@oh-my-pi/pi-mnemopi 18.8.7 → 18.8.8

package/CHANGELOG.md

@@ -2,6 +2,12 @@
 
 ## [Unreleased]
 
+## [18.8.8] - 2026-10-10
+
+### Fixed
+
+- Fixed long retained transcripts exhausting local memory extraction models by limiting each extraction input to 8192 characters while keeping the opening context and newest turns ([#14956](https://github.com/can1357/oh-my-pi/issues/14956)).
+
 ## [18.8.7] - 2026-10-09
 
 ### Fixed

package/dist/types/core/text-window.d.ts

@@ -0,0 +1,4 @@
+/** Default character budget for a single embedding or fact-extraction input. */
+export declare const DEFAULT_INPUT_CHARS = 8192;
+/** Preserve the opening context and newest turns of an oversized chronological input. */
+export declare function clipToWindow(text: string, max: number): string;

package/package.json

@@ -1,7 +1,7 @@
 {
 	"type": "module",
 	"name": "@oh-my-pi/pi-mnemopi",
-	"version": "18.8.7",
+	"version": "18.8.8",
 	"description": "Local SQLite memory engine for omp agents",
 	"homepage": "https://omp.sh",
 	"author": {
@@ -42,10 +42,10 @@
 		"fmt": "oxfmt --no-error-on-unmatched-pattern 'src/**/*.{ts,tsx}' '{test,bench,examples,scripts}/**/*.ts' '*.ts'"
 	},
 	"dependencies": {
-		"@oh-my-pi/pi-ai": "18.8.7",
-		"@oh-my-pi/pi-catalog": "18.8.7",
-		"@oh-my-pi/pi-natives": "18.8.7",
-		"@oh-my-pi/pi-utils": "18.8.7"
+		"@oh-my-pi/pi-ai": "18.8.8",
+		"@oh-my-pi/pi-catalog": "18.8.8",
+		"@oh-my-pi/pi-natives": "18.8.8",
+		"@oh-my-pi/pi-utils": "18.8.8"
 	},
 	"peerDependencies": {
 		"fastembed": "3.0.0",

package/src/core/embeddings.ts

@@ -21,6 +21,7 @@ import {
 	mnemopiDebugEnabled,
 	resolveEmbeddingProvider,
 } from "./runtime-options";
+import { clipToWindow, DEFAULT_INPUT_CHARS } from "./text-window";
 
 export type { EmbeddingOutput } from "./runtime-options";
 export { cosineSimilarity } from "./vector-math";
@@ -166,29 +167,7 @@ function effectiveMaxInputChars(): number {
 	if (override !== undefined) return Math.max(0, Math.trunc(override));
 	const envValue = Number.parseInt($env.MNEMOPI_EMBEDDING_MAX_INPUT_CHARS ?? "", 10);
 	if (Number.isFinite(envValue) && envValue >= 0) return envValue;
-	return 8192;
-}
-
-/** Elision marker injected between the retained head and tail of an oversized input. */
-const EMBEDDING_ELISION_MARKER = "\n\n[...]\n\n";
-
-/**
- * Right-clip a single oversized input to {@link max} chars while preserving
- * both ends. Retention transcripts are chronological (oldest → newest), so a
- * naive `slice(0, max)` would drop the most recent — and most semantically
- * loaded — turns once a session passed the cap, leaving every later retained
- * episode with essentially the same prefix vector. Keeping a head/tail split
- * lets the embedding capture the topic setup at the start AND the latest
- * exchanges at the end. Falls back to a tail-only clip when `max` is too
- * small to fit the elision marker plus a useful slice on either side.
- */
-function clipToWindow(text: string, max: number): string {
-	if (text.length <= max) return text;
-	if (max <= EMBEDDING_ELISION_MARKER.length + 16) return text.slice(text.length - max);
-	const budget = max - EMBEDDING_ELISION_MARKER.length;
-	const headLen = budget >>> 1;
-	const tailLen = budget - headLen;
-	return text.slice(0, headLen) + EMBEDDING_ELISION_MARKER + text.slice(text.length - tailLen);
+	return DEFAULT_INPUT_CHARS;
 }
 
 /**

package/src/core/extraction.ts

@@ -10,6 +10,7 @@ import {
 	type RemoteLlmOptions,
 } from "./local-llm";
 import { getMnemopiRuntimeOptions } from "./runtime-options";
+import { clipToWindow, DEFAULT_INPUT_CHARS } from "./text-window";
 
 const TRUE_VALUES: Record<string, true> = { "1": true, true: true, yes: true, on: true };
 
@@ -378,7 +379,8 @@ export async function extractFactCategories(
 	if (typeof text !== "string" || text.trim() === "") {
 		return emptyFactCategories();
 	}
-	const prompt = buildExtractionPrompt(text);
+	const input = clipToWindow(text, DEFAULT_INPUT_CHARS);
+	const prompt = buildExtractionPrompt(input);
 
 	// Configured completion (host-injected runtime LLM, e.g. the coding-agent's smol
 	// or a local on-device model). Mirrors consolidation's precedence: when a
@@ -389,7 +391,7 @@ export async function extractFactCategories(
 		try {
 			const raw = await callConfiguredCompletion(prompt, 0, {
 				maxTokens: llmMaxTokens(),
-				task: { kind: "memory-extraction", input: text },
+				task: { kind: "memory-extraction", input },
 			});
 			if (typeof raw === "string" && raw.trim() !== "") {
 				const extracted = parseExtractedFactCategories(raw);
@@ -407,7 +409,7 @@ export async function extractFactCategories(
 			console.warn(`extractFacts: configured completion raised: ${safeForLog(exc)}`);
 			return emptyFactCategories();
 		}
-		return localFallback(prompt, text, diag);
+		return localFallback(prompt, input, diag);
 	}
 
 	try {
@@ -424,7 +426,7 @@ export async function extractFactCategories(
 				}
 			}
 			diag.recordNoOutput("host");
-			return localFallback(prompt, text, diag);
+			return localFallback(prompt, input, diag);
 		}
 	} catch (exc) {
 		diag.recordAttempt("host");
@@ -436,7 +438,7 @@ export async function extractFactCategories(
 
 	if (!llmAvailable()) {
 		diag.recordAttempt("local");
-		const heuristic = heuristicExtractFacts(text);
+		const heuristic = heuristicExtractFacts(input);
 		if (heuristic.length > 0) {
 			diag.recordSuccess("local", heuristic.length);
 			diag.recordCall({ succeeded: true });
@@ -465,7 +467,7 @@ export async function extractFactCategories(
 		console.warn(`extractFacts: remote LLM raised: ${safeForLog(exc)}`);
 	}
 
-	return localFallback(prompt, text, diag);
+	return localFallback(prompt, input, diag);
 }
 
 /** Extract legacy flat fact strings from text. */

package/src/core/text-window.ts

@@ -0,0 +1,14 @@
+/** Default character budget for a single embedding or fact-extraction input. */
+export const DEFAULT_INPUT_CHARS = 8192;
+
+const ELISION_MARKER = "\n\n[...]\n\n";
+
+/** Preserve the opening context and newest turns of an oversized chronological input. */
+export function clipToWindow(text: string, max: number): string {
+	if (text.length <= max) return text;
+	if (max <= ELISION_MARKER.length + 16) return text.slice(text.length - max);
+	const budget = max - ELISION_MARKER.length;
+	const headLen = budget >>> 1;
+	const tailLen = budget - headLen;
+	return text.slice(0, headLen) + ELISION_MARKER + text.slice(text.length - tailLen);
+}