk failed: ${String(e)}`);
-						return Promise.resolve(undefined);
-					}
-					return new Promise<string | undefined>(resolve => {
-						pendingInteractive.set(askId, { resolve, options });
-						signal?.addEventListener("abort", () => {
-							if (!pendingInteractive.delete(askId)) return;
-							// Local UI answered: mark the remote action resolved-locally.
-							try {
-								server.resolveLocal(askId, undefined);
-							} catch {}
-							resolve(undefined);
-						});
-					});
-				},
+			const disposeAnswerSource = registerInteractiveAnswerSource(id, server, pendingInteractive, redact, tag);
+			const disposeFileSink = registerTelegramFileSink(id, async file => {
+				try {
+					const data = await fs.promises.readFile(file.path);
+					server.pushFrame(
+						JSON.stringify({
+							type: "file_attachment",
+							sessionId: id,
+							name: path.basename(file.path),
+							data: data.toString("base64"),
+							caption: file.caption,
+						}),
+					);
+					return { ok: true };
+				} catch (e) {
+					return { ok: false, error: e instanceof Error ? e.message : String(e) };
+				}
 			});
 
-			runtimes.set(id, {
+			runtime = {
 				server,
 				idleSeq: 0,
 				pendingInteractive,
 				disposeAnswerSource,
+				disposeFileSink,
 				redact,
 				sessionTag: tag,
 				busy: false,
 				pendingInbound: new Set<number>(),
-			});
+			};
+			runtimes.set(id, runtime);
 			logger.info(`notifications: serving session ${id} at ${endpoint.url} (unattended=${unattended})`);
 
 			if (settingsAvailable && settings && isGloballyConfigured(cfg)) {
@@ -507,6 +578,82 @@ export const createNotificationsExtension: ExtensionFactory = api => {
 		await startSession(ctx);
 	});
 
+	// A session id change within the same process needs reason-aware handling.
+	// `/new` and fork CONTINUE the same terminal thread (e.g. plan "approve and
+	// execute" clears into a fresh session), so re-key the existing runtime
+	// old→new WITHOUT recreating the NotificationServer: the server, its endpoint
+	// discovery file, and the daemon's forum topic are all keyed by the original
+	// session id and the daemon routes by socket, so the existing topic is reused
+	// and the next identity frame renames it in place instead of spawning a new
+	// thread. `resume`, by contrast, loads a DIFFERENT, already-persisted session
+	// that owns its own topic — tear the previous runtime down and start fresh
+	// under the resumed id so the daemon attaches to (or recreates) that
+	// session's own discovery + topic rather than hijacking this terminal's.
+	api.on("session_switch", async (event, ctx) => {
+		const newId = sessionId(ctx);
+		const prevId = sessionIdFromFile(event.previousSessionFile);
+		if (!prevId || prevId === newId) return;
+
+		if (event.reason === "resume") {
+			stopSession(prevId);
+			await startSession(ctx);
+			return;
+		}
+
+		// `/new` / fork: re-key in place and rename the existing topic.
+		if (disabledSessions.delete(prevId)) disabledSessions.add(newId);
+		const rt = runtimes.get(prevId);
+		if (!rt || runtimes.has(newId)) return;
+		runtimes.delete(prevId);
+		runtimes.set(newId, rt);
+		// Re-bind the interactive ask answer source: the ask tool resolves the
+		// source by the current session id, which just changed.
+		try {
+			rt.disposeAnswerSource();
+			rt.disposeFileSink();
+		} catch {}
+		rt.disposeAnswerSource = registerInteractiveAnswerSource(
+			newId,
+			rt.server,
+			rt.pendingInteractive,
+			rt.redact,
+			rt.sessionTag,
+		);
+		rt.disposeFileSink = registerTelegramFileSink(newId, async file => {
+			try {
+				const data = await fs.promises.readFile(file.path);
+				rt.server.pushFrame(
+					JSON.stringify({
+						type: "file_attachment",
+						sessionId: newId,
+						name: path.basename(file.path),
+						data: data.toString("base64"),
+						caption: file.caption,
+					}),
+				);
+				return { ok: true };
+			} catch (e) {
+				return { ok: false, error: e instanceof Error ? e.message : String(e) };
+			}
+		});
+		// Rename the existing topic now when the new session already has a name; a
+		// fresh unnamed session is renamed on its next agent_end re-assert, which
+		// avoids a transient rename to bare "repo/branch".
+		if (ctx.sessionManager.getSessionName()) {
+			try {
+				rt.server.pushFrame(
+					JSON.stringify({
+						type: "identity_header",
+						sessionId: newId,
+						...buildIdentity(ctx.cwd, ctx.sessionManager.getSessionName()),
+					}),
+				);
+			} catch (e) {
+				logger.warn(`notifications: identity_header (switch) failed: ${String(e)}`);
+			}
+		}
+	});
+
 	// Drive the live typing indicator: mark busy when the agent loop starts so
 	// the daemon shows "typing…" in the thread while the agent is thinking,
 	// before any turn output exists. Cleared on `agent_end` below.
@@ -620,18 +767,42 @@ export const createNotificationsExtension: ExtensionFactory = api => {
 	// Redaction suppresses streamed content (only the one-time identity header
 	// survives redaction). The daemon coalesces/throttles these via its shared
 	// rate-limit pool before sending to Telegram.
-	api.on("turn_end", (event, ctx) => {
-		const id = sessionId(ctx);
-		const rt = runtimes.get(id);
-		if (!rt) return;
-		if (rt.redact) return;
-		const text = summaryFromMessage(event.message, 3500);
-		if (!text) return;
+	// Push the in-flight turn's assistant text as a finalized turn_stream, deduped
+	// against what was already flushed for this turn (the pre-ask lead-in).
+	const flushTurnText = (rt: SessionRuntime, id: string, text: string | undefined): void => {
+		if (!text || text === rt.preAskFlushedText) return;
+		rt.preAskFlushedText = text;
 		try {
 			rt.server.pushFrame(JSON.stringify({ type: "turn_stream", sessionId: id, phase: "finalized", text }));
 		} catch (e) {
 			logger.warn(`notifications: pushFrame (turn) failed: ${String(e)}`);
 		}
+	};
+
+	// Emit the assistant text that precedes an ask BEFORE the ask's action_needed
+	// is broadcast, so the remote (e.g. Telegram) shows the lead-in first instead
+	// of only after the ask resolves at turn_end. The text is captured on
+	// message_end (which, like tool_execution_start, is on the awaited extension
+	// path and ordered before it — unlike message_update, which is queued async),
+	// then flushed here before the ask tool's execute calls registerAsk.
+	api.on("tool_execution_start", (event, ctx) => {
+		if (event.toolName !== "ask") return;
+		const id = sessionId(ctx);
+		const rt = runtimes.get(id);
+		if (!rt || rt.redact) return;
+		flushTurnText(rt, id, rt.currentTurnText);
+	});
+
+	api.on("turn_end", (event, ctx) => {
+		const id = sessionId(ctx);
+		const rt = runtimes.get(id);
+		if (!rt) return;
+		const text = rt.redact ? undefined : summaryFromMessage(event.message, 3500);
+		if (text) flushTurnText(rt, id, text);
+		// Reset per-turn streaming state so the next turn starts fresh and a later
+		// turn with identical text is not falsely deduped.
+		rt.currentTurnText = undefined;
+		rt.preAskFlushedText = undefined;
 	});
 
 	// Stream agent-produced images (computer/browser/tool screenshots) as
@@ -640,6 +811,14 @@ export const createNotificationsExtension: ExtensionFactory = api => {
 		const id = sessionId(ctx);
 		const rt = runtimes.get(id);
 		if (!rt || rt.redact) return;
+		// Capture the in-flight ASSISTANT text here (message_end is on the awaited
+		// extension path and ordered before tool_execution_start) so the pre-ask
+		// flush can emit it before the ask prompt. Role-scoped: message_end also
+		// fires for the user prompt, which must never be mirrored back as turn output.
+		if ((event.message as { role?: unknown }).role === "assistant") {
+			const turnText = summaryFromMessage(event.message, 3500);
+			if (turnText) rt.currentTurnText = turnText;
+		}
 		for (const img of imageAttachmentsFromMessage(event.message)) {
 			try {
 				rt.server.pushFrame(

package/src/notifications/managed-daemon.ts

@@ -0,0 +1,163 @@
+import * as path from "node:path";
+import type { Settings } from "../config/settings";
+import { RateLimitPool } from "./rate-limit-pool";
+import {
+	CLIENT_PING_PONG_CAPABILITY,
+	HEARTBEAT_INTERVAL_MS,
+	HEARTBEAT_TTL_MS,
+	NOTIFICATION_PROTOCOL_VERSION,
+} from "./telegram-daemon";
+import { readEndpoint } from "./telegram-reference";
+import { renderThreadedFrame, type ThreadedSend } from "./threaded-render";
+
+export interface ManagedNotificationDaemonFs {
+	readdir(path: string): Promise<string[]>;
+}
+
+export interface ManagedSessionSocket {
+	sessionId: string;
+	token: string;
+	ws: WebSocket;
+	pending: Map<string, { sessionId: string; actionId: string }>;
+	capable: boolean;
+	lastPongAt: number;
+	awaitingNonce: string | undefined;
+	pingTimer: ReturnType<typeof setInterval> | undefined;
+}
+
+export interface ManagedNotificationDaemonOptions {
+	settings: Settings;
+	fs: ManagedNotificationDaemonFs;
+	WebSocketImpl?: typeof WebSocket;
+	now?: () => number;
+	setIntervalImpl?: typeof setInterval;
+	clearIntervalImpl?: typeof clearInterval;
+	rateLimitPool?: RateLimitPool<{ send: ThreadedSend; topicId?: string }>;
+}
+
+export abstract class ManagedNotificationDaemon {
+	readonly sessions = new Map<string, ManagedSessionSocket>();
+	readonly pool: RateLimitPool<{ send: ThreadedSend; topicId?: string }>;
+
+	protected constructor(protected readonly opts: ManagedNotificationDaemonOptions) {
+		this.pool = opts.rateLimitPool ?? new RateLimitPool<{ send: ThreadedSend; topicId?: string }>({ now: opts.now });
+	}
+
+	async scanRoots(): Promise<void> {
+		const rootState = await this.readRoots();
+		for (const root of rootState) {
+			const dir = path.join(root, "notifications");
+			let files: string[];
+			try {
+				files = await this.opts.fs.readdir(dir);
+			} catch {
+				continue;
+			}
+			for (const file of files.filter(item => item.endsWith(".json"))) {
+				const sessionId = path.basename(file, ".json");
+				if (this.sessions.has(sessionId)) continue;
+				try {
+					const endpoint = readEndpoint(path.join(dir, file));
+					this.connectSession(sessionId, endpoint.url, endpoint.token);
+				} catch {}
+			}
+		}
+	}
+
+	connectSession(sessionId: string, url: string, token: string): ManagedSessionSocket {
+		const WS = this.opts.WebSocketImpl ?? WebSocket;
+		const ws = new WS(`${url}/?token=${encodeURIComponent(token)}`);
+		const session: ManagedSessionSocket = {
+			sessionId,
+			token,
+			ws,
+			pending: new Map(),
+			capable: false,
+			lastPongAt: 0,
+			awaitingNonce: undefined,
+			pingTimer: undefined,
+		};
+		this.sessions.set(sessionId, session);
+		ws.addEventListener("open", () => this.sendHello(session));
+		ws.addEventListener("message", ev => {
+			if (this.sessions.get(sessionId) !== session) return;
+			void this.handleSessionMessage(session, JSON.parse(String(ev.data))).catch(() => undefined);
+		});
+		ws.addEventListener("close", () => this.dropSession(session));
+		return session;
+	}
+
+	protected async handleSessionMessage(session: ManagedSessionSocket, msg: Record<string, unknown>): Promise<boolean> {
+		if (msg.type === "hello") {
+			const caps = Array.isArray(msg.capabilities) ? msg.capabilities : [];
+			if (caps.includes(CLIENT_PING_PONG_CAPABILITY)) {
+				session.capable = true;
+				this.startLiveness(session);
+			}
+			return true;
+		}
+		if (msg.type === "pong") {
+			if (typeof msg.nonce === "string" && msg.nonce === session.awaitingNonce) {
+				session.awaitingNonce = undefined;
+				session.lastPongAt = (this.opts.now ?? Date.now)();
+			}
+			return true;
+		}
+		return false;
+	}
+
+	protected renderFrame(frame: Record<string, unknown>): ThreadedSend | undefined {
+		return renderThreadedFrame(frame);
+	}
+
+	protected dropSession(session: ManagedSessionSocket): void {
+		const clearIntervalImpl = this.opts.clearIntervalImpl ?? clearInterval;
+		if (session.pingTimer) {
+			clearIntervalImpl(session.pingTimer);
+			session.pingTimer = undefined;
+		}
+		if (this.sessions.get(session.sessionId) === session) this.sessions.delete(session.sessionId);
+		if (session.ws.readyState !== WebSocket.CLOSED) {
+			try {
+				session.ws.close();
+			} catch {}
+		}
+	}
+
+	protected abstract readRoots(): Promise<string[]>;
+
+	private sendHello(session: ManagedSessionSocket): void {
+		if (session.ws.readyState !== WebSocket.OPEN) return;
+		try {
+			session.ws.send(
+				JSON.stringify({
+					type: "hello",
+					protocolVersion: NOTIFICATION_PROTOCOL_VERSION,
+					capabilities: [CLIENT_PING_PONG_CAPABILITY],
+				}),
+			);
+		} catch {}
+	}
+
+	private startLiveness(session: ManagedSessionSocket): void {
+		if (session.pingTimer) return;
+		const setIntervalImpl = this.opts.setIntervalImpl ?? setInterval;
+		const now = () => (this.opts.now ?? Date.now)();
+		session.lastPongAt = now();
+		session.pingTimer = setIntervalImpl(() => {
+			if (this.sessions.get(session.sessionId) !== session) return;
+			const t = now();
+			if (t - session.lastPongAt >= HEARTBEAT_TTL_MS) {
+				this.dropSession(session);
+				return;
+			}
+			if (session.ws.readyState === WebSocket.OPEN) {
+				const nonce = `${session.sessionId}:${t}:${Math.random().toString(36).slice(2)}`;
+				session.awaitingNonce = nonce;
+				try {
+					session.ws.send(JSON.stringify({ type: "ping", nonce }));
+				} catch {}
+			}
+		}, HEARTBEAT_INTERVAL_MS);
+	}
+}