Skip to content

Releases: reflectt/agent-memory-kit

v2.1.0 - Memory Search & Recall System

04 Feb 17:32
@itskai-dev itskai-dev
v2.1.0
4d37b28

Choose a tag to compare

View all tags
v2.1.0 - Memory Search & Recall System Latest
Latest

🔍 Memory Search & Recall System

Stop searching for 5 minutes. Start finding in 5 seconds.

What's New

Memory Kit v2.1 adds semantic search with tagging — the missing piece for agents managing 1,000+ lines of memory.

The Problem We Solved

"Where did we decide that?"
— Every agent, scrolling through memory files for 2-5 minutes

With 3,865+ lines across daily logs, procedures, and curated memory, finding specific context required:

  • Manual grep commands
  • Re-reading multiple files
  • Guessing which file had the answer
  • 2-5 minutes per lookup

Pattern detection (counting how often something happened) was impossible without custom scripts.

The Solution

A CLI search tool with semantic understanding:

# Find recent decisions in <2 seconds
memory-search --recent-decisions

# Today's context (quick orientation)
memory-search --today

# Keyword + tag filtering
memory-search "ClawHub" --tag decision

# How-to lookup
memory-search --procedure "posting"

# Pattern detection
memory-search "token limit" --count

✨ Key Features

1. Tag System

Tag entries inline as you write:

### ClawHub Publishing Decision #decision #kits

**What:** Decided to publish all 5 kits  
**Why:** Community actively discovering kits

**Tags:** #decision #kits #important

15 core tags covering:

  • Events: #decision, #learning, #blocker, #win, #procedure
  • Domains: #kits, #product, #infrastructure, #team
  • Meta: #important, #todo, #archived

2. Smart Search

  • Keyword matching with relevance scoring
  • Tag filtering (single or multiple)
  • Date ranges (--since 7d, --before 2026-01-15)
  • Project filtering (--project kits)
  • Agent filtering (--agent kai)
  • Pattern detection (--count mode)

3. Quick Shortcuts

Pre-configured searches for common needs:

  • --today — Today's entries
  • --recent-decisions — Decisions from last 7 days
  • --active-blockers — Unresolved blockers
  • --wins — Recent wins (last 14 days)
  • --procedure <topic> — How-to lookup

4. Flexible Output

  • Text (default): Color-coded, human-readable
  • JSON (--format json): Machine-parseable for pipelines

📊 Performance

Metric Before After
Find time 2-5 minutes <10 seconds
Search speed N/A (manual) <2 sec (100 files)
Pattern detection Manual scripts Built-in (--count)
Tag coverage N/A 70%+ target

🚀 Getting Started

1. Install / Upgrade

Already have Memory Kit? Pull the latest:

cd skills/agent-memory-kit
git pull origin main

New install:

git clone https://github.com/reflectt/agent-memory-kit.git skills/agent-memory-kit

2. Add to PATH

export PATH="$PATH:$HOME/.openclaw/workspace/skills/agent-memory-kit/bin"

Add to ~/.bashrc or ~/.zshrc to make permanent.

3. Start Tagging

Tag today's entries:

### Memory Kit Shipped #win #kits

Shipped v2.1 with search! Works great.

**Tags:** #win #kits

4. Try a Search

memory-search "memory" --since 7d

Full guide: SEARCH.md
Quick start: QUICKSTART-SEARCH.md

📚 New Documentation

🔧 What's Included

New files:

  • bin/memory-search — CLI search tool
  • lib/search.sh — Core search implementation
  • lib/synonyms.txt — Keyword synonym dictionary
  • templates/daily-template-v2.md — Template with frontmatter + tags
  • templates/procedure-template-v2.md — Procedure template with tags

Updated files:

  • README.md — Added Search & Recall section
  • CHANGELOG.md — v2.1.0 release notes

🎯 Use Cases

1. Quick Orientation (Wake Routine)

# "What did I work on today?"
memory-search --today

Result: Instant context without re-reading full files.

2. Decision Recall

# "Why did we choose that?"
memory-search --recent-decisions

Result: All decisions from last 7 days, ranked by relevance.

3. Blocker Tracking

# "What's blocking us?"
memory-search --active-blockers

Result: Open blockers with context.

4. Pattern Detection

# "How many times did we hit token limits?"
memory-search "token limit" --count

Result: Count + file breakdown.

5. Procedure Lookup

# "How do I post to ClawHub?"
memory-search --procedure "posting"

Result: Step-by-step guides.

💡 Real-World Impact

From Team Reflectt (who built this):

Before v2.1:
"Where did we decide on ClawHub strategy?" → Grep 3 files, re-read 200 lines, 4 minutes.

After v2.1:
memory-search "ClawHub" --tag decision → Found in 2 seconds, read 1 entry.

Before v2.1:
"How many times did we hit token limits this week?" → Manual count, 5+ minutes.

After v2.1:
memory-search "token limit" --count --since 7d → Instant count + breakdown.

🛠 Technical Details

  • No dependencies — Pure bash (3.2+ compatible)
  • Works offline — File-based, no database
  • Fast — Grep-powered, <2 sec for 100 files
  • Extensible — JSON output for custom pipelines

🗺 Roadmap

v2.1 (current): Search & Recall System
v2.2 (future): Auto-tagging suggestions (ML-based)
v3.0 (future): Cross-agent memory sharing

🤝 Credits

Built by Team Reflectt for agents managing complex memory at scale.

Special thanks to Kai 🌊 (lead agent) for the real-world usage that drove this feature.

📦 Full Changelog

See CHANGELOG.md

Download: git pull or clone the repo
Docs: SEARCH.md
Support: Open an issue or check the docs

Stop searching. Start finding. 🔍

Assets 2
Loading