@vanillagreen/pi-qol 1.2.0 → 1.4.0

package/extensions/qol/compaction.ts

@@ -0,0 +1,264 @@
+import type { AgentMessage } from "@earendil-works/pi-agent-core";
+import { complete, type Message } from "@earendil-works/pi-ai";
+import { convertToLlm, serializeConversation, type ExtensionContext } from "@earendil-works/pi-coding-agent";
+import {
+	DEFAULT_COMPACTION_MAX_TOKENS,
+	DEFAULT_COMPACTION_MODEL,
+	DEFAULT_IDLE_COMPACTION_THRESHOLD_TOKENS,
+	QOL_COMPACTION_SYSTEM_PROMPT,
+} from "./constants.js";
+import { settingBoolean, settingNumber, settingString } from "./settings.js";
+import { stringifyError } from "./util.js";
+
+export type QolSummaryProfile = "concise" | "balanced" | "exhaustive";
+export type QolSummaryPurpose = "compaction" | "branch-summary" | "session-search";
+
+export function compactionNotify(ctx: ExtensionContext, message: string, level: "info" | "warning" | "error" = "info"): void {
+	if (ctx.hasUI && settingBoolean("compaction.notify", true, ctx.cwd)) ctx.ui.notify(message, level);
+}
+
+export function compactionProfile(cwd: string): QolSummaryProfile {
+	const value = settingString("compaction.profile", "balanced", cwd);
+	return value === "concise" || value === "exhaustive" ? value : "balanced";
+}
+
+function compactionProfileInstructions(profile: QolSummaryProfile): string {
+	if (profile === "concise") return "Prefer a compact continuation summary. Include only decisions, current state, modified/read files, blockers, and concrete next steps.";
+	if (profile === "exhaustive") return "Be thorough. The summary may replace substantial conversation history, so preserve all relevant implementation details, alternatives considered, exact file paths, commands, errors, and pending work.";
+	return "Be complete but not verbose. Preserve enough detail for a future assistant to continue without the old transcript.";
+}
+
+function stripThinkingForSummary(messages: Message[]): Message[] {
+	return messages.map((message) => {
+		if (message.role !== "assistant" || !Array.isArray(message.content)) return message;
+		return {
+			...message,
+			content: message.content.filter((part: any) => part?.type !== "thinking"),
+		};
+	});
+}
+
+export function serializeMessagesForSummary(messages: AgentMessage[]): string {
+	return serializeConversation(stripThinkingForSummary(convertToLlm(messages)));
+}
+
+function customMessageContentToText(content: unknown): string {
+	if (typeof content === "string") return content;
+	if (!Array.isArray(content)) return "";
+	const parts: string[] = [];
+	for (const part of content) {
+		if (part?.type === "text" && typeof part.text === "string") parts.push(part.text);
+		else if (part?.type === "image") parts.push(`[image${typeof part.mimeType === "string" ? ` ${part.mimeType}` : ""}]`);
+		else if (part?.type) parts.push(`[${String(part.type)}]`);
+	}
+	return parts.join("\n").trim();
+}
+
+function buildSummaryPrompt(options: {
+	conversationText: string;
+	customInstructions?: string;
+	previousSummary?: string;
+	profile: QolSummaryProfile;
+	purpose: QolSummaryPurpose;
+}): string {
+	const purposeText = options.purpose === "branch-summary"
+		? "the branch being left during /tree navigation"
+		: options.purpose === "session-search"
+			? "the previous session being imported into the current context"
+			: "the conversation span being compacted";
+	const previous = options.previousSummary ? `<previous-summary>\n${options.previousSummary}\n</previous-summary>\n\n` : "";
+	const custom = options.customInstructions?.trim() ? `<custom-instructions>\n${options.customInstructions.trim()}\n</custom-instructions>\n\n` : "";
+	return `${custom}${previous}<conversation>\n${options.conversationText}\n</conversation>\n\nSummarize ${purposeText} for a coding agent that must continue the work.\n\n${compactionProfileInstructions(options.profile)}\n\nUse this markdown shape:\n\n## Goal\n[What the user is trying to accomplish]\n\n## Constraints & Preferences\n- [Requirements, style, safety, or user preferences]\n\n## Progress\n### Done\n- [x] [Completed work]\n\n### In Progress\n- [ ] [Current partial work]\n\n### Blocked\n- [Blockers or none]\n\n## Key Decisions\n- **[Decision]**: [Rationale]\n\n## Files & Commands\n- [Files read/modified and important commands/results]\n\n## Next Steps\n1. [Most important next action]\n\n## Critical Context\n- [Anything easy to lose but needed later]`;
+}
+
+async function summarizeWithRemote(endpoint: string, systemPrompt: string, promptText: string, maxTokens: number, signal?: AbortSignal): Promise<string> {
+	const response = await fetch(endpoint, {
+		body: JSON.stringify({ maxTokens, prompt: promptText, systemPrompt }),
+		headers: { "content-type": "application/json" },
+		method: "POST",
+		signal,
+	});
+	const text = await response.text();
+	if (!response.ok) throw new Error(`Remote compaction endpoint returned ${response.status}: ${text.slice(0, 500)}`);
+	let parsed: unknown;
+	try {
+		parsed = JSON.parse(text);
+	} catch {
+		throw new Error("Remote compaction endpoint did not return JSON");
+	}
+	if (parsed && typeof parsed === "object") {
+		const record = parsed as Record<string, unknown>;
+		if (typeof record.summary === "string") return record.summary;
+		if (typeof record.text === "string") return record.text;
+	}
+	throw new Error("Remote compaction response missing summary");
+}
+
+export function resolveConfiguredModel(ctx: ExtensionContext, configured: string): any | undefined {
+	if (!configured || configured === "current") return ctx.model;
+	const withoutThinking = configured.replace(/:(off|minimal|low|medium|high|xhigh)$/i, "");
+	const slash = withoutThinking.indexOf("/");
+	if (slash > 0) return ctx.modelRegistry.find(withoutThinking.slice(0, slash), withoutThinking.slice(slash + 1));
+	const providers = [ctx.model?.provider, "google", "openai", "anthropic", "mistral", "moonshot", "cloudflare-ai-gateway", "cloudflare-workers-ai"].filter((value): value is string => typeof value === "string");
+	for (const provider of providers) {
+		const model = ctx.modelRegistry.find(provider, withoutThinking);
+		if (model) return model;
+	}
+	return undefined;
+}
+
+export function modelLabel(model: any): string {
+	return model ? `${model.provider}/${model.id}` : "unknown model";
+}
+
+export async function generateQolSummary(ctx: ExtensionContext, options: {
+	conversationText: string;
+	customInstructions?: string;
+	previousSummary?: string;
+	maxTokens?: number;
+	model?: string;
+	purpose: QolSummaryPurpose;
+	signal?: AbortSignal;
+}): Promise<{ model: string; summary: string; via: "model" | "remote" }> {
+	const maxTokens = Math.max(256, Math.floor(options.maxTokens ?? settingNumber("compaction.maxTokens", DEFAULT_COMPACTION_MAX_TOKENS, ctx.cwd)));
+	const promptText = buildSummaryPrompt({
+		conversationText: options.conversationText,
+		customInstructions: options.customInstructions,
+		previousSummary: settingBoolean("compaction.includePreviousSummary", true, ctx.cwd) ? options.previousSummary : undefined,
+		profile: compactionProfile(ctx.cwd),
+		purpose: options.purpose,
+	});
+
+	const remoteEndpoint = settingString("compaction.remoteEndpoint", "", ctx.cwd);
+	if (settingBoolean("compaction.remoteEnabled", false, ctx.cwd) && remoteEndpoint) {
+		try {
+			const summary = await summarizeWithRemote(remoteEndpoint, QOL_COMPACTION_SYSTEM_PROMPT, promptText, maxTokens, options.signal);
+			return { model: remoteEndpoint, summary, via: "remote" };
+		} catch (error) {
+			compactionNotify(ctx, `Remote compaction failed, trying model fallback: ${stringifyError(error)}`, "warning");
+		}
+	}
+
+	const configuredModel = options.model ?? settingString("compaction.model", DEFAULT_COMPACTION_MODEL, ctx.cwd);
+	const model = resolveConfiguredModel(ctx, configuredModel);
+	if (!model) throw new Error(`Summary model not found: ${configuredModel}`);
+	const auth = await ctx.modelRegistry.getApiKeyAndHeaders(model);
+	if (!auth.ok) throw new Error(auth.error);
+	if (!auth.apiKey) throw new Error(`No API key for ${model.provider}`);
+
+	const message: Message = {
+		content: [{ text: promptText, type: "text" }],
+		role: "user",
+		timestamp: Date.now(),
+	};
+	const response = await complete(
+		model,
+		{ messages: [message], systemPrompt: QOL_COMPACTION_SYSTEM_PROMPT },
+		{ apiKey: auth.apiKey, headers: auth.headers, maxTokens, signal: options.signal },
+	);
+	const summary = response.content
+		.filter((content): content is { type: "text"; text: string } => content.type === "text")
+		.map((content) => content.text)
+		.join("\n")
+		.trim();
+	return { model: modelLabel(model), summary, via: "model" };
+}
+
+export async function handleQolCompaction(event: any, ctx: ExtensionContext): Promise<any> {
+	if (!settingBoolean("compaction.customEnabled", false, ctx.cwd)) return undefined;
+	const preparation = event.preparation ?? {};
+	const messages = [...(preparation.messagesToSummarize ?? []), ...(preparation.turnPrefixMessages ?? [])];
+	if (messages.length === 0) return undefined;
+	const tokensBefore = typeof preparation.tokensBefore === "number" ? preparation.tokensBefore : 0;
+	compactionNotify(ctx, `QOL compaction: summarizing ${messages.length} message(s), ${tokensBefore.toLocaleString()} token(s).`, "info");
+	try {
+		const conversationText = serializeMessagesForSummary(messages);
+		const result = await generateQolSummary(ctx, {
+			conversationText,
+			customInstructions: event.customInstructions,
+			previousSummary: preparation.previousSummary,
+			purpose: "compaction",
+			signal: event.signal,
+		});
+		if (!result.summary.trim()) throw new Error("Compaction summary was empty");
+		compactionNotify(ctx, `QOL compaction complete via ${result.via}: ${result.model}`, "info");
+		return {
+			compaction: {
+				details: {
+					messageCount: messages.length,
+					model: result.model,
+					profile: compactionProfile(ctx.cwd),
+					source: "pi-qol",
+					via: result.via,
+				},
+				firstKeptEntryId: preparation.firstKeptEntryId,
+				summary: result.summary,
+				tokensBefore: preparation.tokensBefore,
+			},
+		};
+	} catch (error) {
+		if (event.signal?.aborted) return undefined;
+		compactionNotify(ctx, `QOL compaction failed: ${stringifyError(error)}`, "error");
+		return settingBoolean("compaction.fallbackToDefault", true, ctx.cwd) ? undefined : { cancel: true };
+	}
+}
+
+function summarizeEntryForBranch(entry: any): string[] {
+	if (entry?.type === "message" && entry.message) return [serializeMessagesForSummary([entry.message])];
+	if (entry?.type === "compaction" && typeof entry.summary === "string") return [`[Compaction summary]: ${entry.summary}`];
+	if (entry?.type === "branch_summary" && typeof entry.summary === "string") return [`[Branch summary]: ${entry.summary}`];
+	if (entry?.type === "custom_message") return [`[Custom message${entry.customType ? `:${entry.customType}` : ""}]: ${customMessageContentToText(entry.content) || "[empty]"}`];
+	return [];
+}
+
+export async function handleQolBranchSummary(event: any, ctx: ExtensionContext): Promise<any> {
+	if (!settingBoolean("compaction.branchSummaryEnabled", false, ctx.cwd)) return undefined;
+	const preparation = event.preparation ?? {};
+	if (preparation.userWantsSummary !== true) return undefined;
+	const entries = Array.isArray(preparation.entriesToSummarize) ? preparation.entriesToSummarize : [];
+	const conversationText = entries.flatMap(summarizeEntryForBranch).join("\n\n").trim();
+	if (!conversationText) return undefined;
+	compactionNotify(ctx, `QOL branch summary: summarizing ${entries.length} entr${entries.length === 1 ? "y" : "ies"}.`, "info");
+	try {
+		const result = await generateQolSummary(ctx, {
+			conversationText,
+			customInstructions: event.customInstructions ?? preparation.customInstructions,
+			purpose: "branch-summary",
+			signal: event.signal,
+		});
+		if (!result.summary.trim()) throw new Error("Branch summary was empty");
+		return {
+			summary: {
+				details: { entryCount: entries.length, model: result.model, profile: compactionProfile(ctx.cwd), source: "pi-qol", via: result.via },
+				summary: result.summary,
+			},
+		};
+	} catch (error) {
+		if (event.signal?.aborted) return undefined;
+		compactionNotify(ctx, `QOL branch summary failed: ${stringifyError(error)}`, "error");
+		return undefined;
+	}
+}
+
+function contextUsage(ctx: ExtensionContext): { contextWindow?: number; tokens: number } | undefined {
+	const usage = ctx.getContextUsage?.() as { tokens?: unknown; contextWindow?: unknown } | undefined;
+	const tokens = Number(usage?.tokens);
+	if (!Number.isFinite(tokens) || tokens <= 0) return undefined;
+	const contextWindow = Number(usage?.contextWindow ?? ctx.model?.contextWindow);
+	return { contextWindow: Number.isFinite(contextWindow) && contextWindow > 0 ? contextWindow : undefined, tokens };
+}
+
+export function compactionTriggerReason(ctx: ExtensionContext): string | undefined {
+	const usage = contextUsage(ctx);
+	if (!usage) return undefined;
+	const tokenLimit = settingNumber("compaction.thresholdTokens", -1, ctx.cwd);
+	if (tokenLimit > 0 && usage.tokens >= tokenLimit) return `${usage.tokens.toLocaleString()} tokens >= ${Math.floor(tokenLimit).toLocaleString()} token limit`;
+	const percentLimit = settingNumber("compaction.thresholdPercent", -1, ctx.cwd);
+	if (percentLimit > 0 && usage.contextWindow) {
+		const percent = (usage.tokens / usage.contextWindow) * 100;
+		if (percent >= percentLimit) return `${percent.toFixed(1)}% context >= ${percentLimit}% limit`;
+	}
+	const idleLimit = settingNumber("compaction.idleThresholdTokens", DEFAULT_IDLE_COMPACTION_THRESHOLD_TOKENS, ctx.cwd);
+	if (usage.tokens >= idleLimit) return `${usage.tokens.toLocaleString()} tokens >= ${Math.floor(idleLimit).toLocaleString()} idle threshold`;
+	return undefined;
+}

package/extensions/qol/constants.ts

@@ -0,0 +1,107 @@
+export const INSTALL_SYMBOL = Symbol.for("vstack.pi-qol.installed");
+export const CONFIG_ID = "@vanillagreen/pi-qol";
+
+export const STATUS_KEY = "qol-attachments";
+export const SESSION_SEARCH_STATUS_KEY = "qol-session-search";
+export const SESSION_SEARCH_CONTEXT_TYPE = "qol-session-context";
+export const CONTEXT_USAGE_MESSAGE_TYPE = "qol-context-usage";
+export const SESSION_MANAGER_STATUS_KEY = "session-manager";
+
+export const IMAGE_EXTENSIONS = new Set([".png", ".jpg", ".jpeg", ".gif", ".webp", ".bmp", ".tif", ".tiff", ".heic", ".heif"]);
+export const IMAGE_PATH_PATTERN = /(^|[\s(\[{<"'`])(@?(?:~|\.\.?|\/)[^\s)\]}>"'`]+?\.(?:png|jpe?g|gif|webp|bmp|tiff?|heic|heif))(?=$|[\s)\]}>"'`,.;:!?])/gi;
+
+export const QUESTION_SERVICE_SYMBOL = Symbol.for("vstack.pi-questions.service");
+export const QOL_NOTIFICATION_SERVICE_SYMBOL = Symbol.for("vstack.pi-qol.notification-service");
+export const VSTACK_MODAL_LOCK_SYMBOL = Symbol.for("vstack.pi.modal-lock");
+export const CAVEMAN_BRIDGE_SYMBOL = Symbol.for("vstack.pi.caveman");
+export const PI_AGENTS_STATUSLINE_SYMBOL = Symbol.for("vstack.pi-agents-tmux.statusline");
+
+export const CAVEMAN_ICON_ACTIVE = "\uee9a";
+export const CAVEMAN_ICON_INACTIVE = "\u{f19e0}";
+
+export const THINKING_TIMER_STORE_SYMBOL = Symbol.for("vstack.pi-qol.thinking-timer.store");
+export const THINKING_TIMER_PATCH_SYMBOL = Symbol.for("vstack.pi-qol.thinking-timer.patch");
+export const SESSION_SEARCH_PENDING_SYMBOL = Symbol.for("vstack.pi-qol.session-search.pending-context");
+export const PENDING_QUEUE_THEME_PATCH_SYMBOL = Symbol.for("vstack.pi-qol.pending-queue.theme-patch");
+export const STATUS_TEXT_ALIGNMENT_PATCH_SYMBOL = Symbol.for("vstack.pi-qol.status-text-alignment-patch");
+
+export const QUESTION_OPENED_EVENT = "vstack:pi-questions:opened";
+export const QUESTION_NOTIFY_DEDUP_MS = 2000;
+
+export const DEFAULT_NOTIFICATION_TITLE = "Pi";
+export const DEFAULT_NOTIFICATION_COOLDOWN_SECONDS = 8;
+export const DEFAULT_NOTIFICATION_BODY_MAX_CHARS = 240;
+export const DEFAULT_TMUX_MESSAGE_DURATION_MS = 5000;
+
+export const DEFAULT_COMPACTION_MODEL = "google/gemini-2.5-flash";
+export const DEFAULT_COMPACTION_MAX_TOKENS = 8192;
+export const DEFAULT_IDLE_COMPACTION_THRESHOLD_TOKENS = 200000;
+export const DEFAULT_IDLE_COMPACTION_SECONDS = 300;
+
+export const DEFAULT_PERMISSION_GATE_COMMANDS = "rm -Rf";
+export const DEFAULT_PERMISSION_GATE_PREVIEW_LINES = 12;
+export const DEFAULT_PERMISSION_GATE_PREVIEW_CHARS = 1200;
+export const DEFAULT_PERMISSION_GATE_PREVIEW_LINE_WIDTH = 120;
+
+export const DEFAULT_SESSION_SEARCH_LIMIT = 40;
+export const DEFAULT_SESSION_SEARCH_PREVIEW_SNIPPETS = 6;
+export const DEFAULT_SESSION_SEARCH_SHORTCUT = "f2";
+export const DEFAULT_SESSION_SEARCH_SUMMARY_INPUT_CHARS = 180_000;
+export const DEFAULT_SESSION_SEARCH_SUMMARY_MAX_TOKENS = 4096;
+export const DEFAULT_SESSION_SEARCH_CACHE_TTL_SECONDS = 0;
+export const SESSION_SEARCH_OVERLAY_HEIGHT_RATIO = 0.9;
+
+export const DEFAULT_AUTO_RENAME_MODEL = "openai-codex/gpt-5.4-mini";
+export const DEFAULT_AUTO_RENAME_FALLBACK_MODEL = "current";
+export const DEFAULT_AUTO_RENAME_INPUT_CHARS = 2000;
+export const DEFAULT_AUTO_RENAME_NAME_CHARS = 80;
+export const DEFAULT_AUTO_RENAME_MAX_TOKENS = 96;
+export const DEFAULT_AUTO_RENAME_TIMEOUT_MS = 12_000;
+export const DEFAULT_INPUT_BOTTOM_PADDING_LINES = 0;
+
+export const SESSION_TITLE_SYNC_INTERVAL_MS = 1000;
+
+// tmux trims plain leading/trailing spaces from pane-border-format output, so
+// use NBSP padding. Terminals render it as a space, but tmux keeps it visible.
+export const TMUX_SESSION_TITLE_PAD = "\u00a0";
+export const TMUX_SESSION_TITLE_BORDER_FORMAT = `${TMUX_SESSION_TITLE_PAD}#{pane_title}${TMUX_SESSION_TITLE_PAD}`;
+
+export const THINKING_LABEL_DEFAULT = "\ue28c ";
+
+export const QOL_COMPACTION_SYSTEM_PROMPT = "You summarize coding-agent sessions for continuation. Preserve exact technical facts, filenames, commands, constraints, decisions, blockers, and next actions. Do not invent details.";
+
+export const AUTO_RENAME_SYSTEM_PROMPT = "You create short, descriptive session names for coding-agent chats. Use 2-6 words in Title Case. Respond with only the name, no quotes, explanations, markdown, emoji, or trailing punctuation.";
+
+export const DEFAULT_AUTO_RENAME_PROMPT = `Generate a short, descriptive title for this Pi coding-agent session based on the first user message.
+
+Rules:
+- Use 2-6 words
+- Use Title Case
+- Be specific about the user's task or topic
+- Do not mention Pi unless Pi itself is the task
+- Return only the title
+
+First user message:
+{{message}}`;
+
+export const HANDOFF_SYSTEM_PROMPT = `You are a context transfer assistant. Given a conversation history and the user's goal for a new thread, generate a focused prompt that:
+
+1. Summarizes relevant context from the conversation (decisions made, approaches taken, key findings)
+2. Lists any relevant files that were discussed or modified
+3. Clearly states the next task based on the user's goal
+4. Is self-contained - the new thread should be able to proceed without the old conversation
+
+Format your response as a prompt the user can send to start the new thread. Be concise but include all necessary context. Do not include any preamble like "Here's the prompt" - just output the prompt itself.
+
+Example output format:
+## Context
+We've been working on X. Key decisions:
+- Decision 1
+- Decision 2
+
+Files involved:
+- path/to/file1.ts
+- path/to/file2.ts
+
+## Task
+[Clear description of what to do next based on user's goal]`;