inform on max configuration

README.md

@@ -98,16 +98,46 @@ tasks:
 
 ## Configuration
 
-Create `.ralpi/config.yaml`:
+Create config files. Both are optional:
+
+| Scope | Path |
+|-------|------|
+| **Global** | `~/.pi/ralpi/config.yaml` |
+| **Project** | `./.ralpi/config.yaml` |
 
 ```yaml
-maxRetries: 3
+execution:
-retryDelayMs: 5000
+  maxParallel: 3          # ralpi-level concurrency only
-timeoutMs: 1800000
+  models:                 # round-robin in <provider>/<model> format
-maxParallel: 3
+    - google/gemini-3.5-flash # 1st and 3rd task in parallel
+    - openai/gpt-5.5 # 2nd task in parallel
+prompts:
   projectContext: "Additional context for all tasks"
 ```
 
+> ralpi deliberately does **not** set timeouts or retries — those are inherited
+> from Pi's own settings. Tasks run until they complete or Pi's own flow stops them.
+>
+> `execution.models` uses slot-aware round-robin: with 3 models and 2 concurrent
+> tasks, only the first two models are used. The third model is only touched when
+> a third concurrent task starts. Freed model slots are reused before new ones
+> are allocated.
+>
+> **Automatic failover**: if a provider/API is unreachable (rate limit, 503, etc.),
+> the task automatically cycles to the next model in the list without counting it
+> as a task failure. Each model is tried once before the task is marked as failed.
+
+The keys mirror the nested structure of `RalpiConfig` in `src/types.ts`.
+
+### Precedence (highest wins)
+
+| Priority | Source |
+|----------|--------|
+| **1st** | In-memory overrides (`model`, `thinkingLevel` from parent Pi session) |
+| **2nd** | `./.ralpi/config.yaml` — project-level |
+| **3rd** | `~/.pi/ralpi/config.yaml` — global, shared across projects |
+| **4th** | `DEFAULT_CONFIG` in `src/types.ts` |
+
 ### Task-Level Timeout
 
 You can set a timeout for individual tasks using a meta block in the task file:

index.ts

@@ -66,9 +66,10 @@ async function selectExecutionMode(
   ctx: ExtensionContext,
   project: import("./src/types").Project,
   taskFile: string,
+  config: import("./src/types").RalpiConfig,
 ): Promise<ExecutionMode> {
   const mode = await ctx.ui.select("Execution mode for this run?", [
-    "Parallel (where dependencies allow)",
+    `Parallel (where dependencies allow)-[${config.execution.maxParallel} max]`,
     "Sequential (one at a time)",
   ]);
   const isParallel = mode?.startsWith("Parallel") ?? false;
@@ -117,7 +118,7 @@ async function executePlanBatches(
   plan: ReturnType<typeof buildPlanByMode>,
   project: Parameters<typeof buildExecutionPlan>[0],
   taskFile: string,
-  config: import("./src/types").ralpiConfig,
+  config: import("./src/types").RalpiConfig,
   progress: ProgressTracker,
   ctx: ExtensionContext,
   mode: ExecutionMode,
@@ -220,6 +221,16 @@ export default function ralpiLoopExtension(pi: ExtensionAPI): void {
     },
   );
 
+  // Register the extension's prompts/ directory so Pi discovers @task-manager
+  pi.on("resources_discover", async (_event, _ctx) => {
+    const promptsDir = fs.existsSync(path.resolve(__dirname, "prompts"))
+      ? path.resolve(__dirname, "prompts")
+      : path.resolve(__dirname, "..", "prompts");
+    return {
+      promptPaths: [promptsDir],
+    };
+  });
+
   pi.registerCommand("ralpi", {
     description:
       "Execute tasks from a task file using DAG-based dependency resolution",
@@ -248,21 +259,45 @@ export default function ralpiLoopExtension(pi: ExtensionAPI): void {
         return handlePlan(ctx, parts);
       }
       if (looksLikePath(parts[0])) {
-        return handleRun(ctx, parts, sendProgress);
+        return handleRun(
+          ctx,
+          parts,
+          sendProgress,
+          ctx.model,
+          pi.getThinkingLevel(),
+        );
       }
 
       const command = parts[0];
       switch (command) {
         case "run":
-          return handleRun(ctx, parts.slice(1), sendProgress);
+          return handleRun(
+            ctx,
+            parts.slice(1),
+            sendProgress,
+            ctx.model,
+            pi.getThinkingLevel(),
+          );
         case "plan":
           return handlePlan(ctx, parts.slice(1));
         case "status":
           return handleStatus(ctx, parts.slice(1));
         case "resume":
-          return handleResume(ctx, parts.slice(1), sendProgress);
+          return handleResume(
+            ctx,
+            parts.slice(1),
+            sendProgress,
+            ctx.model,
+            pi.getThinkingLevel(),
+          );
         case "next":
-          return handleNext(ctx, parts.slice(1), sendProgress);
+          return handleNext(
+            ctx,
+            parts.slice(1),
+            sendProgress,
+            ctx.model,
+            pi.getThinkingLevel(),
+          );
         case "reset":
           return handleReset(ctx, parts.slice(1));
         default: {
@@ -316,6 +351,8 @@ async function handleRun(
   ctx: ExtensionContext,
   args: string[],
   sendChatMessage?: SendChatMessage,
+  parentModel?: unknown,
+  parentThinkingLevel?: unknown,
 ): Promise<void> {
   const taskFile = resolveTaskArg(args[0] || "README.md", process.cwd());
 
@@ -323,7 +360,13 @@ async function handleRun(
   // auto-resume instead of starting fresh
   const existingProgress = findProgressFile(process.cwd(), taskFile);
   if (existingProgress) {
-    return handleResume(ctx, args.slice(0, 1), sendChatMessage);
+    return handleResume(
+      ctx,
+      args.slice(0, 1),
+      sendChatMessage,
+      parentModel,
+      parentThinkingLevel,
+    );
   }
 
   // No existing progress for this task — check for any progress at all
@@ -336,7 +379,13 @@ async function handleRun(
     );
 
     if (shouldResume?.startsWith("Yes")) {
-      return handleResume(ctx, [], sendChatMessage);
+      return handleResume(
+        ctx,
+        [],
+        sendChatMessage,
+        parentModel,
+        parentThinkingLevel,
+      );
     }
   }
 
@@ -346,16 +395,12 @@ async function handleRun(
 
   const project = parseTaskFile(taskFile);
   const config = loadConfig(projectDir);
+  config.model = parentModel ?? ctx.model;
+  config.thinkingLevel = parentThinkingLevel;
   const progress = new ProgressTracker(projectDir, taskFile);
 
-  // Set initial status
-  ctx.ui.setStatus(
-    "ralpi",
-    `Starting ${project.tasks.length} tasks from ${path.basename(taskFile)}`,
-  );
-
   const completed = buildCompletedSet(progress, project);
-  const mode = await selectExecutionMode(ctx, project, taskFile);
+  const mode = await selectExecutionMode(ctx, project, taskFile, config);
   const plan = buildPlanByMode(mode, project, completed);
 
   // Show execution plan before starting so user can see batch breakdown
@@ -434,6 +479,8 @@ async function handleResume(
   ctx: ExtensionContext,
   args: string[],
   sendChatMessage?: SendChatMessage,
+  parentModel?: unknown,
+  parentThinkingLevel?: unknown,
 ): Promise<void> {
   let taskFile: string;
   let projectDir: string;
@@ -473,15 +520,14 @@ async function handleResume(
     );
   }
   const config = loadConfig(projectDir);
+  config.model = parentModel ?? ctx.model;
+  config.thinkingLevel = parentThinkingLevel;
   const progress = new ProgressTracker(projectDir, taskFile, found.prdKey);
 
   progress.setPaused(false);
 
-  // Set resume status
-  ctx.ui.setStatus("ralpi", `Resuming from ${path.basename(taskFile)}`);
-
   const completed = buildCompletedSet(progress, project);
-  const mode = await selectExecutionMode(ctx, project, taskFile);
+  const mode = await selectExecutionMode(ctx, project, taskFile, config);
   const plan = buildPlanByMode(mode, project, completed);
 
   await executePlanBatches(
@@ -505,6 +551,8 @@ async function handleNext(
   ctx: ExtensionContext,
   args: string[],
   sendChatMessage?: SendChatMessage,
+  parentModel?: unknown,
+  parentThinkingLevel?: unknown,
 ): Promise<void> {
   let taskFile: string;
   let projectDir: string;
@@ -540,6 +588,8 @@ async function handleNext(
     );
   }
   const config = loadConfig(projectDir);
+  config.model = parentModel ?? ctx.model;
+  config.thinkingLevel = parentThinkingLevel;
   const progress = new ProgressTracker(projectDir, taskFile, found?.prdKey);
 
   const completed = buildCompletedSet(progress, project);

package.json

@@ -1,7 +1,7 @@
 {
-  "name": "ralpi-loop",
+  "name": "ralpi",
-  "version": "1.0.0",
+  "version": "0.1.0",
-  "description": "Execute tasks from task files using DAG-based dependency resolution with persistent progress tracking",
+  "description": "Execute tasks from task files/PRD's using DAG-based dependency resolution with persistent progress tracking",
   "main": "dist/index.js",
   "keywords": [
     "pi-package",

src/executor.ts

@@ -5,7 +5,6 @@ import type { ProgressTracker } from "./progress";
 import type { ExtensionContext } from "@earendil-works/pi-coding-agent";
 import { buildTaskPrompt } from "./prompts";
 import { extractReflection } from "./reflection";
-import { WidgetBatcher } from "./widget-batcher";
 import {
 	runAgentSession,
 	writeFileSafe,
@@ -26,6 +25,78 @@ export interface ToolCallEntry {
 	label: string;
 }
 
+// ─── Widget Expand/Collapse ───────────────────────────────────────────────
+
+/** Max tool calls shown in a live widget before truncating. Widgets don't
+ *  support message-style Ctrl+O expansion (that's only for chat-history
+ *  messages rendered by registerMessageRenderer). */
+const MAX_COLLAPSED = 3;
+
+const SPINNER_FRAMES = ["⠋", "⠙", "⠹", "⠸", "⠼", "⠴", "⠦", "⠧", "⠇", "⠏"];
+
+// ─── Model Round-Robin ─────────────────────────────────────────────────────
+
+/**
+ * Round-robin model assignment with slot reuse.
+ *
+ * With models [A, B, C] and 2 concurrent tasks, only A and B are used.
+ * Model C is only touched when a third concurrent task starts.
+ * Freed slots are reused before new slots are allocated.
+ */
+class ModelRoundRobin {
+	private models: unknown[];
+	private freeSlots: number[];
+	private nextIndex = 0;
+	private assignments = new Map<string, number>();
+
+	constructor(models: unknown[]) {
+		this.models = models;
+		this.freeSlots = [];
+	}
+
+	get length(): number {
+		return this.models.length;
+	}
+
+	assign(taskId: string): unknown {
+		let index: number;
+		if (this.freeSlots.length > 0) {
+			// Reuse a freed model slot first
+			index = this.freeSlots.shift()!;
+		} else if (this.nextIndex < this.models.length) {
+			// Allocate a new slot
+			index = this.nextIndex++;
+		} else {
+			// All models in use — wrap around
+			index = this.nextIndex % this.models.length;
+			this.nextIndex++;
+		}
+		this.assignments.set(taskId, index);
+		return this.models[index];
+	}
+
+	release(taskId: string): void {
+		const index = this.assignments.get(taskId);
+		if (index !== undefined) {
+			this.freeSlots.push(index);
+			this.freeSlots.sort((a, b) => a - b);
+			this.assignments.delete(taskId);
+		}
+	}
+}
+
+/** Shared state for parallel-batch widget. Each running task writes its
+ *  tool calls and spinner frame; the batch widget reads them in task-ID order. */
+interface ParallelWidgetEntry {
+	taskHeader: string;
+	frameIndex: number;
+	done: boolean;
+	success: boolean;
+	toolCalls: ToolCallEntry[];
+}
+
+type ParallelWidgetState = Map<string, ParallelWidgetEntry>;
+
 // ─── Run Single Task ────────────────────────────────────────────────────────
 
 /**
@@ -40,7 +111,8 @@ export async function runTask(
 	ctx: ExtensionContext,
 	sendChatMessage?: SendChatMessage,
 	projectDir: string = project.sourceDir,
-  batcher?: WidgetBatcher,
+	parallelState?: ParallelWidgetState,
+	assignedModel?: unknown,
 ): Promise<{
 	success: boolean;
 	reflection?: Reflection;
@@ -68,59 +140,76 @@ export async function runTask(
 	const promptFile = path.join(ralpiDir, `prompt-${startMs}.md`);
 	writeFileSafe(promptFile, prompt);
 
-  // Footer shows just the task title (no batch prefix)
-  ctx.ui.setStatus("ralpi", task.title);
-
 	const taskHeader = `${task.id} · ${task.title}`;
 
-  // Live progress widget above the editor — animated spinner + tool call tree
+	// When running in parallel, all tasks share a single widget so ordering
-  // Using setWidget instead of setWorkingMessage because the working message area
+	// is deterministic (sorted by task ID). In sequential mode each task gets
-  // is only visible during parent agent streaming, not during extension command execution.
+	// its own widget.
-  // Widget key is unique per task so parallel tasks each get their own widget.
+	const isParallel = !!parallelState;
 	const widgetKey = `ralpi-task-${task.id}`;
-  const SPINNER_FRAMES = ["⠋", "⠙", "⠹", "⠸", "⠼", "⠴", "⠦", "⠧", "⠇", "⠏"];
 	let frameIndex = 0;
-  const theme = ctx.ui.theme;
-  const MAX_COLLAPSED = 3;
-
 	const toolCalls: ToolCallEntry[] = [];
+	let widgetTui: { requestRender(): void } | null = null;
 
-  const updateWidget = () => {
+	if (isParallel) {
-    const frame = theme.fg("accent", SPINNER_FRAMES[frameIndex]);
+		parallelState!.set(task.id, {
+			taskHeader,
+			frameIndex: 0,
+			done: false,
+			success: false,
+			toolCalls: [],
+		});
+	} else {
+		// Build widget lines from current state. Live widgets can't expand/collapse
+		// like chat messages, so we always truncate to MAX_COLLAPSED recent calls.
+		const buildLines = (t: typeof ctx.ui.theme): string[] => {
+			const frame = t.fg("accent", SPINNER_FRAMES[frameIndex]);
 			const lines = [`${frame} ${taskHeader}`];
 
 			if (toolCalls.length > 0) {
+				if (toolCalls.length <= MAX_COLLAPSED) {
+					for (let i = 0; i < toolCalls.length; i++) {
+						const entry = toolCalls[i];
+						const isLast = i === toolCalls.length - 1;
+						const branch = isLast ? "  └── " : "  ├── ";
+						const tag = t.fg("accent", `[${entry.name}]`);
+						lines.push(`${branch}${tag} ${entry.label}`);
+					}
+				} else {
 					const shown = toolCalls.slice(-MAX_COLLAPSED);
 					const remaining = toolCalls.length - shown.length;
-
+					lines.push(t.fg("dim", `  ├── …${remaining} earlier`));
-      if (remaining > 0) {
-        lines.push(theme.fg("dim", `  ├── ${remaining} more`));
-      }
-
 					for (let i = 0; i < shown.length; i++) {
 						const entry = shown[i];
 						const isLast = i === shown.length - 1;
 						const branch = isLast ? "  └── " : "  ├── ";
-        const tag = theme.fg("accent", `[${entry.name}]`);
+						const tag = t.fg("accent", `[${entry.name}]`);
 						lines.push(`${branch}${tag} ${entry.label}`);
 					}
 				}
-
-    if (batcher) {
-      batcher.schedule(widgetKey, lines);
-    } else {
-      ctx.ui.setWidget(widgetKey, lines);
 			}
+			return lines;
 		};
 
-  // Smooth spinner animation at 100ms intervals
+		ctx.ui.setWidget(widgetKey, (tui, t) => {
-  const spinnerTimer = setInterval(() => {
+			widgetTui = tui;
-    frameIndex = (frameIndex + 1) % SPINNER_FRAMES.length;
+			return {
-    updateWidget();
+				render: () => buildLines(t),
-  }, 100);
+				invalidate: () => widgetTui?.requestRender(),
+			};
+		});
+	}
 
-  // Initial display
+	const requestRender = () => widgetTui?.requestRender();
-  updateWidget();
+
+	// Spinner animation (sequential only — parallel uses a single batch timer)
+	let spinnerTimer: NodeJS.Timeout | undefined;
+	if (!isParallel) {
+		spinnerTimer = setInterval(() => {
+			frameIndex = (frameIndex + 1) % SPINNER_FRAMES.length;
+			requestRender();
+		}, 100);
+	}
 
 	// Use task-level timeout if set, otherwise fall back to config
 	const timeoutMs = task.timeoutMs ?? config.execution.timeoutMs;
@@ -142,23 +231,34 @@ export async function runTask(
 					name: event.toolName,
 					label,
 				});
-        updateWidget();
+				if (isParallel) {
+					const entry = parallelState!.get(task.id);
+					if (entry) {
+						entry.toolCalls.push({ name: event.toolName, label });
+					}
+				}
+				requestRender();
 			}
 		},
 		undefined, // no abort signal
 		sessionFilePath, // stream events to file
+		assignedModel ?? config.model,
+		config.thinkingLevel,
 	);
 
 	const durationMs = Date.now() - startMs;
 
 	// Clear progress widget and status after task finishes
-  clearInterval(spinnerTimer);
+	if (spinnerTimer) clearInterval(spinnerTimer);
-  if (batcher) {
+	if (isParallel) {
-    batcher.scheduleRemove(widgetKey);
+		const entry = parallelState!.get(task.id);
+		if (entry) {
+			entry.done = true;
+			entry.success = output.success;
+		}
 	} else {
 		ctx.ui.setWidget(widgetKey, undefined);
 	}
-  ctx.ui.setStatus("ralpi", undefined);
 
 	if (!output.success) {
 		sendChatMessage?.(`✗ ${taskHeader} — ${output.error}`);
@@ -227,6 +327,37 @@ export async function executeBatch(
 		);
 	}
 
+	// Set up model round-robin if configured.
+	// Config entries are "<provider>/<model>" strings — resolve via modelRegistry.
+	let roundRobin: ModelRoundRobin | null = null;
+	if (config.execution.models.length > 0) {
+		const resolvedModels: unknown[] = [];
+		for (const entry of config.execution.models) {
+			const slashIdx = entry.indexOf("/");
+			if (slashIdx === -1) {
+				ctx.ui.notify(
+					`ralpi config: skipping model "${entry}" — expected <provider>/<model> format`,
+					"warning",
+				);
+				continue;
+			}
+			const provider = entry.slice(0, slashIdx);
+			const modelId = entry.slice(slashIdx + 1);
+			const resolved = ctx.modelRegistry?.find(provider, modelId);
+			if (resolved) {
+				resolvedModels.push(resolved);
+			} else {
+				ctx.ui.notify(
+					`ralpi config: model "${entry}" not found in registry — skipping`,
+					"warning",
+				);
+			}
+		}
+		if (resolvedModels.length > 0) {
+			roundRobin = new ModelRoundRobin(resolvedModels);
+		}
+	}
+
 	// Check if we should run parallel
 	const shouldParallel =
 		options?.parallel && tasks.length > 1 && config.execution.maxParallel > 0;
@@ -240,12 +371,14 @@ export async function executeBatch(
 			ctx,
 			sendChatMessage,
 			projectDir,
+			roundRobin,
 		);
 		return;
 	}
 
 	// Execute sequentially
 	for (const task of tasks) {
+		const model = roundRobin?.assign(task.id);
 		await executeTask(
 			task,
 			project,
@@ -254,6 +387,9 @@ export async function executeBatch(
 			ctx,
 			sendChatMessage,
 			projectDir,
+			undefined,
+			model,
+			roundRobin,
 		);
 	}
 }
@@ -269,12 +405,76 @@ async function executeBatchParallel(
 	ctx: ExtensionContext,
 	sendChatMessage?: SendChatMessage,
 	projectDir?: string,
+	roundRobin?: ModelRoundRobin | null,
 ): Promise<void> {
 	const maxParallel = config.execution.maxParallel;
-  const batcher = new WidgetBatcher(ctx);
+	const sharedState: ParallelWidgetState = new Map();
+
+	// Register a single batch widget that renders ALL parallel tasks in ID order.
+	const widgetKey = `ralpi-parallel-${Date.now()}`;
+	let widgetTui: { requestRender(): void } | null = null;
+
+	const buildBatchLines = (t: typeof ctx.ui.theme): string[] => {
+		const lines: string[] = [];
+		const sortedIds = Array.from(sharedState.keys()).sort();
+
+		for (const id of sortedIds) {
+			const entry = sharedState.get(id)!;
+			const frame = entry.done
+				? entry.success
+					? "✓"
+					: "✗"
+				: t.fg("accent", SPINNER_FRAMES[entry.frameIndex]);
+			lines.push(`${frame} ${entry.taskHeader}`);
+
+			if (entry.toolCalls.length > 0) {
+				if (entry.toolCalls.length <= MAX_COLLAPSED) {
+					for (let i = 0; i < entry.toolCalls.length; i++) {
+						const tc = entry.toolCalls[i];
+						const isLast = i === entry.toolCalls.length - 1;
+						const branch = isLast ? "  └── " : "  ├── ";
+						const tag = t.fg("accent", `[${tc.name}]`);
+						lines.push(`${branch}${tag} ${tc.label}`);
+					}
+				} else {
+					const shown = entry.toolCalls.slice(-MAX_COLLAPSED);
+					const remaining = entry.toolCalls.length - shown.length;
+					lines.push(t.fg("dim", `  ├── …${remaining} earlier`));
+					for (let i = 0; i < shown.length; i++) {
+						const tc = shown[i];
+						const isLast = i === shown.length - 1;
+						const branch = isLast ? "  └── " : "  ├── ";
+						const tag = t.fg("accent", `[${tc.name}]`);
+						lines.push(`${branch}${tag} ${tc.label}`);
+					}
+				}
+			}
+		}
+		return lines;
+	};
+
+	ctx.ui.setWidget(widgetKey, (tui, t) => {
+		widgetTui = tui;
+		return {
+			render: () => buildBatchLines(t),
+			invalidate: () => widgetTui?.requestRender(),
+		};
+	});
+
+	// Single spinner timer drives all tasks in the batch
+	const spinnerTimer = setInterval(() => {
+		for (const entry of sharedState.values()) {
+			if (!entry.done) {
+				entry.frameIndex = (entry.frameIndex + 1) % SPINNER_FRAMES.length;
+			}
+		}
+		widgetTui?.requestRender();
+	}, 100);
+
 	const results: Array<{ task: Task; result: Promise<any> }> = [];
 
 	for (const task of tasks) {
+		const assignedModel = roundRobin?.assign(task.id);
 		results.push({
 			task,
 			result: executeTask(
@@ -285,7 +485,9 @@ async function executeBatchParallel(
 				ctx,
 				sendChatMessage,
 				projectDir,
-        batcher,
+				sharedState,
+				assignedModel,
+				roundRobin,
 			),
 		});
 
@@ -301,8 +503,8 @@ async function executeBatchParallel(
 		await result;
 	}
 
-  // Flush and stop the batcher after all tasks complete
+	clearInterval(spinnerTimer);
-  batcher.stop();
+	ctx.ui.setWidget(widgetKey, undefined);
 }
 
 // ─── Execute Single Task with Retry ──────────────────────────────────────────
@@ -315,11 +517,26 @@ async function executeTask(
 	ctx: ExtensionContext,
 	sendChatMessage?: SendChatMessage,
 	projectDir: string = project.sourceDir,
-  batcher?: WidgetBatcher,
+	parallelState?: ParallelWidgetState,
+	assignedModel?: unknown,
+	roundRobin?: ModelRoundRobin | null,
 ): Promise<void> {
 	const maxRetries = config.execution.maxRetries;
-  let retries = 0;
 
+	// Model failover: when a provider/API is down, cycle through available models.
+	// result.success === false always means an agent-session failure (API error,
+	// provider unreachable, etc.), not a task-work error.
+	const maxModelAttempts = roundRobin ? roundRobin.length : 1;
+	let modelAttempt = 0;
+	let currentModel: unknown = assignedModel ?? config.model;
+
+	while (modelAttempt < maxModelAttempts) {
+		// Get the next model from round-robin (on first try, use the pre-assigned model)
+		if (modelAttempt > 0 && roundRobin) {
+			currentModel = roundRobin.assign(task.id);
+		}
+
+		let retries = 0;
 		while (retries <= maxRetries) {
 			try {
 				// Mark as in progress
@@ -339,7 +556,8 @@ async function executeTask(
 					ctx,
 					sendChatMessage,
 					projectDir,
-        batcher,
+					parallelState,
+					currentModel,
 				);
 
 				if (result.success) {
@@ -359,10 +577,23 @@ async function executeTask(
 						result.commitMessages,
 						result.commitSummary,
 					);
+					roundRobin?.release(task.id);
 					return;
 				}
 
-      // Task failed, check if we should retry
+				// Agent session failed (provider error).
+				// If we have more models, cycle immediately — don't waste retries.
+				if (roundRobin && modelAttempt < maxModelAttempts - 1) {
+					roundRobin.release(task.id);
+					modelAttempt++;
+					ctx.ui.notify(
+						`Task ${task.id}: model failed, trying next (${modelAttempt + 1}/${maxModelAttempts}): ${result.error}`,
+						"warning",
+					);
+					break; // exit retry loop, cycle to next model
+				}
+
+				// No more models — use normal retry logic
 				if (retries < maxRetries) {
 					retries = progress.incrementRetry(task.id);
 					ctx.ui.notify(
@@ -379,11 +610,20 @@ async function executeTask(
 					throw new Error(`Task ${task.id} failed: ${result.error}`);
 				}
 			} catch (error) {
+				roundRobin?.release(task.id);
 				const errorMsg = error instanceof Error ? error.message : String(error);
 				progress.markFailed(task.id, errorMsg);
 				throw error;
 			}
 		}
+
+		// If we broke out (model cycling), continue the outer loop
+		modelAttempt++;
+	}
+
+	// All models exhausted
+	progress.markFailed(task.id, "All configured models exhausted");
+	throw new Error(`Task ${task.id} failed: all configured models exhausted`);
 }
 
 // ─── Save Reflection to File ────────────────────────────────────────────────

src/types.ts

@@ -153,6 +153,8 @@ export interface RalpiConfig {
 		timeoutMs: number;
 		/** Maximum parallel tasks (0 = unlimited) */
 		maxParallel: number;
+		/** Round-robin model list for parallel tasks (empty = inherit parent model) */
+		models: string[];
 	};
 	prompts: {
 		/** Additional context injected into every task prompt */
@@ -160,6 +162,10 @@ export interface RalpiConfig {
 		/** Custom prompt suffix for reflection extraction */
 		reflectionPrompt: string;
 	};
+	/** Parent session model to inherit in child agent sessions */
+	model?: unknown;
+	/** Parent session thinking level to inherit in child agent sessions */
+	thinkingLevel?: unknown;
 }
 
 export const DEFAULT_CONFIG: RalpiConfig = {
@@ -168,10 +174,11 @@ export const DEFAULT_CONFIG: RalpiConfig = {
 		reflectionsDir: ".ralpi/reflections",
 	},
 	execution: {
-		maxRetries: 3,
+		maxRetries: 0,
-		retryDelayMs: 5000,
+		retryDelayMs: 0,
-		timeoutMs: 30 * 60 * 1000, // 30 minutes
+		timeoutMs: 0, // 0 = inherit Pi's own defaults (no ralpi-level timeout)
 		maxParallel: 3,
+		models: [],
 	},
 	prompts: {
 		projectContext: "",

src/utils.ts

@@ -82,32 +82,35 @@ export function findProgressFile(
 
 // ─── Config ──────────────────────────────────────────────────────────────────
 
-function parseSimpleYaml(content: string): Record<string, any> {
+/** Try to use the `yaml` package (real dependency in package.json).
+ *  Falls back to a flat key:value parser when unavailable. */
+const parseSimpleYaml: (content: string) => Record<string, any> = (() => {
+	try {
+		// eslint-disable-next-line @typescript-eslint/no-var-requires
+		const { parse } = require("yaml");
+		return (content: string) => parse(content) ?? {};
+	} catch {
+		return (content: string) => {
 			const result: Record<string, any> = {};
-	const lines = content.split("\n");
+			for (const line of content.split("\n")) {
-
-	for (const line of lines) {
 				const trimmed = line.trim();
 				if (!trimmed || trimmed.startsWith("#")) continue;
-
+				const match = trimmed.match(/^([^:]+):\s*(.*)$/);
-		const match = trimmed.match(/^([^:]+):\s*(.+)$/);
 				if (match) {
-			const key = match[1].trim();
+					const value = match[2].trim();
-			let value: string | boolean | number = match[2].trim();
+					if (value === "true") result[match[1].trim()] = true;
-
+					else if (value === "false") result[match[1].trim()] = false;
-			// Parse booleans
+					else if (/^\d+$/.test(value))
-			if (value === "true") value = true;
+						result[match[1].trim()] = parseInt(value, 10);
-			else if (value === "false") value = false;
+					else if (/^\d+\.\d+$/.test(value))
-			// Parse numbers
+						result[match[1].trim()] = parseFloat(value);
-			else if (/^\d+$/.test(value)) value = parseInt(value, 10);
+					else result[match[1].trim()] = value;
-			else if (/^\d+\.\d+$/.test(value)) value = parseFloat(value);
-
-			result[key] = value;
 				}
 			}
-
 			return result;
+		};
 	}
+})();
 
 /**
  * Deep merge configuration objects
@@ -129,25 +132,44 @@ function mergeConfig(
 	return result as RalpiConfig;
 }
 
+/** Path to the global ralpi config under the user's Pi home directory. */
+const GLOBAL_CONFIG_PATH = path.join(
+	process.env.HOME || "/tmp",
+	".pi",
+	"ralpi",
+	"config.yaml",
+);
+
 /**
- * Load configuration from .ralpi/config.yaml or return defaults
+ * Load and merge config from global and project sources.
+ *
+ * Precedence (highest wins):
+ *   1. Project-level: `<projectDir>/.ralpi/config.yaml`
+ *   2. Global: `~/.pi/ralpi/config.yaml`
+ *   3. `DEFAULT_CONFIG` in `src/types.ts`
  */
 export function loadConfig(projectDir: string): RalpiConfig {
-	const configPath = path.join(projectDir, ".ralpi", "config.yaml");
+	// Start with defaults
+	const merged: RalpiConfig = { ...DEFAULT_CONFIG };
 
-	// Return defaults silently when config file does not exist
+	// Layer 1: global config (~/.pi/ralpi/config.yaml)
-	if (!fs.existsSync(configPath)) {
+	tryLoadConfigFile(GLOBAL_CONFIG_PATH, merged);
-		return { ...DEFAULT_CONFIG };
-	}
 
+	// Layer 2: project config (.ralpi/config.yaml) — overrides global
+	tryLoadConfigFile(path.join(projectDir, ".ralpi", "config.yaml"), merged);
+
+	return merged;
+
+	/** Attempt to load a single config file and merge into `acc` in place. */
+	function tryLoadConfigFile(filePath: string, acc: RalpiConfig): void {
+		if (!fs.existsSync(filePath)) return;
 		try {
-		const content = fs.readFileSync(configPath, "utf-8");
+			const content = fs.readFileSync(filePath, "utf-8");
-		// Simple YAML parsing (key: value format)
+			const parsed = parseSimpleYaml(content);
-		const config = parseSimpleYaml(content);
+			Object.assign(acc, mergeConfig(acc, parsed));
-		return mergeConfig(DEFAULT_CONFIG, config);
 		} catch {
-		// Malformed config — fall back to defaults silently
+			// Malformed config — skip silently
-		return { ...DEFAULT_CONFIG };
+		}
 	}
 }
 
@@ -339,6 +361,8 @@ export async function runAgentSession(
 	onEvent?: (event: AgentSessionEvent) => void,
 	signal?: AbortSignal,
 	sessionFile?: string,
+	model?: unknown,
+	thinkingLevel?: unknown,
 ): Promise<{
 	success: boolean;
 	text: string;
@@ -361,10 +385,13 @@ export async function runAgentSession(
 		? fs.createWriteStream(sessionFile, { flags: "a" })
 		: null;
 
-	// Wire timeout via abort signal
+	// Wire timeout via abort signal (only when set; 0 means inherit Pi's defaults)
-	const timeoutHandle = setTimeout(() => {
+	let timeoutHandle: NodeJS.Timeout | null = null;
+	if (timeoutMs > 0) {
+		timeoutHandle = setTimeout(() => {
 			if (sessionRef?.session) sessionRef.session.agent.abort();
 		}, timeoutMs);
+	}
 
 	const sessionRef: {
 		session?: Awaited<ReturnType<typeof createAgentSession>>["session"];
@@ -387,6 +414,8 @@ export async function runAgentSession(
 			sessionManager: SessionManager.inMemory(),
 			resourceLoader: loader,
 			tools: ["read", "bash", "edit", "write", "grep", "find", "ls"],
+			model: model as any,
+			thinkingLevel: thinkingLevel as any,
 		});
 		sessionRef.session = result.session;
 
@@ -437,7 +466,7 @@ export async function runAgentSession(
 		unsubscribe();
 		result.session.dispose();
 		signal?.removeEventListener("abort", abortHandler);
-		clearTimeout(timeoutHandle);
+		if (timeoutHandle) clearTimeout(timeoutHandle);
 
 		// Flush and close the event stream before returning
 		if (eventStream) {
@@ -463,7 +492,7 @@ export async function runAgentSession(
 			events: [], // streamed to file
 		};
 	} catch (error) {
-		clearTimeout(timeoutHandle);
+		if (timeoutHandle) clearTimeout(timeoutHandle);
 		if (eventStream && !eventStream.destroyed) {
 			eventStream.end();
 		}