workspace-memory

name	workspace-memory
description	Semantic search and memory retrieval using txtai. Use when searching through workspace files, finding relevant documents, or answering questions about prior work, decisions, or content stored in the workspace. Provides semantic search across all indexed text files (markdown, code, configs, etc.).

txtai Memory Search

Semantic search for workspace files using txtai embeddings. Replaces manual file searching with AI-powered semantic retrieval.

Quick Start

1. Index the Workspace (one-time setup)

cd scripts/
python index_workspace.py

This indexes all text files in the workspace (see SKILLS.md for workspace root).

2. Search

# Basic search
python search_memory.py "what was the config issue"

# With options
python search_memory.py "agent personality" --limit 3 --full

Python API Usage

For integration into other scripts:

import sys
sys.path.insert(0, "<python-packages-path>")  # See SKILLS.md for path

from txtai import Embeddings

# Load index
with Embeddings().load("<txtai-index-path>") as embeddings:  # See SKILLS.md for path
    # Search
    results = embeddings.search("your query", limit=5)
    for result in results:
        print(f"{result['id']}: {result['score']:.3f}")

Re-indexing

Re-run index_workspace.py after:

• Adding new files
• Major content changes
• Wanting fresh embeddings

The index is stored at the txtai index path (see SKILLS.md).

Indexed File Types

• Documentation: .md, .txt
• Code: .py, .js, .ts, .rs, .go, etc.
• Configs: .json, .yaml, .yml, .toml
• Scripts: .sh, .zsh
• Data: .sql, .graphql, .proto

When to Use

• Memory questions: "What did we decide about...", "When did..."
• Finding files: "Where is the config for...", "Show me the script that..."
• Context retrieval: Prior work, decisions, code patterns
• Semantic search: Concepts, not just keywords

Scripts

Workspace Memory (Your Files)

• scripts/index_workspace.py - Build/update workspace index
• scripts/search_memory.py - Query workspace index

OpenClaw Documentation

• scripts/index_openclaw.py - Build/update OpenClaw source index
• scripts/search_openclaw.py - Query OpenClaw documentation/code

Use OpenClaw search for questions about OpenClaw internals, APIs, or how to implement features.

Usage Examples

Use the Python binary from SKILLS.md to run scripts. Example:

# Search your workspace memory
<python-binary> search_memory.py "config issue" --limit 5

# Search OpenClaw documentation
<python-binary> search_openclaw.py "how to create a skill" --limit 3

# Re-index workspace after changes
<python-binary> index_workspace.py

# Re-index OpenClaw source (weekly)
<python-binary> index_openclaw.py

Notes

• Uses sentence-transformers/all-MiniLM-L6-v2 for embeddings
• Content stored in SQLite (no external DB needed)
• Skips: .git, node_modules, venv, __pycache__, etc.