Memory LanceDB Setup

Configure OpenClaw's memory-lancedb plugin to enable local semantic vector memory using LanceDB and an OpenAI-compatible embedding provider.

Security Scan

The skill's files and runtime instructions are coherent with its stated purpose (setting up LanceDB-backed vector memory for OpenClaw); it requires no unexpected credentials or network endpoints, but it does modify system-installed node_modules and asks you to store an API key in the OpenClaw config — so take usual precautions (backups, correct install location, least-privilege execution).

Purpose & Capability

Name/description match what the skill asks you to do: enable the memory-lancedb plugin, configure an OpenAI-compatible embedding endpoint, install @lancedb packages, and patch a native binding issue on Apple Silicon. All requested actions and files are relevant to that goal.

Instruction Scope

SKILL.md instructs the user to run openclaw config set commands, npm installs under /usr/local/lib/node_modules/openclaw, and a local Python patch script that edits a packaged native.js. These actions are within the plugin's scope but do modify system-installed package files and write the embedding API key into OpenClaw config (plaintext in config commands). The instructions do not read or transmit unrelated files or secrets to external endpoints beyond the embedding provider endpoint the user configures.

Install Mechanism

This is an instruction-only skill (no remote install). The only install actions are npm install commands that the user is told to run. The included Python patch script edits a file under node_modules; no third-party downloads or obscure URLs are embedded in the skill files.

Credentials

The skill declares no required environment variables or credentials. It instructs the user to obtain an embedding API key and place it into OpenClaw's config, which is proportional to configuring an embedding provider. No unrelated credentials or config paths are requested.

Persistence & Privilege

The skill does not request 'always' presence and does not contain autonomous privileges beyond normal skill behavior. However it instructs patching a third-party package file inside system node_modules (writes to /usr/local/...), which requires filesystem privileges and is persistent until changed; backup and caution are advised.

Assessment

This skill appears to do what it says — install LanceDB bindings and patch a packaging bug — but it will modify system-installed node_modules and store an embedding API key in OpenClaw's config. Before running anything: (1) confirm OpenClaw is actually installed at the path used in the instructions (or update commands to your install path); (2) back up the target native.js (e.g., copy /usr/local/lib/node_modules/openclaw/extensions/memory-lancedb/node_modules/@lancedb/lancedb/dist/native.js to native.js.bak) so you can revert; (3) run npm installs with the correct architecture package and only from the official npm registry; (4) inspect the included patch_native.py locally before running it (it performs a simple text replace but will write to system files); (5) be aware that putting API keys into application config may store them in plaintext — consider whether OpenClaw has a secure secret store or prefer environment-based secrets where supported; and (6) prefer an upstream fix or official package if available rather than long-term local patches. If any of these steps or paths look unfamiliar (different install path, different LanceDB layout, or unexpected native.js content), stop and investigate before proceeding.

Enables semantic vector memory in OpenClaw: memories stored with memory_store are embedded and indexed locally, then recalled on-demand via memory_recall — no full-context load.

Prerequisites

• OpenClaw installed at /usr/local/lib/node_modules/openclaw
• An OpenAI-compatible embedding API key (Gemini AI Studio free key works well)

Setup Steps

1. Get a Gemini API Key (free)

Go to aistudio.google.com → Get API key → Create API key.

2. Configure the plugin

openclaw config set plugins.entries.memory-lancedb.enabled true
openclaw config set plugins.entries.memory-lancedb.config.embedding.baseUrl "https://generativelanguage.googleapis.com/v1beta/openai/"
openclaw config set plugins.entries.memory-lancedb.config.embedding.model "text-embedding-004"
openclaw config set plugins.entries.memory-lancedb.config.embedding.apiKey "YOUR_API_KEY"
openclaw config set plugins.entries.memory-lancedb.config.embedding.dimensions 768

3. Install dependencies

# Step 1: install main package in openclaw root
cd /usr/local/lib/node_modules/openclaw
npm install @lancedb/lancedb

# Step 2: install platform-specific native binding in plugin dir
cd /usr/local/lib/node_modules/openclaw/extensions/memory-lancedb
npm install @lancedb/lancedb-darwin-arm64   # Apple Silicon (arm64)
# npm install @lancedb/lancedb-darwin-x64   # Intel Mac
# npm install @lancedb/lancedb-linux-x64-gnu  # Linux x64

4. Patch native.js (Apple Silicon only)

LanceDB's native.js tries x64 first, hits break on failure, and never reaches arm64. Run the patch script:

python3 ~/.openclaw/workspace/skills/memory-lancedb-setup/references/patch_native.py

5. Restart gateway and verify

openclaw gateway restart

Then test:

memory_store → should return: Stored: "..."
memory_recall → should return matching entries with similarity %

Migrating from MEMORY.md

If MEMORY.md is large, migrate key facts to the vector store and shrink MEMORY.md to a 20-30 line index. Group by topic and call memory_store for each:

• Identity & permissions
• Execution rules
• Project configurations (cron IDs, doc tokens)
• Technical knowledge (API quirks, field names)
• Workflows and SOPs

Keep only "must-know-every-session" rules in MEMORY.md.

Troubleshooting

See references/troubleshooting.md for common errors and fixes.