@vanillagreen/pi-qol 1.4.3 → 1.5.0

package/extensions/qol/transcript-risk.ts

@@ -0,0 +1,28 @@
+// Transcript risk for /context: estimate serialized request payload size and
+// warn before it crosses a char budget, even when the token count alone has
+// not reached the context window. Keeps /context independent of compaction.ts
+// (no summarizer/fs/artifact dependencies on the UI path).
+
+import type { AgentMessage } from "@earendil-works/pi-agent-core";
+import { convertToLlm, serializeConversation } from "@earendil-works/pi-coding-agent";
+import { evaluateTranscriptRisk, type TranscriptRiskResult } from "./budget-guard.js";
+
+function stripThinking(messages: any[]): any[] {
+	return messages.map((message: any) => {
+		if (message.role !== "assistant" || !Array.isArray(message.content)) return message;
+		return { ...message, content: message.content.filter((part: any) => part?.type !== "thinking") };
+	});
+}
+
+export function transcriptRiskState(messages: AgentMessage[], threshold: number): TranscriptRiskResult {
+	if (!Array.isArray(messages) || messages.length === 0 || threshold <= 0) {
+		return { chars: 0, exceeded: false, messageCount: messages?.length ?? 0, threshold };
+	}
+	try {
+		const serialized = serializeConversation(stripThinking(convertToLlm(messages)));
+		return evaluateTranscriptRisk({ chars: serialized.length, messageCount: messages.length, threshold });
+	} catch (error) {
+		const message = error instanceof Error ? error.message : String(error);
+		return evaluateTranscriptRisk({ chars: 0, error: message, messageCount: messages.length, threshold });
+	}
+}

package/extensions/qol.ts

@@ -2,7 +2,9 @@ import type { ExtensionAPI, ExtensionCommandContext, ExtensionContext } from "@e
 import type { TUI } from "@earendil-works/pi-tui";
 import { criticalInfo, lastAssistantTextFromAgentEnd, needsDirection, taskStats } from "./qol/agent-end.js";
 import { getQuestionService, readCavemanBridge, type QuestionOpenedEventLike } from "./qol/bridges.js";
+import { BudgetGuardDriver } from "./qol/budget-guard-runtime.js";
 import {
+	budgetGuardTrigger,
 	compactionTriggerReason,
 	handleQolBranchSummary,
 	handleQolCompaction,
@@ -257,6 +259,35 @@ export default function qol(pi: ExtensionAPI): void {
 		idleCompactionTimer = undefined;
 	};
 
+	const budgetGuardDriver = new BudgetGuardDriver();
+
+	const resetBudgetGuard = () => {
+		budgetGuardDriver.reset();
+	};
+
+	const maybeFireBudgetGuard = (ctx: ExtensionContext) => {
+		let trigger;
+		try {
+			trigger = budgetGuardTrigger(ctx);
+		} catch (error) {
+			if (isStaleCtxError(error)) return;
+			throw error;
+		}
+		const notifySafely = (message: string, level: "info" | "warning" | "error") => {
+			try {
+				if (ctx.hasUI && settingBoolean("compaction.notify", true, ctx.cwd)) ctx.ui.notify(message, level);
+			} catch (error) {
+				if (isStaleCtxError(error)) return;
+				throw error;
+			}
+		};
+		budgetGuardDriver.dispatch({
+			compact: typeof ctx.compact === "function" ? ctx.compact.bind(ctx) : undefined,
+			notify: notifySafely,
+			trigger,
+		});
+	};
+
 	const scheduleIdleCompaction = (ctx: ExtensionContext) => {
 		clearIdleCompactionTimer();
 		if (!settingBoolean("compaction.idleEnabled", false, ctx.cwd)) return;
@@ -509,6 +540,7 @@ export default function qol(pi: ExtensionAPI): void {
 		subscribeCavemanBridge();
 		latestSystemPromptOptions = undefined;
 		resetAutoRename();
+		resetBudgetGuard();
 		resetThinkingTimer(ctx);
 		void consumePendingSessionSearchContext(pi, ctx, event.reason);
 		installAutocompleteHintStyling(ctx);
@@ -569,6 +601,7 @@ export default function qol(pi: ExtensionAPI): void {
 		cavemanUnsubscribe?.();
 		cavemanUnsubscribe = undefined;
 		resetAutoRename();
+		resetBudgetGuard();
 		clearIdleCompactionTimer();
 		clearQuestionSubscribeTimer();
 		if (sessionSearchWarmupTimer) clearTimeout(sessionSearchWarmupTimer);
@@ -637,6 +670,7 @@ export default function qol(pi: ExtensionAPI): void {
 			void refreshStatusline(ctx);
 			requestRender();
 		}
+		maybeFireBudgetGuard(ctx);
 		scheduleIdleCompaction(ctx);
 		void attemptAutoRename(ctx);
 		const text = lastAssistantTextFromAgentEnd(event, ctx);
@@ -660,6 +694,9 @@ export default function qol(pi: ExtensionAPI): void {
 		sendQolNotification(ctx, "ready", settingString("notification.readyMessage", "Ready for input", ctx.cwd), "info", "ready");
 	});
 	pi.on("session_compact", (_event, ctx) => {
+		// After a successful compaction usage drops below the budget, so reset the
+		// crossing key so the next threshold crossing re-fires the guard.
+		budgetGuardDriver.noteSessionCompacted();
 		if (!ctx.hasUI) return;
 		void refreshStatusline(ctx);
 		requestRender();

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@vanillagreen/pi-qol",
-  "version": "1.4.3",
+  "version": "1.5.0",
   "description": "Pi quality-of-life extension: compact statusline/π prompt, reliable multiline input, styled pasted-image chips, session naming/search/context import, handoff, permission prompts, notifications, custom compaction, and a collapsed-thinking timer.",
   "license": "MIT",
   "keywords": [
@@ -16,6 +16,9 @@
     ],
     "image": "https://raw.githubusercontent.com/vanillagreencom/vstack/main/pi-extensions/pi-qol/assets/settings-panel.png"
   },
+  "scripts": {
+    "test": "bun test ./tests"
+  },
   "vstack": {
     "extensionManager": {
       "displayName": "QOL",
@@ -30,6 +33,19 @@
           "apply": "reload",
           "requiresReload": true
         },
+        {
+          "key": "glyphStyle",
+          "label": "Glyph style",
+          "description": "Glyphs for vstack-owned UI chrome. Unicode is the default; ASCII uses plain box/status/separator characters for terminal compatibility. pi-tool-renderer's global override wins when set.",
+          "type": "enum",
+          "enumValues": [
+            "unicode",
+            "ascii"
+          ],
+          "default": "unicode",
+          "category": "Rendering",
+          "apply": "live"
+        },
         {
           "key": "replaceFooter",
           "label": "Replace built-in footer",
@@ -654,9 +670,9 @@
         {
           "key": "compaction.model",
           "label": "Compaction model",
-          "description": "Summarizer model as provider/model, a bare model id, or current. Default mirrors Pi's custom-compaction example.",
+          "description": "Summarizer model as provider/model, a bare model id, or current. Default uses Pi's active model.",
           "type": "string",
-          "default": "google/gemini-2.5-flash",
+          "default": "current",
           "category": "Compaction",
           "apply": "live"
         },
@@ -782,6 +798,60 @@
           "category": "Compaction Threshold",
           "apply": "live"
         },
+        {
+          "key": "compaction.budgetGuardEnabled",
+          "label": "Long-session budget guard",
+          "description": "Trigger compaction on agent_end (no idle wait) when context usage crosses budgetPercent or budgetTokens. Designed for autonomous/Flightdeck runs so transcript growth cannot silently hit provider/buffer limits. Fires once per threshold crossing, not in a loop.",
+          "type": "boolean",
+          "default": true,
+          "category": "Compaction Budget Guard",
+          "apply": "live"
+        },
+        {
+          "key": "compaction.budgetPercent",
+          "label": "Budget guard percent",
+          "description": "Context-window percentage that fires the budget guard at agent_end. Set -1 to disable percent-based firing and rely on absolute tokens instead.",
+          "type": "number",
+          "default": 85,
+          "category": "Compaction Budget Guard",
+          "apply": "live"
+        },
+        {
+          "key": "compaction.budgetTokens",
+          "label": "Budget guard token limit",
+          "description": "Absolute token count above which the budget guard fires at agent_end. Set -1 to use only budgetPercent.",
+          "type": "number",
+          "default": -1,
+          "category": "Compaction Budget Guard",
+          "apply": "live"
+        },
+        {
+          "key": "compaction.maxInputChars",
+          "label": "Chunked compaction input cap",
+          "description": "Maximum serialized characters sent to the summarizer in a single request. Inputs larger than this are split into chunks, summarized chunk-by-chunk, then merged via a summary-of-summaries pass so the compaction request payload itself cannot blow past provider buffer limits. Set 0 to disable chunking.",
+          "type": "number",
+          "default": 240000,
+          "category": "Compaction Budget Guard",
+          "apply": "live"
+        },
+        {
+          "key": "compaction.handoffArtifactEnabled",
+          "label": "Write pre-compaction handoff artifact",
+          "description": "Before compaction, write a JSON handoff artifact under ~/.pi/agent/vstack/sessions/<session>/pi-qol/handoff/ containing the previous summary, last task state, and recently-referenced file/artifact paths. Always writes a stamped file and a latest.json pointer. Runs whenever the QOL bounded compaction handler runs — that is, on every budget-guard-triggered compaction (regardless of compaction.customEnabled), and on manual/idle compactions only when compaction.customEnabled is on. Write failures surface a QOL warning and a handoffArtifactError field in the compaction details.",
+          "type": "boolean",
+          "default": true,
+          "category": "Compaction Budget Guard",
+          "apply": "live"
+        },
+        {
+          "key": "compaction.transcriptRiskWarnChars",
+          "label": "Transcript-risk warn budget (chars)",
+          "description": "Show a transcript-risk warning in /context when the serialized payload of the messages-to-send exceeds this many characters, even if the token count is still below the model context window. Set 0 to disable the warning.",
+          "type": "number",
+          "default": 600000,
+          "category": "Compaction Budget Guard",
+          "apply": "live"
+        },
         {
           "key": "thinkingLabel.text",
           "label": "Hidden thinking label",