pi-ui-extend 1.0.38 → 1.0.40

package/external/pi-tools-suite/src/async-subagents/async-subagents.sample.jsonc

@@ -234,6 +234,14 @@
       "description": "Use when the sub-agent should make or plan code changes for a feature, bug fix, or refactor.",
       "model": "openai-codex/gpt-5.6-sol",
       "fallbackModels": ["zai/glm-5.3"],
+      // Avoid Sol recursively doing routine implementation work for a Sol parent.
+      // Luna also escalates substantial implementation to Terra. modelByParent
+      // wins over preset role models, so this prevents both Luna -> Sol and
+      // Sol -> Sol for implement tasks under the built-in `gpt` preset.
+      "modelByParent": {
+        "openai-codex/gpt-5.6-luna*": { "model": "openai-codex/gpt-5.6-terra", "fallbackModels": ["zai/glm-5.3"] },
+        "openai-codex/gpt-5.6-sol*": { "model": "openai-codex/gpt-5.6-terra", "fallbackModels": ["zai/glm-5.3"] }
+      },
       "thinking": "high"
     },
 

package/external/pi-tools-suite/src/async-subagents/core/agent-strategy.ts

@@ -1,6 +1,6 @@
 import { isGptLikeModel } from "./ultrawork-auto.js";
 
-export type AgentStrategyName = "parallel-first" | "deep-work";
+export type AgentStrategyName = "parallel-first" | "deep-work" | "escalation-aware" | "cost-aware-orchestrator";
 
 export interface AgentStrategyOptions {
 	modelRef?: string;
@@ -27,13 +27,40 @@ Default: autonomous deep worker. Build context directly, make progress, edit, an
 For broad work, keep delegation explicit and bounded: focused review/research/tests/frontend/deep tracks, plus one oracle only for high-stakes uncertainty or final plan checks. Read compact results, decide in the parent session, and report only what matters. If compressing unfinished work, preserve active objective + next step via todo/DCP rules.
 </agent_strategy>`;
 
+const ESCALATION_AWARE_STRATEGY_PROMPT = `<agent_strategy name="escalation-aware">
+Execution hint for Pi, not a replacement for system/developer/user instructions.
+
+Default: self-sufficient with tiered escalation. Solve narrow, well-bounded work directly, but do not grind through a large or uncertain task at the current model tier when a stronger focused subagent is appropriate. Use todo for the plan and async subagents for escalation; read compact results first and keep the parent context lean.
+
+For a Luna parent, prefer Terra workers for substantial multi-file research, tests, or implementation, and escalate deep root-cause analysis, architecture/security review, high-risk decisions, or repeatedly failing complex work to Sol through the deep/review roles. For a Terra parent, handle routine research/tests/implementation directly; escalate deep root-cause analysis, architecture/security review, high-risk decisions, or stubborn complex failures to Sol through deep/review. Do not escalate merely because a plan has several steps, and do not delegate a tiny known-file edit or exact lookup.
+
+Keep user questions, plan/todo changes, integration decisions, and the final report in the parent. Independent read-only escalations may run in parallel; serialize overlapping edits unless scopes are clearly disjoint. When work is delegated, synchronize its todo lifecycle: mark it in progress, collect and verify the worker result, then complete/update it before moving on.
+</agent_strategy>`;
+
+const COST_AWARE_ORCHESTRATOR_STRATEGY_PROMPT = `<agent_strategy name="cost-aware-orchestrator">
+Execution hint for Pi, not a replacement for system/developer/user instructions.
+
+Default: cost-aware orchestration. You are an expensive parent model, so keep the parent session focused on planning, decisions, integration, verification, and the final user-facing answer. For non-trivial todo work, prefer focused async subagents for repo scanning, multi-file research, documentation, tests, frontend work, and implementation steps that would otherwise require several repository tool calls. Read compact subagent results first; inspect raw artifacts or redo work in the parent only when verification or uncertainty requires it.
+
+Keep user questions, todo/plan changes, architecture tradeoffs, cross-worker integration decisions, high-stakes review, and the final report in the parent. Do not delegate merely to avoid one cheap exact lookup or a tiny known-file edit. Independent read-only tracks may run in parallel; serialize overlapping edits unless scopes are clearly disjoint. When a todo item is delegated, keep its lifecycle synchronized: mark it in progress, collect and verify the worker result, then complete/update it before moving on.
+</agent_strategy>`;
+
 export function agentStrategyPrompt(options: AgentStrategyOptions = {}): string | undefined {
 	const env = options.env ?? process.env;
 	const override = strategyOverride(env);
 	if (override === "off") return undefined;
 	if (options.customPrompt && shouldSkipCustomPrompt(env)) return undefined;
 
-	const strategy = override ?? (isGptLikeModel(options.modelRef) ? "deep-work" : "parallel-first");
+	const strategy = override
+		?? (isExpensiveGptParent(options.modelRef)
+			? "cost-aware-orchestrator"
+			: isEscalationAwareGptParent(options.modelRef)
+				? "escalation-aware"
+			: isGptLikeModel(options.modelRef)
+				? "deep-work"
+				: "parallel-first");
+	if (strategy === "cost-aware-orchestrator") return COST_AWARE_ORCHESTRATOR_STRATEGY_PROMPT;
+	if (strategy === "escalation-aware") return ESCALATION_AWARE_STRATEGY_PROMPT;
 	return strategy === "deep-work" ? DEEP_WORK_STRATEGY_PROMPT : PARALLEL_FIRST_STRATEGY_PROMPT;
 }
 
@@ -49,10 +76,22 @@ function strategyOverride(env: NodeJS.ProcessEnv): AgentStrategyName | "off" | u
 	if (FALSE_ENV_PATTERN.test(value)) return "off";
 	if (value === "parallel-first") return "parallel-first";
 	if (value === "deep-work") return "deep-work";
+	if (value === "escalation" || value === "escalation-aware") return "escalation-aware";
+	if (value === "cost-aware" || value === "cost-aware-orchestrator" || value === "orchestrator") return "cost-aware-orchestrator";
 	if (TRUE_ENV_PATTERN.test(value)) return undefined;
 	return undefined;
 }
 
+function isExpensiveGptParent(modelRef: string | undefined): boolean {
+	if (!modelRef) return false;
+	return /(?:^|\/)gpt-5\.6-sol(?:$|[-.:])/i.test(modelRef.trim());
+}
+
+function isEscalationAwareGptParent(modelRef: string | undefined): boolean {
+	if (!modelRef) return false;
+	return /(?:^|\/)gpt-5\.6-(?:luna|terra)(?:$|[-.:])/i.test(modelRef.trim());
+}
+
 function shouldSkipCustomPrompt(env: NodeJS.ProcessEnv): boolean {
 	const raw = firstEnv(env, "PI_AGENT_STRATEGY_WITH_CUSTOM_PROMPT", "ASYNC_SUBAGENTS_AGENT_STRATEGY_WITH_CUSTOM_PROMPT");
 	return raw ? !TRUE_ENV_PATTERN.test(raw.trim()) : true;

package/external/pi-tools-suite/src/async-subagents/core/config.ts

@@ -224,6 +224,10 @@ const BUILTIN_CONFIG: SubagentConfig = {
 		},
 		implement: {
 			description: "Use when the sub-agent should make or plan code changes for a feature, bug fix, or refactor.",
+			modelByParent: {
+				"openai-codex/gpt-5.6-luna*": { model: "openai-codex/gpt-5.6-terra", fallbackModels: ["zai/glm-5.3"] },
+				"openai-codex/gpt-5.6-sol*": { model: "openai-codex/gpt-5.6-terra", fallbackModels: ["zai/glm-5.3"] },
+			},
 			thinking: "high",
 		},
 		tests: {
@@ -234,10 +238,18 @@ const BUILTIN_CONFIG: SubagentConfig = {
 		},
 		review: {
 			description: "Use for review/audit of existing code or changes: correctness, security, performance, maintainability, API risks, quality. Do not implement new code.",
+			modelByParent: {
+				"openai-codex/gpt-5.6-luna*": { model: "openai-codex/gpt-5.6-sol", fallbackModels: ["zai/glm-5.3"] },
+				"openai-codex/gpt-5.6-terra*": { model: "openai-codex/gpt-5.6-sol", fallbackModels: ["zai/glm-5.3"] },
+			},
 			thinking: "high",
 		},
 		deep: {
 			description: "Use for broad hard reasoning: architecture, root-cause analysis, cross-module impact, complex debugging or tradeoffs.",
+			modelByParent: {
+				"openai-codex/gpt-5.6-luna*": { model: "openai-codex/gpt-5.6-sol", fallbackModels: ["zai/glm-5.3"] },
+				"openai-codex/gpt-5.6-terra*": { model: "openai-codex/gpt-5.6-sol", fallbackModels: ["zai/glm-5.3"] },
+			},
 			thinking: "high",
 		},
 		oracle: {

package/external/pi-tools-suite/src/async-subagents/index.ts

@@ -32,6 +32,7 @@ import { registerSubagentsTool } from "./tools/subagents.js";
 import type { LiveAgent, SubagentsLiveStateEvent } from "./types.js";
 import type { AgentState } from "./core/types.js";
 import { publishStartupSection } from "../startup-section.js";
+import { publishRpcSessionState } from "../lib/rpc-session-state.js";
 
 function isTerminalAgentStatus(status: AgentState["status"]): boolean {
 	return status === "done" || status === "failed" || status === "stopped";
@@ -83,8 +84,8 @@ function createLiveStatePayload(
 }
 
 function agentMatchesSession(agent: LiveAgent, sessionFile: string | undefined): boolean {
-	if (!sessionFile || !agent.parentSession) return true;
-	return pathsEqual(sessionFile, agent.parentSession);
+	if (!sessionFile) return true;
+	return agent.parentSession !== undefined && pathsEqual(sessionFile, agent.parentSession);
 }
 
 function isStaleExtensionContextError(error: unknown): boolean {
@@ -100,6 +101,7 @@ export default function (pi: ExtensionAPI) {
 	const subagentOverlay = new SubagentOverlay(liveAgents);
 	let sawAutoUltraworkCandidate = false;
 	let currentSessionFile: string | undefined;
+	let currentSessionStateContext: Parameters<typeof publishRpcSessionState>[0];
 	let completionWatchTimer: ReturnType<typeof setInterval> | undefined;
 	publishSubagentPresetsStartupSection();
 
@@ -109,6 +111,7 @@ export default function (pi: ExtensionAPI) {
 			const liveState = createLiveStatePayload(liveAgents, currentSessionFile);
 			pi.events?.emit?.(SUBAGENTS_LIVE_COUNT_EVENT, { count: liveState.count });
 			pi.events?.emit?.(SUBAGENTS_LIVE_STATE_EVENT, liveState);
+			publishRpcSessionState(currentSessionStateContext, SUBAGENTS_LIVE_STATE_EVENT, liveState);
 			updateCompletionWatcher();
 		} catch (error) {
 			ignoreStaleExtensionContextError(error);
@@ -168,6 +171,7 @@ export default function (pi: ExtensionAPI) {
 		try {
 			sawAutoUltraworkCandidate = false;
 			currentSessionFile = sessionFileFromContext(ctx);
+			currentSessionStateContext = ctx;
 			subagentOverlay.restoreRunningAgents(ctx.cwd, currentSessionFile);
 			refreshSubagentOverlay();
 		} catch (error) {

package/external/pi-tools-suite/src/async-subagents/subagent-overlay.ts

@@ -23,7 +23,7 @@ export class SubagentOverlay {
 				if (liveRun.has(agent.id)) continue;
 				const agentDir = path.join(runDir, agent.id);
 				const agentParentSession = readParentSessionLink(agentDir);
-				if (parentSession && agentParentSession && !pathsEqual(parentSession, agentParentSession)) continue;
+				if (parentSession && (!agentParentSession || !pathsEqual(parentSession, agentParentSession))) continue;
 				liveRun.set(agent.id, { runDir, agentId: agent.id, parentSession: agentParentSession, completed: Promise.resolve() });
 			}
 			if (liveRun.size === 0) this.liveAgents.delete(runDir);

package/external/pi-tools-suite/src/coding-discipline/index.ts

@@ -3,7 +3,7 @@ import * as path from "node:path";
 import type { Api, AssistantMessage, ImageContent, Model, ProviderHeaders, TextContent } from "@earendil-works/pi-ai";
 import { Type } from "typebox";
 
-import { loadPiToolsSuiteConfig, DEFAULT_CODING_DISCIPLINE_STRICTNESS, type CodingDisciplineStrictness } from "../config.js";
+import { loadPiToolsSuiteConfig } from "../config.js";
 import { ignoreStaleExtensionContextError } from "../context-usage.js";
 import { completeWithModelRegistry, type ModelCompletionRegistry } from "../model-completion.js";
 
@@ -41,13 +41,6 @@ const DEFAULT_LOOKUP_MAX_IMAGES = 6;
 const DEFAULT_LOOKUP_MAX_TOKENS = 1_600;
 const DEFAULT_LOOKUP_TIMEOUT_MS = 120_000;
 const MAX_IMAGE_BYTES = 16 * 1024 * 1024;
-// Keep recovery nudges sparse: newer GLM coding models are substantially better at
-// long-horizon tool use, and repeated developer reminders can become prompt noise.
-const SILENCE_REMINDER_MIN_VIOLATION_GAP = 5;
-const SILENCE_REMINDER_MIN_MESSAGE_GAP = 20;
-// When the visible history shrinks by more than this many messages (compaction or
-// truncation), the chatter baseline is reset so it isn't measured against a stale peak.
-const SILENCE_REMINDER_COMPACTION_MARGIN = 8;
 const LOOKUP_TOOL_NAME = "lookup";
 
 const LOOKUP_TOOL_PARAMS = Type.Object(
@@ -65,48 +58,46 @@ const LOOKUP_TOOL_PARAMS = Type.Object(
 	{ additionalProperties: false },
 );
 
-type DisciplinePromptOptions = { lookupEnabled?: boolean; strictness?: CodingDisciplineStrictness };
+type DisciplinePromptOptions = { lookupEnabled?: boolean };
 
-/**
- * Builds the tool-only contract lines. The working-state rule is strictness-aware:
- *   - "strict"  — one tool call per turn, no assistant text at all (Opus-like);
- *   - "lenient" — independent (especially read-only) calls may be batched, and brief
- *                 substantive reasoning text is acceptable when thinking is off.
- */
-function buildDisciplineLines(strictness: CodingDisciplineStrictness): readonly string[] {
-	const isStrict = strictness === "strict";
-	return [
-		"TOOL-ONLY CODING AGENT CONTRACT.",
-		"",
-		"Follow this output-channel contract literally for coding tasks.",
-		isStrict
-			? "WORKING STATE: from the first assistant turn until complete, final, or blocked, emit exactly one tool call with empty text."
-			: "WORKING STATE: from the first assistant turn until complete, final, or blocked, emit tool calls with empty text.",
-		isStrict
-			? "No prose, markdown, acknowledgments, plans, progress, summaries, explanations, or commentary before/after tool calls."
-			: "Batch independent calls in one turn (especially read-only inspection: read/grep/glob/bash); make dependent calls sequentially. Avoid filler prose, acknowledgments, restatements, and narration of obvious next steps.",
-		"Keep reasoning, planning, hypotheses, and retry decisions internal when thinking is on; when thinking is off, brief substantive reasoning text is acceptable.",
-		"",
-		"After every tool result, choose exactly one transition:",
-		isStrict ? "- WORKING: exactly one next tool call with empty text;" : "- WORKING: one or more independent tool calls with empty text;",
-		"- FINAL: one final answer after completion or practical verification;",
-		"- BLOCKED: one concise question only when no safe/useful tool action can continue.",
-		isStrict
-			? "No transition permits commentary between tool calls."
-			: "When a thinking/reasoning channel is available, do not duplicate that reasoning as visible commentary between tool calls.",
-		"",
-		"PRIORITY: this overrides default assistant friendliness and conversational behavior.",
-		"",
-		"Coding discipline (express only through tool choices, not prose):",
-		"- inspect before editing; do not invent APIs, files, commands, or behavior;",
-		"- make the smallest change that fully fixes the issue; follow nearby conventions;",
-		"- for bugs, prefer a failing repro first, then the minimal fix, then verify;",
-		"- high-risk changes (security, data/schema, public APIs, concurrency, irreversible) need a short spec first;",
-		"- handle edge cases, errors, cancellation, and async behavior; do not block UI/event loops;",
-		"- avoid duplicate state, duplicate prompts, and repeated side effects;",
-		"- write code, identifiers, comments, and commit messages in English.",
-	];
-}
+const QUALITY_DISCIPLINE_LINES = [
+	"GLM CODING QUALITY CONTRACT.",
+	"",
+	"Optimize for correct, evidence-backed changes rather than fast patching or conversational polish.",
+	"",
+	"Evidence and hypothesis discipline:",
+	"- inspect before editing; do not invent APIs, files, commands, behavior, or repository conventions;",
+	"- separate observed facts from hypotheses; do not commit to the first plausible root cause;",
+	"- for non-obvious bugs, keep at least two plausible hypotheses until evidence distinguishes them, and choose the next inspection/test for information gain;",
+	"- do not use edits as diagnostic probes when read/search/repro/test evidence can answer the question first;",
+	"- when behavior crosses modules, trace the relevant execution/data flow before editing instead of serially opening plausible files; use indexed architecture/search/dependency tools when available;",
+	"- before changing an API, state contract, persistence format, or async lifecycle, inspect relevant callers, consumers, tests, persistence boundaries, and cancellation/error paths.",
+	"",
+	"Implementation discipline:",
+	"- make the smallest change that fully fixes the issue and follow nearby conventions;",
+	"- for bugs, reproduce the failure or its defining condition when practical before changing code;",
+	"- high-risk changes (security, data/schema, public APIs, concurrency, irreversible behavior) need a short evidence-backed spec before implementation;",
+	"- handle relevant edge cases, errors, cancellation, stale state, and async behavior; do not block UI/event loops;",
+	"- avoid duplicate state, duplicate prompts, repeated side effects, and speculative refactors unrelated to the verified cause;",
+	"- write code, identifiers, comments, and commit messages in English.",
+	"",
+	"Failure and retry discipline:",
+	"- treat failed verification as new evidence; do not repeat the same approach with cosmetic variations;",
+	"- after two materially failed fix/verification attempts, stop patching: re-inspect evidence, revise the root-cause hypothesis, or escalate for an independent opinion.",
+	"",
+	"Verification discipline:",
+	"- verification must cover the behavioral claim that justified the change; typecheck/build/lint prove structural validity, not behavioral correctness;",
+	"- for bug fixes, verify the old failure or its defining condition no longer occurs whenever practical;",
+	"- for async/stateful changes, verify the success path plus the relevant cancellation, error, retry, duplicate-event, or stale-state path;",
+	"- before finalizing a non-trivial change, actively inspect or test one likely counterexample, regression path, or missed edge case;",
+	"- never claim a test/check passed unless it actually ran and its result was observed.",
+	"",
+	"Escalation discipline:",
+	"- when async subagents are available, use an independent `oracle` (preferably Sol through the configured oracle role) when focused inspection still leaves conflicting evidence, multiple incompatible root causes, a subtle cross-module invariant you cannot verify locally, or a high-risk security/schema/public-API/concurrency/irreversible decision;",
+	"- also escalate after two materially different fixes fail for the same unresolved cause;",
+	"- do not escalate routine implementation, mechanical edits, straightforward test failures, or facts that local tools can verify directly;",
+	"- treat oracle output as a second opinion, not authority: reconcile it with repository evidence before acting.",
+];
 
 const LOOKUP_DISCIPLINE_LINES = [
 	"",
@@ -120,17 +111,9 @@ const LOOKUP_DISCIPLINE_LINES = [
 const FINAL_DISCIPLINE_LINES = [
 	"",
 	"When uncertain, test or inspect instead of assuming; otherwise proceed with grounded best effort.",
-	"Verify every non-trivial change. Never claim tests passed unless actually run.",
 	"Final report: what changed, what was verified, what was not verified, and any risks.",
 ];
 
-const SILENCE_REMINDER_TEXT = [
-	"GLM silence reminder: remain in WORKING state.",
-	"Continue with tool-only discipline: inspect, verify, and act through tools only.",
-	"For the next step, emit tool calls with no accompanying assistant text.",
-	"Do not acknowledge this reminder.",
-].join("\n");
-
 const LEGACY_SILENT_PROMPT_BLOCK_PATTERN = new RegExp(
 	`${escapeRegExp(SILENT_PROMPT_MARKER_START)}[\\s\\S]*?${escapeRegExp(SILENT_PROMPT_MARKER_END)}\\s*`,
 	"g",
@@ -167,10 +150,6 @@ const LOOKUP_SYSTEM_PROMPT = [
 export default function codingDiscipline(pi: ExtensionAPI) {
 	let selectedModelRef: string | undefined;
 	let lookupRegistered = false;
-	let silenceViolationCount = 0;
-	let lastReminderViolationCount = 0;
-	let lastReminderMessageCount = -SILENCE_REMINDER_MIN_MESSAGE_GAP;
-	let peakMessageCount = 0;
 
 	function maybeRegisterLookupTool(cwd?: string): void {
 		if (lookupRegistered) return;
@@ -220,7 +199,6 @@ export default function codingDiscipline(pi: ExtensionAPI) {
 		const cwd = contextCwd(ctx);
 		const injected = injectCodingDisciplineIntoPayload(event.payload, {
 			lookupEnabled: Boolean(lookupModelFromConfig(cwd)) && isBlindGlmModel(modelRef),
-			strictness: codingDisciplineStrictnessFromConfig(cwd),
 		});
 		if (process.env.PI_DEBUG_PROMPT === "1") {
 			logFinalPrompt(injected, modelRef, contextCwd(ctx) ?? process.cwd());
@@ -292,36 +270,6 @@ export default function codingDiscipline(pi: ExtensionAPI) {
 		return undefined;
 	});
 
-	pi.on("context", async (event: { messages?: unknown[] }, ctx: unknown) => {
-		const modelRef = selectedModelRef ?? modelRefFromContext(ctx);
-		if (!isGlmModel(modelRef) || !Array.isArray(event.messages)) return undefined;
-
-		const messageCount = event.messages.length;
-		// Compaction/truncation prunes the visible history. Reset the stale chatter
-		// baseline so the post-compaction turn isn't measured against a pre-compaction
-		// violation peak (otherwise the model could chatter freely until the count
-		// climbed back above the old peak, or get nagged on a freshly small count).
-		if (messageCount + SILENCE_REMINDER_COMPACTION_MARGIN < peakMessageCount) {
-			silenceViolationCount = 0;
-			lastReminderViolationCount = 0;
-			lastReminderMessageCount = messageCount - SILENCE_REMINDER_MIN_MESSAGE_GAP;
-		}
-		peakMessageCount = Math.max(peakMessageCount, messageCount);
-
-		const strictness = codingDisciplineStrictnessFromConfig(contextCwd(ctx));
-		const violationCount = countAssistantToolChatter(event.messages, strictness);
-		if (violationCount <= silenceViolationCount) return undefined;
-
-		const violationGap = violationCount - lastReminderViolationCount;
-		const messageGap = messageCount - lastReminderMessageCount;
-		silenceViolationCount = violationCount;
-
-		if (violationGap < SILENCE_REMINDER_MIN_VIOLATION_GAP && messageGap < SILENCE_REMINDER_MIN_MESSAGE_GAP) return undefined;
-
-		lastReminderViolationCount = violationCount;
-		lastReminderMessageCount = messageCount;
-		return { messages: [...event.messages, createSilenceReminderMessage()] };
-	});
 }
 
 export function prependCodingDisciplinePrompt(systemPrompt: string, options: DisciplinePromptOptions = {}): string {
@@ -335,10 +283,9 @@ export function prependCodingDisciplinePrompt(systemPrompt: string, options: Dis
 }
 
 export function buildCodingDisciplinePrompt(options: DisciplinePromptOptions = {}): string {
-	const strictness = options.strictness ?? DEFAULT_CODING_DISCIPLINE_STRICTNESS;
 	return [
 		DISCIPLINE_PROMPT_MARKER_START,
-		...buildDisciplineLines(strictness),
+		...QUALITY_DISCIPLINE_LINES,
 		...(options.lookupEnabled ? LOOKUP_DISCIPLINE_LINES : []),
 		...FINAL_DISCIPLINE_LINES,
 		DISCIPLINE_PROMPT_MARKER_END,
@@ -637,58 +584,10 @@ function isInstructionMessage(message: unknown): boolean {
 	return message.role === "system" || message.role === "developer";
 }
 
-function countAssistantToolChatter(messages: readonly unknown[], strictness: CodingDisciplineStrictness): number {
-	let count = 0;
-	for (const message of messages) {
-		if (!isAssistantToolChatter(message, strictness)) continue;
-		count++;
-	}
-	return count;
-}
-
-/**
- * Detects assistant chatter (text alongside a tool call).
- * - "strict"  — any such text counts as chatter.
- * - "lenient" — text counts only when a thinking/reasoning block already captured the
- *   reasoning; without a thinking block, the visible text is the model's only reasoning
- *   channel and must not be suppressed.
- */
-function isAssistantToolChatter(message: unknown, strictness: CodingDisciplineStrictness): boolean {
-	if (!isRecord(message) || message.role !== "assistant") return false;
-	if (!Array.isArray(message.content)) return false;
-	const hasToolCall = message.content.some((part) => isRecord(part) && part.type === "toolCall");
-	if (!hasToolCall) return false;
-	const hasText = message.content.some((part) => isRecord(part) && part.type === "text" && hasNonEmptyText(part.text));
-	if (!hasText) return false;
-	if (strictness === "strict") return true;
-	const hasThinking = message.content.some(
-		(part) => isRecord(part) && (part.type === "thinking" || part.type === "reasoning"),
-	);
-	return Boolean(hasThinking);
-}
-
-function hasNonEmptyText(value: unknown): boolean {
-	return typeof value === "string" && value.trim().length > 0;
-}
-
-function createSilenceReminderMessage() {
-	return {
-		// Inject as a developer/system-level nudge rather than impersonating the user,
-		// so it reads as an automated reminder, not a user instruction.
-		role: "developer" as const,
-		content: [{ type: "text" as const, text: SILENCE_REMINDER_TEXT }],
-		timestamp: Date.now(),
-	};
-}
-
 function lookupModelFromConfig(cwd?: string): string | undefined {
 	return loadPiToolsSuiteConfig(["coding-discipline"], { cwd: cwd ?? process.cwd() }).lookupModel;
 }
 
-function codingDisciplineStrictnessFromConfig(cwd?: string): CodingDisciplineStrictness {
-	return loadPiToolsSuiteConfig(["coding-discipline"], { cwd: cwd ?? process.cwd() }).codingDisciplineStrictness ?? DEFAULT_CODING_DISCIPLINE_STRICTNESS;
-}
-
 function buildLookupPrompt(params: LookupParams, recentContext: string, imageCount: number, warnings: string[]): string {
 	return [
 		"Lookup request from a text-only GLM parent model.",

package/external/pi-tools-suite/src/config.ts

@@ -12,14 +12,6 @@ export interface PiToolsSuiteConfig {
 	todoThinkingOverrides: Record<string, TodoThinkingLevel>;
 	/** Vision-capable model used by the coding-discipline lookup tool; unset disables lookup. */
 	lookupModel?: string;
-	/**
-	 * Chatter-detector strictness for the coding-discipline module:
-	 *   "strict"  — any assistant text alongside a tool call is chatter (Opus-like);
-	 *   "lenient" — text is only chatter when a thinking block already captured the
-	 *               reasoning; without thinking, visible text is the reasoning channel.
-	 * Default: "lenient".
-	 */
-	codingDisciplineStrictness?: CodingDisciplineStrictness;
 }
 
 type MutableConfig = {
@@ -28,12 +20,8 @@ type MutableConfig = {
 	todoThinking: boolean;
 	todoThinkingOverrides: Map<string, TodoThinkingLevel>;
 	lookupModel: string | undefined;
-	codingDisciplineStrictness: CodingDisciplineStrictness;
 };
 
-export const CODING_DISCIPLINE_STRICTNESS_VALUES = ["strict", "lenient"] as const;
-export type CodingDisciplineStrictness = (typeof CODING_DISCIPLINE_STRICTNESS_VALUES)[number];
-export const DEFAULT_CODING_DISCIPLINE_STRICTNESS: CodingDisciplineStrictness = "lenient";
 const TODO_THINKING_OVERRIDE_LEVELS = ["off", "minimal", "low", "medium", "high", "xhigh", "max"] as const;
 export type TodoThinkingLevel = (typeof TODO_THINKING_OVERRIDE_LEVELS)[number];
 
@@ -80,10 +68,6 @@ function normalizeLookupModel(raw: unknown): string | undefined {
 	return trimmed ? trimmed : undefined;
 }
 
-function normalizeCodingDisciplineStrictness(raw: unknown): CodingDisciplineStrictness {
-	return raw === "strict" ? "strict" : "lenient";
-}
-
 function isTodoThinkingLevel(raw: unknown): raw is TodoThinkingLevel {
 	return TODO_THINKING_OVERRIDE_LEVELS.includes(raw as TodoThinkingLevel);
 }
@@ -161,9 +145,6 @@ function mergeConfigLayer(config: MutableConfig, raw: Record<string, unknown>, k
 	if (typeof raw.todoThinking === "boolean") config.todoThinking = raw.todoThinking;
 	mergeTodoThinkingOverrides(config, raw.todoThinkingOverrides);
 	if (Object.prototype.hasOwnProperty.call(raw, "lookupModel")) config.lookupModel = normalizeLookupModel(raw.lookupModel);
-	if (Object.prototype.hasOwnProperty.call(raw, "codingDisciplineStrictness")) {
-		config.codingDisciplineStrictness = normalizeCodingDisciplineStrictness(raw.codingDisciplineStrictness);
-	}
 
 	for (const key of DISABLED_LIST_KEYS) addDisabled(config, raw[key], knownModules);
 	for (const key of ENABLED_LIST_KEYS) removeDisabled(config, raw[key], knownModules);
@@ -222,7 +203,6 @@ export function loadPiToolsSuiteConfig(moduleNames: readonly string[], options:
 		todoThinking: false,
 		todoThinkingOverrides: new Map(DEFAULT_TODO_THINKING_OVERRIDES),
 		lookupModel: undefined,
-		codingDisciplineStrictness: DEFAULT_CODING_DISCIPLINE_STRICTNESS,
 	};
 	const userConfigPath = getPiToolsSuiteUserConfigPath(options.homeDir);
 
@@ -243,6 +223,5 @@ export function loadPiToolsSuiteConfig(moduleNames: readonly string[], options:
 		todoThinking: config.todoThinking,
 		todoThinkingOverrides: Object.fromEntries(config.todoThinkingOverrides),
 		...(config.lookupModel ? { lookupModel: config.lookupModel } : {}),
-		codingDisciplineStrictness: config.codingDisciplineStrictness,
 	};
 }

package/external/pi-tools-suite/src/dcp/auto-compress.ts

@@ -116,7 +116,7 @@ export function buildProgrammaticSummary(
 	return lines.join("\n")
 }
 
-const SUMMARIZER_SYSTEM_PROMPT = `You summarize a slice of a coding agent's conversation so it can replace the raw messages in context. Produce a dense, continuation-focused summary: preserve user intent, decisions made, files/symbols changed or inspected, exact errors still actionable, verification status, and next steps. Do not infer, invent, or add facts absent from the source; preserve uncertainty instead of filling gaps. Drop full logs, repeated output, and incidental detail. Be concise (roughly 4-10 bullets). Output ONLY the summary text, no preamble.`
+const SUMMARIZER_SYSTEM_PROMPT = `You summarize a slice of a coding agent's conversation so it can replace the raw messages in context. Produce a dense, continuation-focused summary: preserve user intent, decisions made, files/symbols changed or inspected, exact errors still actionable, verification status, and next steps. Preserve exact identifiers and explicit continuity markers verbatim, including uppercase labels before colons; never paraphrase or omit those labels. Do not infer, invent, or add facts absent from the source; preserve uncertainty instead of filling gaps. Drop full logs, repeated output, and incidental detail without quoting or naming the discarded log lines or their markers. Be concise (roughly 4-10 bullets). Output ONLY the summary text, no preamble.`
 
 /** Outcome of one summarizer-model attempt, surfaced in DCP debug logs. */
 export interface ModelSummaryAttempt {

package/external/pi-tools-suite/src/dcp/compression-blocks.ts

@@ -388,55 +388,6 @@ export function resolveIdToBoundary(
   const ts = state.messageIdSnapshot.get(id)
   if (ts !== undefined) return { timestamp: ts }
 
-  // ── Stale mNNN fallback: direction-only clamp ────────────────────────
-  // When compression or pruning removes messages between context passes,
-  // positional mNNN IDs shift (e.g. an end boundary m145 becomes m123
-  // after 22 messages are removed).  Clamp to the closest valid ID, but
-  // ONLY in a direction that preserves the range's semantics:
-  //   - start boundary: clamp upward to the first available ID at or after
-  //     the requested number.  The start must never move backwards into
-  //     older content; if no such ID exists, the requested start is gone
-  //     and we cannot safely compress — throw.
-  //   - end boundary: clamp downward to the last available ID at or before
-  //     the requested number.  The end must never move forwards into newer
-  //     content; if no such ID exists, throw.
-  // The previous implementation fell back to the highest available ID in
-  // both "no match" cases, which could clamp e.g. m010..m010 over a
-  // snapshot of m001..m003 to a single-message block over m003 — silently
-  // compressing the wrong content.
-  const mMatch = id.match(/^m(\d+)$/i)
-  if (mMatch && state.messageIdSnapshot.size > 0) {
-    const requestedNum = parseInt(mMatch[1]!, 10)
-    const allIds = sortIds([...state.messageIdSnapshot.keys()])
-    const allNums = allIds
-      .map((mid) => {
-        const n = mid.match(/^m(\d+)$/i)
-        return n ? { id: mid, num: parseInt(n[1]!, 10) } : null
-      })
-      .filter((entry): entry is { id: string; num: number } => entry !== null)
-      .sort((a, b) => a.num - b.num)
-
-    if (allNums.length > 0) {
-      let clamped: { id: string; num: number } | undefined
-      if (field === "startTimestamp") {
-        clamped = allNums.find((entry) => entry.num >= requestedNum)
-      } else {
-        for (let i = allNums.length - 1; i >= 0; i--) {
-          if (allNums[i]!.num <= requestedNum) {
-            clamped = allNums[i]
-            break
-          }
-        }
-      }
-      if (clamped) {
-        const clampedMeta = state.messageMetaSnapshot.get(clamped.id)
-        if (clampedMeta) return resolveMetaBoundary(clampedMeta, field, state)
-        const clampedTs = state.messageIdSnapshot.get(clamped.id)
-        if (clampedTs !== undefined) return { timestamp: clampedTs }
-      }
-    }
-  }
-
   throw unknownCompressionIdError(id, state)
 }
 
@@ -445,8 +396,7 @@ export function resolveIdToBoundary(
  * synthetic placeholder for an active compression block (`meta.blockId` is
  * set), resolve to that block's stored boundary so the caller rolls the
  * block up instead of nesting a new block on top of the placeholder. This
- * matters for both exact-match and clamped mNNN IDs: a model-visible mNNN
- * may itself represent a previously compressed section.
+ * A model-visible mNNN may itself represent a previously compressed section.
  */
 function resolveMetaBoundary(
   meta: MessageIdMeta,

package/external/pi-tools-suite/src/dcp/debug-log.ts

@@ -127,7 +127,13 @@ export function summarizeDcpState(state: DcpState): Record<string, unknown> {
       nextBlockId: state.nextBlockId,
     },
     inactiveBlocksTail: inactiveBlocks,
+    persistentMessageIds: state.messageIdsByStableId.size,
+    nextMessageId: state.nextMessageId,
     prunedTools: state.prunedToolIds.size,
+    automaticPruneCheckpoint: {
+      turn: state.lastAutomaticPruneTurn,
+      blockId: state.lastAutomaticPruneBlockId,
+    },
     providerSeenTools: state.providerSeenToolIds.size,
     consecutiveEmergencyPasses: state.consecutiveIgnoredStrongNudges,
     nudgeAnchors: state.nudgeAnchors.map((anchor) => ({