chat-memory（聊天记忆）

Chat Memory

Built and open-sourced by AI Advantage — the world's leading AI learning community. We teach 40,000+ people how to actually use AI. This skill is how our own agents remember everything. Want to learn more? Join us at aiadvantage.ai.

Solve the #1 problem with long-running AI agents: knowledge loss after context compaction.

The Problem

When sessions compact (summarize old messages to free context), specific details are lost: names, decisions, file paths, reasoning. The agent retains a summary but loses the ability to recall "What exactly did Annika say?" or "When did we decide to use v6 format?"

Most memory skills on ClawHub are just SKILL.md instructions — "write stuff to MEMORY.md." That's not a solution. This skill ships real scripts that do real work.

The Solution: Three-Layer Memory Architecture

Layer 1: MEMORY.md          — Curated long-term memory (human-edited)
Layer 2: SESSION-GLOSSAR.md — Auto-generated structured index (people/projects/decisions/timeline)
Layer 3: memory/sessions/   — Full session transcripts as searchable Markdown

All three layers live under memory/ and are automatically vectorized by OpenClaw's memory search, creating a navigational hierarchy: glossary finds the right session, session provides the details.

Setup (run once)

Step 1: Convert existing sessions to Markdown

python3 scripts/session-to-memory.py

This scans all JSONL session logs in ~/.openclaw/agents/*/sessions/ and converts them to memory/sessions/session-YYYY-MM-DD-HHMM-*.md. Truncates long assistant responses to 2KB, skips system messages, tracks state to avoid re-processing.

Options:

• --new — Only convert sessions not yet processed (for incremental runs)
• --agent main — Specify agent ID (default: main)

Step 2: Build the glossary

python3 scripts/build-glossary.py

Scans all session transcripts and builds memory/SESSION-GLOSSAR.md with:

• People — Who was mentioned, in how many sessions, date ranges
• Projects — Which projects discussed, with relevant topic tags
• Topics — Categorized themes (Email Drafts, Website Build, Security, etc.)
• Timeline — Per-day summary (session count, people, topics)
• Decisions — Extracted decision-like statements with dates

Options:

• --incremental — Only process new sessions (uses cached scan state)

Step 3: Set up cron jobs for auto-updates

Create two cron jobs (use a cheap model like Gemini Flash):

Job 1: Session sync + glossary rebuild (every 4-6 hours)

Task: Run `python3 scripts/session-to-memory.py --new` then
      `python3 scripts/build-glossary.py --incremental`.
      Report how many new sessions were converted and indexed.

Optional Job 2: Pre-compaction memory flush check Already built into AGENTS.md by default — just ensure the agent writes to memory/YYYY-MM-DD.md before each compaction.

Customizing Entity Detection

Edit scripts/build-glossary.py to add your own known people and projects:

KNOWN_PEOPLE = {
    "alice": "Alice Smith — Project Manager",
    "bob": "Bob Jones — CTO",
}

KNOWN_PROJECTS = {
    "website-redesign": "Website Redesign — Q1 Initiative",
    "api-migration": "API Migration — v2 to v3",
}

The glossary also detects topics via regex patterns. Add new patterns in the topic_patterns dict for your domain.

How It Works With memory_search

Once set up, memory_search("Alice project decision") will find:

1. The glossary entry for Alice (which sessions she appears in)
2. The actual session transcript where the decision was discussed
3. Any MEMORY.md entry about Alice

This gives the agent a navigation layer (glossary) plus detail access (transcripts) — much better than either alone.

File Structure After Setup

memory/
├── MEMORY.md                    — Curated (you maintain this)
├── SESSION-GLOSSAR.md           — Auto-generated index
├── YYYY-MM-DD.md                — Daily notes
├── .glossary-state.json         — Glossary builder state
├── .glossary-scans.json         — Cached scan results
└── sessions/
    ├── .state.json              — Converter state
    ├── session-2026-01-15-0830-abc123.md
    ├── session-2026-01-15-1200-def456.md
    └── ...

Cron Memory Optimizer

Cron jobs run in isolated sessions with zero memory context. The optimizer analyzes your cron jobs and suggests memory-enhanced versions:

python3 scripts/cron-optimizer.py

This scans ~/.openclaw/cron/jobs.json, identifies jobs that would benefit from memory context, and generates memory/cron-optimization-report.md with before/after prompts and implementation guidance.

Example optimization:

Original: "Run daily research scout..."
Enhanced: "Before starting: Use memory_search to find recent context about research activities. Check memory/SESSION-GLOSSAR.md for relevant people, projects, and recent decisions. Then proceed with the original task using this context.

Run daily research scout..."

The script is conservative (suggests only, never auto-modifies) and skips monitoring jobs that don't need context.

Sharing Memory Context with Cron Jobs, Subagents & Telegram Sessions

One of the biggest challenges in multi-session AI systems is context isolation. Here's how to share memory context across different execution environments:

For Cron Jobs

The problem: Cron jobs run in isolated sessions with zero memory context, making them blind to recent activities, people, and decisions.

The solution: Prepend a "memory preamble" to cron job prompts that instructs the agent to search memory before starting:

Before starting this task: Use memory_search to find recent context relevant to this task. Check memory/SESSION-GLOSSAR.md for people, projects, and recent decisions that may be relevant. Then proceed with the original task using this context.

The cron-optimizer.py script analyzes your existing cron jobs and automatically suggests which ones would benefit from memory context. It generates a detailed report with before/after prompts.

Example transformation:

Before: "You are a research scout. Find AI tools and report findings..."

After:  "Before starting this task: Use memory_search to find recent context relevant to this task. Check memory/SESSION-GLOSSAR.md for people, projects, and recent decisions that may be relevant. Then proceed with the original task using this context.

You are a research scout. Find AI tools and report findings..."

For Subagents (sessions_spawn)

The problem: Subagents start with empty context and don't know about recent activities or ongoing projects.

The solution: Include memory instructions in the task prompt when spawning subagents:

Before starting: Use memory_search("relevant keywords") to find recent context. 
Check memory/SESSION-GLOSSAR.md for people, projects, decisions.
Check MEMORY.md for long-term context.
Then proceed:

[your actual task...]

Tips:

• Be specific with memory_search keywords for best results
• Include both recent (SESSION-GLOSSAR.md) and long-term (MEMORY.md) context
• Consider what the subagent needs to know to do its job effectively

For Telegram Group Sessions

The problem: Group sessions share the workspace but don't automatically know about the memory system or recent activities discussed in other sessions.

The solution: Two approaches depending on your setup:

Method 1: Push context via sessions_send

# From main session, send relevant context to group session
sessions_send telegram-group "Memory context: Recent project status - [summary]"

Method 2: Add memory awareness to AGENTS.md Add guidance to your AGENTS.md so group sessions know to search memory:

## Group Chat Guidelines
When answering questions about past work or ongoing projects, 
always use memory_search first to check for relevant context.

Tips:

• Group sessions can access the memory system if they know to use it
• Include memory search instructions in your group-specific agent guidelines
• Consider pushing critical updates from main to group sessions when decisions are made

For Knowledge Bases (Vectorized Databases)

If you have custom vectorized knowledge bases (e.g., using sentence-transformers), make them accessible across sessions:

Method 1: Query scripts

# Create a query script that any session can call
python3 scripts/query-knowledge-base.py "search terms"

Method 2: Workspace storage

# Store the database in workspace for universal access
mkdir -p knowledge-base/
# Include database path in task prompts
"Use the knowledge base at ./knowledge-base/db.pkl for additional context..."

Method 3: Integration prompts Include the script path in cron job and subagent prompts:

Before starting: Run `python3 scripts/query-knowledge-base.py "project context"` 
for additional background. Then proceed with the task.

The key is making knowledge discovery automatic and standardized across all execution contexts — main session, cron jobs, subagents, and group sessions should all follow the same memory-aware patterns.

Tips

• Run the full rebuild (python3 scripts/build-glossary.py without --incremental)

occasionally to pick up improvements to entity detection

• The glossary is most useful when KNOWN_PEOPLE and KNOWN_PROJECTS are populated —

spend 5 minutes adding your key contacts and projects

• For agents that run 24/7, the cron job keeps everything current automatically
• Session transcripts can get large (our 297 sessions = 24MB) — this is fine,

OpenClaw's vector search handles it efficiently

• Use the cron optimizer after setting up memory to enhance existing automation

Why This Exists

We run OpenClaw agents 24/7 for real work — client projects, research pipelines, content production. After a week we had 300+ sessions and our agents kept forgetting critical details after compaction. We built this to fix it, and it worked so well we open-sourced it.

What makes this different from other memory skills:

• ✅ Real Python scripts — not just "instructions for the agent"
• ✅ Three-layer architecture — curated + auto-glossary + raw transcripts
• ✅ Cron automation — runs in the background, zero manual work
• ✅ Glossary with entity detection — people, projects, decisions, timeline
• ✅ Cron optimizer — makes your existing cron jobs context-aware
• ✅ Clean security score — no suspicious flags, no external dependencies
• ✅ Battle-tested — 300+ sessions, running in production daily

Built with 🔥 by AI Advantage — Join 40,000+ people learning to build with AI. We don't just teach AI — we build with it every day. This skill is proof.