@arhen/pi-core-subagent 1.1.5 → 1.1.7

package/README.md

@@ -1,6 +1,6 @@
-# @arhen/pi-subagent
+# @arhen/pi-core-subagent
 
-[![npm version](https://img.shields.io/npm/v/%40arhen%2Fpi-subagent?color=cb3837&logo=npm)](https://www.npmjs.com/package/@arhen/pi-subagent)
+[![npm version](https://img.shields.io/npm/v/%40arhen%2Fpi-core-subagent?color=cb3837&logo=npm)](https://www.npmjs.com/package/@arhen/pi-core-subagent)
 [![license](https://img.shields.io/badge/license-MIT-blue.svg)](./LICENSE)
 [![pi extension](https://img.shields.io/badge/pi-extension-7c3aed)](https://github.com/earendil-works/pi)
 
@@ -10,7 +10,7 @@ Built for one job: delegate work to isolated subagents **without bloating the pa
 
 ## Design principles
 
-- **No agent files required, no discovery overhead.** The leader defines every subagent inline per call — name, system prompt, toolset. Existing files (`.agents/`, `.pi/agents/`, `~/.pi/agent/agents/`) are *looked up by name only when no inline prompt is given*; nothing is ever created or scanned per request.
+- **No agent files, no discovery.** The leader defines every subagent inline per call — name, system prompt, toolset. Nothing is read from or written to disk.
 - **Two toolsets only.** Read-only (`read, grep, find, ls` — default) or write (`read, grep, find, ls, bash, edit, write` — `write: true`). No per-agent tool config surface.
 - **In-process** — children are `AgentSession`s in the same runtime. No process spawn, no context bleed.
 - **Zero parent-context injection.** No catalog, no context hook. 6 slim tools total.
@@ -20,13 +20,13 @@ Built for one job: delegate work to isolated subagents **without bloating the pa
 ## Install
 
 ```sh
-pi install npm:@arhen/pi-subagent
+pi install npm:@arhen/pi-core-subagent
 # or locally: pi install /path/to/pi-subagents
 ```
 
 ## Usage — the leader invents the agents
 
-Define agents inline (never creates files). If a task has **no inline `prompt`**, the extension looks up an existing file by agent name — `.agents/<name>.md` or `.pi/agents/<name>.md` (nearest project dir), then `~/.pi/agent/agents/<name>.md` — and inherits its prompt + tools. Inline params always win over the file. Model resolution: explicit `provider/model-id` (or bare id) via the pi model registry → agent-file `model` → the parent's current model → settings default.
+Define agents inline per call — never creates or reads agent files. Model resolution: explicit `provider/model-id` (or bare id) via the pi model registry → agent-file `model` → the parent's current model → settings default.
 
 ```json
 {
@@ -84,7 +84,7 @@ Background + intercom:
 
 ### Per-task fields
 
-`agent` (name you invent, or an existing file's name — required), `task` (required), `prompt` (system prompt, optional — falls back to file lookup, then a minimal default), `write` (toolset, default read-only), plus optional `model` (`provider/model-id`), `thinking` (validated enum: `off|minimal|low|medium|high|xhigh|max`), `tools` (explicit allowlist, overrides everything), `maxRuntimeMs`, `id`. Top-level only: `background`, `notifyPerTask`, `allowIntercom`, `concurrency`.
+`agent` (name you invent — required), `task` (required), `prompt` (system prompt, optional — minimal default used), `write` (toolset, default read-only), plus optional `model` (`provider/model-id`), `thinking` (validated enum: `off|minimal|low|medium|high|xhigh|max`), `tools` (explicit allowlist), `cwd`, `maxRuntimeMs`, `id`. Top-level only: `background`, `notifyPerTask`, `allowIntercom`, `concurrency`.
 
 ### Child talk tools (when `allowIntercom: true`)
 
@@ -99,7 +99,7 @@ Background + intercom:
 
 - Parent tools: 6 schemas with short descriptions. **No catalog, no context hook** — nothing injected per request.
 - Background completion: 3-line notice. Full text only via `subagent_result`.
-- Children: isolated sessions; talk tools injected only when `allowIntercom`; each child's prompt states its own task id and its siblings' so mailbox addressing works.
+- Children: isolated sessions; talk tools injected only when `allowIntercom`; each child's prompt states its own task id and its siblings' so mailbox addressing works. Model resolution: explicit `provider/model-id` or bare id via the pi model registry → the parent's current model → settings default. Thinking levels validated against the resolved model's `thinkingLevelMap`.
 
 ## Development
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@arhen/pi-core-subagent",
-  "version": "1.1.5",
+  "version": "1.1.7",
   "type": "module",
   "description": "pi extension: fast in-process subagents with single/parallel/chain, background runs, intercom and agent-to-agent mailbox. Leader defines agents inline.",
   "license": "MIT",
@@ -36,7 +36,7 @@
   },
   "repository": {
     "type": "git",
-    "url": "git+https://github.com/arhen/pi-subagent.git"
+    "url": "git+https://github.com/arhen/pi-core-subagent.git"
   },
   "publishConfig": {
     "access": "public",

package/src/index.ts

@@ -98,7 +98,7 @@ interface RunSnapshot {
 	concurrency: number;
 	/** True once the parent awaited this run — completion notices are redundant then. */
 	awaited?: boolean;
-	/** Wake the parent (queued follow-up turn) as each task completes. Default true. */
+	/** Wake the parent (queued follow-up turn) as each task completes. Default false. */
 	notifyPerTask: boolean;
 	tasks: TaskSnapshot[];
 	aggregateUsage: UsageStats;
@@ -135,7 +135,9 @@ function aggregateUsage(tasks: TaskSnapshot[]): UsageStats {
 }
 function truncateText(text: string, max = FINAL_OUTPUT_CAP): string {
 	if (Buffer.byteLength(text, "utf8") <= max) return text;
-	return `${text.slice(0, max)}\n\n[Output truncated. Full child session is available in the session file.]`;
+	let out = text.slice(0, max);
+	while (Buffer.byteLength(out, "utf8") > max) out = out.slice(0, -1); // multibyte-safe
+	return `${out}\n\n[Output truncated. Full child session is available in the session file.]`;
 }
 function getFirstText(message: AssistantMessage): string {
 	for (const part of message?.content ?? []) {
@@ -396,7 +398,7 @@ class SubagentManager {
 	private liveChildren = new Map<string, { abort: () => void; dispose: () => void; touchWatchdog: () => void }>();
 	private mailboxes: Mailbox = createMailbox();
 	private runControllers = new Map<string, AbortController>();
-	private widgetTimer: ReturnType<typeof setTimeout> | undefined;
+	private widgetTimers = new Map<string, ReturnType<typeof setTimeout>>(); // per-run stream throttle
 	private widgetRuns: RunSnapshot[] = [];
 
 	turnActivity = false;
@@ -431,14 +433,19 @@ class SubagentManager {
 		return runId ? this.runs.get(runId) : undefined;
 	}
 	clearRuns(): void {
+		for (const child of this.liveChildren.values()) {
+			child.abort();
+			child.dispose();
+		}
+		this.liveChildren.clear();
 		this.runs.clear();
 		this.settlers.clear();
 		this.pendingReplies.clear();
 		this.runControllers.clear();
 		this.mailboxes = createMailbox();
 		this.widgetTui = null; // force re-registration on the next session
-		if (this.widgetTimer) clearTimeout(this.widgetTimer);
-		this.widgetTimer = undefined;
+		for (const t of this.widgetTimers.values()) clearTimeout(t);
+		this.widgetTimers.clear();
 		this.widgetRuns = [];
 	}
 
@@ -455,9 +462,17 @@ class SubagentManager {
 			if (!Array.isArray(raw)) return;
 			runs = (raw as RunSnapshot[]).map((run) => {
 				const interrupted = run.tasks.some((t) => !TERMINAL.includes(t.status));
+				// A persisted "running" run whose tasks are all terminal (crash between
+				// task end and run end) must not stay "running" forever.
+				let status = interrupted ? ("aborted" as RunStatus) : run.status;
+				if (!TERMINAL.includes(status)) {
+					const anyFailed = run.tasks.some((t) => t.status === "failed");
+					const anyAborted = run.tasks.some((t) => t.status === "aborted");
+					status = anyFailed ? "failed" : anyAborted ? "aborted" : "completed";
+				}
 				return {
 					...run,
-					status: interrupted ? ("aborted" as RunStatus) : run.status,
+					status,
 					endedAt: interrupted ? Date.now() : run.endedAt,
 					tasks: run.tasks.map((t) => (TERMINAL.includes(t.status) ? t : { ...t, status: "aborted" as TaskStatus, error: t.error || "Interrupted by session reload" })),
 				};
@@ -481,7 +496,7 @@ class SubagentManager {
 			const parentFile = getParentSessionFile(ctx);
 			if (!parentFile) return;
 			const sidecar = parentFile.replace(/\.jsonl$/, ".subagents.json");
-			import("fs").then(({ writeFileSync }) => writeFileSync(sidecar, JSON.stringify(this.listRuns().map(cloneRun), null, 2)));
+			import("fs").then(({ writeFileSync }) => writeFileSync(sidecar, JSON.stringify(this.listRuns().slice(0, 50).map(cloneRun), null, 2)));
 		} catch {
 			/* ignore */
 		}
@@ -531,22 +546,23 @@ class SubagentManager {
 	}
 	private scheduleWidget(run: RunSnapshot | undefined, ctx?: ExtensionContext, onUpdate?: (partial: any) => void): void {
 		this.upsertWidgetRun(run);
-		if (this.widgetTimer || this.widgetRuns.length === 0) return;
-		this.widgetTimer = setTimeout(() => {
-			this.widgetTimer = undefined;
-			if (this.widgetRuns.length === 0) return;
-			const targets = [...this.widgetRuns]; // read at fire: never render stale runs
+		if (!run || this.widgetTimers.has(run.id)) return;
+		this.widgetTimers.set(run.id, setTimeout(() => {
+			this.widgetTimers.delete(run.id);
 			if (ctx?.hasUI) {
 				this.ensureWidget(ctx);
 				this.widgetTui?.requestRender();
 			}
-			onUpdate?.({ content: [{ type: "text", text: compactLines(run ?? targets[0]!).join("\n") }] });
-		}, WIDGET_THROTTLE_MS);
+			onUpdate?.({ content: [{ type: "text", text: compactLines(run).join("\n") }] });
+		}, WIDGET_THROTTLE_MS));
 	}
 	private flushWidget(run: RunSnapshot | undefined, ctx?: ExtensionContext, onUpdate?: (partial: any) => void): void {
-		if (this.widgetTimer) {
-			clearTimeout(this.widgetTimer);
-			this.widgetTimer = undefined;
+		if (run) {
+			const t = this.widgetTimers.get(run.id);
+			if (t) {
+				clearTimeout(t);
+				this.widgetTimers.delete(run.id);
+			}
 		}
 		if (!run || this.widgetRuns.length === 0) return;
 		if (ctx?.hasUI) {
@@ -586,6 +602,13 @@ class SubagentManager {
 				this.updateTask(run, task, { status: "awaiting_parent" }, ctx);
 				this.liveChildren.get(`${run.id}:${task.id}`)?.touchWatchdog();
 				this.notifyParent(run, "asked", { taskId: task.id, question });
+				// A blocking run's parent can't reply mid-tool (followUp only fires after the
+				// tool returns) — only background runs can truly wait for the answer.
+				if (!run.background) {
+					this.updateTask(run, task, { status: "running" }, ctx);
+					this.liveChildren.get(`${run.id}:${task.id}`)?.touchWatchdog();
+					return "Parent cannot answer while this run is blocking. Continue autonomously with your best judgment.";
+				}
 				const reply = await this.awaitParentReply(run.id, task.id);
 				this.updateTask(run, task, { status: "running" }, ctx);
 				this.liveChildren.get(`${run.id}:${task.id}`)?.touchWatchdog();
@@ -593,10 +616,24 @@ class SubagentManager {
 			},
 			onNotifyParent: (_taskId, message, level) => {
 				this.emit("subagent:intercom", { runId: run.id, taskId: task.id, kind: "notify", level, message });
+				if (!run.awaited) {
+					try {
+						this.pi.sendUserMessage(`[Subagent ${task.agent}] ${message}`, { deliverAs: "followUp" });
+					} catch {
+						/* parent mid-stream */
+					}
+				}
 			},
 			onSendMessage: (_taskId, to, text) => {
 				if (to === "leader") {
 					this.emit("subagent:intercom", { runId: run.id, taskId: task.id, kind: "notify", level: "info", message: text });
+					if (!run.awaited) {
+						try {
+							this.pi.sendUserMessage(`[Subagent ${task.agent}] ${text}`, { deliverAs: "followUp" });
+						} catch {
+							/* parent mid-stream */
+						}
+					}
 					return true;
 				}
 				// Run-scoped keys: sibling ids are run-local; cross-run task_1 can never collide.
@@ -706,7 +743,7 @@ class SubagentManager {
 			});
 
 			unsubscribe = child.subscribe((event: AgentSessionEvent) => {
-				const active = event.type === "message_update" || event.type === "message_end" || event.type === "tool_execution_start" || event.type === "tool_execution_end" || event.type === "agent_settled";
+				const active = event.type === "message_update" || event.type === "message_end" || event.type === "tool_execution_start" || event.type === "tool_execution_update" || event.type === "tool_execution_end" || event.type === "bash_execution_update" || event.type === "agent_settled";
 				if (active) {
 					watchdog.touch();
 					this.emit("subagent:session-event", { runId: run.id, taskId: task.id, seq: eventSeq++, event: { type: event.type } });
@@ -745,7 +782,10 @@ class SubagentManager {
 				}
 			});
 
-			const abortChild = () => void child?.abort();
+			const abortChild = () => {
+				void child?.abort();
+				this.runControllers.get(run.id)?.abort(); // parent abort kills ALL siblings, not just this child
+			};
 			const runController = this.runControllers.get(run.id);
 			if (signal) signal.addEventListener("abort", abortChild, { once: true });
 			if (runController) runController.signal.addEventListener("abort", abortChild, { once: true });
@@ -820,6 +860,13 @@ class SubagentManager {
 				? params.tasks!
 				: params.chain!;
 		if (inputs.length > MAX_TASKS) throw new Error(`Too many subagent tasks (${inputs.length}). Max is ${MAX_TASKS}.`);
+		const ids = new Set<string>();
+		for (const input of inputs) {
+			if (input.id !== undefined) {
+				if (ids.has(input.id)) throw new Error(`Duplicate task id: ${input.id}`);
+				ids.add(input.id);
+			}
+		}
 
 		const run: RunSnapshot = {
 			id: newId("run"),
@@ -879,7 +926,7 @@ class SubagentManager {
 				const input = { ...inputs[i]!, task: next };
 				task.task = input.task;
 				await this.runChild(run, task, input, ctx, signal, onUpdate);
-				if (run.notifyPerTask && TERMINAL.includes(task.status)) {
+				if (run.notifyPerTask && run.background && TERMINAL.includes(task.status)) {
 					this.notifyTask(run, task, task.status as "completed" | "failed" | "aborted");
 				}
 				if (task.status !== "completed") break;
@@ -889,7 +936,7 @@ class SubagentManager {
 			await mapWithConcurrency(run.tasks, run.mode === "single" ? 1 : run.concurrency, async (task) => {
 				const index = run.tasks.indexOf(task);
 				await this.runChild(run, task, inputs[index]!, ctx, signal, onUpdate);
-				if (run.notifyPerTask && TERMINAL.includes(task.status)) {
+				if (run.notifyPerTask && run.background && TERMINAL.includes(task.status)) {
 					this.notifyTask(run, task, task.status as "completed" | "failed" | "aborted");
 				}
 			});
@@ -939,7 +986,10 @@ class SubagentManager {
 				}
 				this.settleRun(run.id, run);
 				this.runControllers.delete(run.id);
+				for (const task of run.tasks) this.mailboxes.close(`${run.id}:${task.id}`);
+				this.emit("subagent:run-completed", { runId: run.id, status: "failed", run: cloneRun(run) });
 				this.notifyParent(run, "failed");
+				this.persist(ctx);
 			});
 		return { run: cloneRun(run), background: true };
 	}
@@ -947,6 +997,7 @@ class SubagentManager {
 	cancelRun(runId: string): { aborted: number } {
 		const run = this.runs.get(runId);
 		if (!run) return { aborted: 0 };
+		if (TERMINAL.includes(run.status)) return { aborted: 0 }; // never corrupt a finished run
 		let aborted = 0;
 		this.runControllers.get(runId)?.abort();
 		for (const [key, child] of this.liveChildren) {
@@ -957,7 +1008,7 @@ class SubagentManager {
 		for (const task of run.tasks) {
 			if (TERMINAL.includes(task.status)) continue;
 			task.status = "aborted";
-			task.error = task.error || "Canceled by subagent_cancel";
+			task.error = task.error || "Canceled by subagent_cancel"; // never overwrite a real error
 			task.endedAt = Date.now();
 			aborted += 1;
 		}
@@ -995,9 +1046,10 @@ class SubagentManager {
 		if (!timeoutMs) return settled;
 		return Promise.race([
 			settled,
-			new Promise<RunSnapshot | undefined>((resolve) =>
-				setTimeout(() => resolve(this.runs.get(runId) ? cloneRun(this.runs.get(runId)!) : undefined), timeoutMs),
-			),
+			new Promise<RunSnapshot | undefined>((resolve) => {
+				const timer = setTimeout(() => resolve(this.runs.get(runId) ? cloneRun(this.runs.get(runId)!) : undefined), timeoutMs);
+				settled.then(() => clearTimeout(timer));
+			}),
 		]);
 	}
 }
@@ -1082,8 +1134,7 @@ export default function (pi: ExtensionAPI) {
 				ctx.ui.notify("No subagent runs in this session.", "info");
 				return;
 			}
-			ctx.ui.setWidget("subagents", runs.flatMap((run) => compactLines(run).concat("")), { placement: "aboveEditor" });
-			ctx.ui.notify(`Showing ${runs.length} subagent run(s).`, "info");
+			ctx.ui.notify(runs.flatMap((run) => compactLines(run).concat("")).join("\n") || "No subagent runs in this session.", "info");
 		},
 	});