nfemmanuel/iranti

v0.3.7 Breaking

This release includes 1 breaking change for platform teams planning a safe upgrade.

Published 1mo AI Agents & Assistants

View tool

✓ No known CVEs patched

Read the diff → Tool health → What is this tool? →

✓ No known CVEs patched in this version

Summary

AI summary

Fixed irrelevant profile-fact injection and optimized relevance ranking while removing expectedCallSequence.

Full changelog

v0.3.7 — Context efficiency & attend quality

Fixed

Attend profile-fact injection — classifyMemoryScope regex matched nearly all English (/\b(my|me|i)\b/), injecting irrelevant user profile facts every turn. Now uses 5 specific patterns for genuine personal-data questions only.
Observe relevance ranking — Added token-overlap relevance scoring and profile-fact penalty (-50 on zero overlap) so top-N slots go to contextually relevant facts.
next_step O(N²) accumulation — Each checkpoint concatenated the entire prior chain. Now does a clean overwrite; history preserved in checkpoint_summary.

Changed

IRANTI.md — Full protocol now lives in a host-neutral IRANTI.md file generated by setup commands. CLAUDE.md slimmed to a 3-line pointer (~80 tokens, down from ~800). Saves ~720 tokens per API turn.
Removed expectedCallSequence from attend responses (~300 tokens/call savings). Protocol delivery via IRANTI.md instead.

Breaking Changes

Removed `expectedCallSequence` from attend responses

View diff on GitHub

Weekly OSS security release digest.

The CVE patches and breaking changes that affected production tools this week. One email, every Sunday.

No spam, unsubscribe anytime.

Share this release

Share on X Share on Bluesky

Track nfemmanuel/iranti

Get notified when new releases ship.

About nfemmanuel/iranti

Persistent shared memory for AI coding agents. Stores facts as `entity/key/value` triples with hybrid semantic search, task checkpoints, and conflict resolution — shared across Claude Code, Codex CLI, and GitHub Copilot.

All releases →