← Back to Skills Marketplace
MeiliSearch Memory
by
Eric Kariuki
· GitHub ↗
· v2.0.0
· MIT-0
34
Downloads
0
Stars
0
Active Installs
2
Versions
Install in OpenClaw
/install meili-memory
Description
Local full-text search for OpenClaw memory recall using MeiliSearch. No API keys, no cloud services — runs entirely on your server.
README (SKILL.md)
MeiliSearch Memory
Local full-text search for memory recall. No API keys, no cloud services — runs entirely on your server.
Architecture
Three-tier indexing system that mimics human memory:
- Core — MEMORY.md sections (highest importance, always current)
- Working — Daily notes from the last 7 days (recent context)
- Archive — Older daily notes (lower priority)
Search queries hit all tiers in order: core → working → archive. Results are ranked by tier priority and importance score.
Features
- Synonym expansion — "debts" also matches "liabilities", "rent", "KES", etc.
- Relationship tags — Each chunk is auto-tagged (#finance/debts, #project/molthub, #identity, etc.)
- Tiered search — Query core only for fast answers, or all tiers for deep recall
- Auto-distillation — Weekly cron extracts key facts from daily notes into MEMORY.md
- Typo tolerance — Built-in fuzzy matching via MeiliSearch
Search
bash scripts/search.sh "your query" [limit] [tier]
Examples:
bash scripts/search.sh "debts" # Search all tiers, top 5
bash scripts/search.sh "projects" 3 # Top 3 from all tiers
bash scripts/search.sh "Eric" 5 core # Core only (fast)
Indexing
bash scripts/indexer.sh # Incremental update
bash scripts/indexer.sh --full # Full reindex (wipes and rebuilds)
Distillation
bash scripts/distill.sh # Extract facts → MEMORY.md
bash scripts/distill.sh --dry-run # Preview what would be added
Cron Jobs
- Hourly — Re-index new daily notes
- Weekly (Sunday 2 AM) — Distill facts into MEMORY.md
Service
systemctl status meilisearch
journalctl -u meilisearch -f
Usage Guidance
Review before installing. This appears to be a local-only memory search and indexing skill, not an exfiltration tool, but it should be treated as Review because it may make secrets and personal details more persistent and searchable. Do not use it on notes containing passwords, API keys, tokens, private contact information, or sensitive financial data unless you first remove or change the distillation rules, move credentials out of source code, and ensure temporary index files are protected.
Capability Tags
Capability Assessment
Purpose & Capability
Local indexing and search fit the stated purpose, and distillation is disclosed, but the implementation treats credentials, tokens, keys, contact data, URLs, and endpoints as memory-worthy facts and can append them to MEMORY.md.
Instruction Scope
The skill is broader than passive search: it reads OpenClaw memory files, indexes them, expands broad synonym groups, appends distilled facts, and triggers reindexing. Some of this is documented, but sensitive-data retention is not clearly scoped or guarded.
Install Mechanism
No installer or hidden startup mechanism is present in the artifacts. The documentation describes hourly and weekly cron jobs, but the package does not itself create them.
Credentials
Using localhost MeiliSearch is proportionate for local memory search, but the scripts hardcode a bearer/master key while the description says no API keys, and the indexer writes full memory document batches to predictable /tmp files.
Persistence & Privilege
The distillation script can append extracted facts to /root/.openclaw/workspace/MEMORY.md and reindex them, creating durable retention of potentially sensitive information without a required preview or approval step.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install meili-memory - After installation, invoke the skill by name or use
/meili-memory - Provide required inputs per the skill's parameter spec and get structured output
Version History
v2.0.0
Version 2.0.0 — Major upgrade with tiered memory architecture and new features
- Introduces a three-tier indexing system: core, working, and archive, to mimic human memory and prioritize search results.
- Adds synonym expansion, relationship/tagging, and tiered search options in search queries.
- Implements auto-distillation: weekly extraction of key facts from notes into MEMORY.md.
- Enhances scripts: new distillation script, improved search and indexing with tier and full rebuild options.
- Provides support for typo tolerance and synonym management using a synonyms.json file.
v1.0.0
- Initial release of meili-memory skill.
- Provides local full-text search for OpenClaw memory recall using MeiliSearch.
- No API keys or cloud services required; operates entirely on your server.
- Supports indexing of MEMORY.md (by section) and daily notes (by 1500 character chunks).
- Includes scripts for searching and indexing, with automated hourly indexing via cron.
- Easy service management instructions provided for MeiliSearch.
Metadata
Frequently Asked Questions
What is MeiliSearch Memory?
Local full-text search for OpenClaw memory recall using MeiliSearch. No API keys, no cloud services — runs entirely on your server. It is an AI Agent Skill for Claude Code / OpenClaw, with 34 downloads so far.
How do I install MeiliSearch Memory?
Run "/install meili-memory" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is MeiliSearch Memory free?
Yes, MeiliSearch Memory is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does MeiliSearch Memory support?
MeiliSearch Memory is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created MeiliSearch Memory?
It is built and maintained by Eric Kariuki (@enjuguna); the current version is v2.0.0.
More Skills