@oh-my-pi/pi-mnemopi 16.0.4 → 16.0.5

package/CHANGELOG.md

@@ -2,6 +2,13 @@
 
 ## [Unreleased]
 
+## [16.0.5] - 2026-06-17
+
+### Fixed
+
+- Capped `sleep_consolidation` episodic rows at `maxEpisodeChars` (default 100KB, `MNEMOPI_MAX_EPISODE_CHARS`) so raw session transcripts cannot be stored and extracted as multi-megabyte episodes. ([#2869](https://github.com/can1357/oh-my-pi/issues/2869))
+- Skipped regex-only entity and pattern fact extraction for oversized raw transcripts so progress/log noise cannot flood MEMORIA with junk facts. ([#2868](https://github.com/can1357/oh-my-pi/issues/2868))
+
 ## [15.13.1] - 2026-06-15
 
 ### Added

package/dist/types/config.d.ts

@@ -33,6 +33,7 @@ export declare function apiEmbeddingsAvailable(env?: Env): boolean;
 export declare function workingMemoryMaxItems(env?: Env): number;
 export declare function workingMemoryTtlHours(env?: Env): number;
 export declare function episodicRecallLimit(env?: Env): number;
+export declare function maxEpisodeChars(env?: Env): number;
 export declare function sleepBatchSize(env?: Env): number;
 export declare function scratchpadMaxItems(env?: Env): number;
 export declare function recencyHalflifeHours(env?: Env): number;

package/dist/types/core/beam/types.d.ts

@@ -38,6 +38,7 @@ export interface BeamConfig {
     importanceWeight: number;
     useCloud: boolean;
     localLlmEnabled: boolean;
+    maxEpisodeChars: number;
 }
 export interface BeamMemoryOptions {
     sessionId?: string;

package/dist/types/core/entities.d.ts

@@ -1,4 +1,6 @@
 export declare const ENTITY_EXTRACTION_STOP_WORDS: ReadonlySet<string>;
+/** Maximum raw text size for regex-only entity extraction. */
+export declare const REGEX_EXTRACTION_MAX_INPUT_CHARS = 50000;
 export declare function levenshteinDistance(s1: string, s2: string): number;
 export declare function similarity(s1: string, s2: string): number;
 export declare function extractEntitiesRegex(text: string): string[];

package/package.json

@@ -1,7 +1,7 @@
 {
 	"type": "module",
 	"name": "@oh-my-pi/pi-mnemopi",
-	"version": "16.0.4",
+	"version": "16.0.5",
 	"description": "Local SQLite memory engine for Oh My Pi agents",
 	"homepage": "https://omp.sh",
 	"author": "Can Boluk",
@@ -39,9 +39,9 @@
 		"fmt": "biome format --write ."
 	},
 	"dependencies": {
-		"@oh-my-pi/pi-ai": "16.0.4",
-		"@oh-my-pi/pi-catalog": "16.0.4",
-		"@oh-my-pi/pi-utils": "16.0.4",
+		"@oh-my-pi/pi-ai": "16.0.5",
+		"@oh-my-pi/pi-catalog": "16.0.5",
+		"@oh-my-pi/pi-utils": "16.0.5",
 		"lru-cache": "11.5.1"
 	},
 	"peerDependencies": {

package/src/config.ts

@@ -126,6 +126,10 @@ export function episodicRecallLimit(env: Env = process.env): number {
 	return envInt("MNEMOPI_EP_LIMIT", 50000, env);
 }
 
+export function maxEpisodeChars(env: Env = process.env): number {
+	return Math.max(1, envInt("MNEMOPI_MAX_EPISODE_CHARS", 100000, env));
+}
+
 export function sleepBatchSize(env: Env = process.env): number {
 	return envInt("MNEMOPI_SLEEP_BATCH", 5000, env);
 }

package/src/core/beam/consolidate.ts

@@ -1,6 +1,7 @@
 import type { SQLQueryBindings } from "bun:sqlite";
 import { generateId, stableMemoryId } from "../../util/ids";
 import { aaakEncode } from "../aaak";
+import { REGEX_EXTRACTION_MAX_INPUT_CHARS } from "../entities";
 import { EpisodicGraph } from "../episodic-graph";
 import { heuristicExtractFacts } from "../extraction";
 import { clampVeracity } from "../veracity-consolidation";
@@ -57,6 +58,70 @@ const TIER2_DAYS = envInt("MNEMOPI_TIER2_DAYS", 30);
 const TIER3_DAYS = envInt("MNEMOPI_TIER3_DAYS", 180);
 const DEGRADE_BATCH_SIZE = envInt("MNEMOPI_DEGRADE_BATCH", 100);
 const TIER3_MAX_CHARS = envInt("MNEMOPI_TIER3_MAX_CHARS", 300);
+const DEFAULT_MAX_EPISODE_CHARS = 100_000;
+const SLEEP_SUMMARY_SEPARATOR = " | ";
+const SLEEP_TRUNCATION_MARKER = "\n[... sleep_consolidation episode truncated by maxEpisodeChars ...]";
+const PATTERN_FACT_EXTRACTION_MAX_INPUT_CHARS = REGEX_EXTRACTION_MAX_INPUT_CHARS;
+
+type SleepSummary = {
+	summary: string;
+	originalChars: number;
+	truncated: boolean;
+	maxChars: number;
+};
+type SleepChunk = {
+	items: Row[];
+	originalChars: number;
+};
+
+function normalizedMaxEpisodeChars(beam: BeamMemoryState): number {
+	const configured = Math.trunc(beam.config?.maxEpisodeChars ?? DEFAULT_MAX_EPISODE_CHARS);
+	return Number.isFinite(configured) && configured > 0 ? configured : DEFAULT_MAX_EPISODE_CHARS;
+}
+
+function markTruncated(content: string, maxChars: number): string {
+	if (maxChars <= 0) return "";
+	if (maxChars <= SLEEP_TRUNCATION_MARKER.length) return content.slice(0, maxChars);
+	const bodyChars = maxChars - SLEEP_TRUNCATION_MARKER.length;
+	return `${content.slice(0, bodyChars).trimEnd()}${SLEEP_TRUNCATION_MARKER}`;
+}
+
+function splitSleepItems(beam: BeamMemoryState, source: string, items: readonly Row[]): SleepChunk[] {
+	const maxChars = normalizedMaxEpisodeChars(beam);
+	const prefixChars = `[${source}] `.length;
+	const joinedLimit = Math.max(0, maxChars - prefixChars);
+	const chunks: SleepChunk[] = [];
+	let current: Row[] = [];
+	let currentChars = 0;
+
+	for (const item of items) {
+		const contentChars = (rowValue(item, "content") ?? "").length;
+		const separatorChars = current.length === 0 ? 0 : SLEEP_SUMMARY_SEPARATOR.length;
+		if (current.length > 0 && currentChars + separatorChars + contentChars > joinedLimit) {
+			chunks.push({ items: current, originalChars: currentChars });
+			current = [];
+			currentChars = 0;
+		}
+		current.push(item);
+		currentChars += (current.length === 1 ? 0 : SLEEP_SUMMARY_SEPARATOR.length) + contentChars;
+	}
+	if (current.length > 0) chunks.push({ items: current, originalChars: currentChars });
+	return chunks;
+}
+
+function buildSleepSummary(beam: BeamMemoryState, source: string, chunk: SleepChunk): SleepSummary {
+	const maxChars = normalizedMaxEpisodeChars(beam);
+	const prefix = `[${source}] `;
+	const joined = chunk.items.map(item => rowValue(item, "content") ?? "").join(SLEEP_SUMMARY_SEPARATOR);
+	const uncapped = `${prefix}${aaakEncode(joined)}`;
+	const truncated = uncapped.length > maxChars;
+	return {
+		summary: truncated ? markTruncated(uncapped, maxChars) : uncapped,
+		originalChars: chunk.originalChars,
+		truncated,
+		maxChars,
+	};
+}
 
 function isoNow(): string {
 	return new Date().toISOString();
@@ -440,6 +505,7 @@ export function extractAndStoreFacts(
 		decision: 0,
 	};
 	const text = String(content ?? "");
+	if (text.length > PATTERN_FACT_EXTRACTION_MAX_INPUT_CHARS) return counts;
 	for (const match of text.matchAll(
 		/(\d+(?:[.,]\d+)?)\s*(ms|sec|seconds?|minutes?|hours?|days?|weeks?|months?|%|KB|MB|GB|TB|rows?|columns?|roles?|features?|bugs?|commits?|cards?|users?|items?|tests?|APIs?|endpoints?|sprints?|tickets?)\b/gi,
 	)) {
@@ -878,26 +944,34 @@ export function sleep(beam: BeamMemoryState, dryRun = false): SleepResult {
 	const consolidatedIds: string[] = [];
 	let summariesCreated = 0;
 	for (const [source, items] of grouped) {
-		const lines = items.map(item => rowValue(item, "content") ?? "");
-		const ids = items.map(item => rowValue(item, "id")).filter((id): id is string => id !== null);
-		let scope = "session";
-		let validUntil: string | null = null;
-		for (const item of items) {
-			if (rowValue(item, "scope") === "global") scope = "global";
-			const itemValidUntil = rowValue(item, "valid_until");
-			if (itemValidUntil && (validUntil === null || itemValidUntil < validUntil)) validUntil = itemValidUntil;
-		}
-		const summary = `[${source}] ${aaakEncode(lines.join(" | "))}`;
-		if (!dryRun) {
-			consolidateToEpisodic(beam, summary, ids, "sleep_consolidation", 0.6, {
-				scope,
-				validUntil,
-				veracity: aggregateEpisodicVeracity(items.map(item => rowValue(item, "veracity") ?? "unknown")),
-				metadata: { original_count: items.length, source, llm_used: false },
-			});
+		for (const chunk of splitSleepItems(beam, source, items)) {
+			const ids = chunk.items.map(item => rowValue(item, "id")).filter((id): id is string => id !== null);
+			let scope = "session";
+			let validUntil: string | null = null;
+			for (const item of chunk.items) {
+				if (rowValue(item, "scope") === "global") scope = "global";
+				const itemValidUntil = rowValue(item, "valid_until");
+				if (itemValidUntil && (validUntil === null || itemValidUntil < validUntil)) validUntil = itemValidUntil;
+			}
+			const sleepSummary = buildSleepSummary(beam, source, chunk);
+			const metadata: Metadata = { original_count: chunk.items.length, source, llm_used: false };
+			if (sleepSummary.truncated) {
+				metadata.truncated = true;
+				metadata.original_chars = sleepSummary.originalChars;
+				metadata.max_chars = sleepSummary.maxChars;
+			}
+			const summary = sleepSummary.summary;
+			if (!dryRun) {
+				consolidateToEpisodic(beam, summary, ids, "sleep_consolidation", 0.6, {
+					scope,
+					validUntil,
+					veracity: aggregateEpisodicVeracity(chunk.items.map(item => rowValue(item, "veracity") ?? "unknown")),
+					metadata,
+				});
+			}
+			consolidatedIds.push(...ids);
+			summariesCreated++;
 		}
-		consolidatedIds.push(...ids);
-		summariesCreated++;
 	}
 	if (!dryRun) {
 		beam.db.run(

package/src/core/beam/index.ts

@@ -1,6 +1,6 @@
 import type { Database } from "bun:sqlite";
 import { existsSync } from "node:fs";
-import { ftsWeight, importanceWeight, vectorWeight } from "../../config";
+import { ftsWeight, importanceWeight, maxEpisodeChars, vectorWeight } from "../../config";
 import { closeQuietly, openDatabase } from "../../db";
 import { AnnotationStore } from "../annotations";
 import { EpisodicGraph } from "../episodic-graph";
@@ -68,6 +68,7 @@ const DEFAULT_CONFIG: BeamConfig = {
 	importanceWeight: 0.2,
 	useCloud: false,
 	localLlmEnabled: false,
+	maxEpisodeChars: 100_000,
 };
 
 function normalizeConfig(options: BeamMemoryOptions): BeamConfig {
@@ -82,6 +83,7 @@ function normalizeConfig(options: BeamMemoryOptions): BeamConfig {
 		importanceWeight: configured.importanceWeight ?? importanceWeight(),
 		useCloud,
 		localLlmEnabled: configured.localLlmEnabled ?? DEFAULT_CONFIG.localLlmEnabled,
+		maxEpisodeChars: configured.maxEpisodeChars ?? maxEpisodeChars(),
 	};
 }
 function autoMigrateAnnotations(db: Database, dbPath: string | undefined): void {

package/src/core/beam/types.ts

@@ -53,6 +53,7 @@ export interface BeamConfig {
 	importanceWeight: number;
 	useCloud: boolean;
 	localLlmEnabled: boolean;
+	maxEpisodeChars: number;
 }
 
 export interface BeamMemoryOptions {

package/src/core/entities.ts

@@ -118,6 +118,9 @@ const ENTITY_EXTRACTION_STOP_WORD_VALUES = [
 ] as const;
 
 export const ENTITY_EXTRACTION_STOP_WORDS: ReadonlySet<string> = new Set(ENTITY_EXTRACTION_STOP_WORD_VALUES);
+/** Maximum raw text size for regex-only entity extraction. */
+export const REGEX_EXTRACTION_MAX_INPUT_CHARS = 50_000;
+
 const ENTITY_PATTERNS: readonly RegExp[] = [
 	/@(\w{2,30})/g,
 	/#(\w{2,30})/g,
@@ -192,6 +195,7 @@ function isPureNumber(entity: string): boolean {
 
 export function extractEntitiesRegex(text: string): string[] {
 	if (typeof text !== "string" || text.length === 0) return [];
+	if (text.length > REGEX_EXTRACTION_MAX_INPUT_CHARS) return [];
 
 	const entities = new Set<string>();
 	for (const sourcePattern of ENTITY_PATTERNS) {