Aister Vector Memory

# Vector Memory Skill

Vector memory for Aister — search by meaning, not by grep!

## Description

Vector memory using PostgreSQL + pgvector + e5-large-v2. Enables searching information by MEANING, not just keywords.

## Environment Variables

**Required:**
- `VECTOR_MEMORY_DB_PASSWORD` — PostgreSQL password for database access

**Optional:**
| Variable | Default | Description |
|----------|---------|-------------|
| `VECTOR_MEMORY_DB_HOST` | `localhost` | PostgreSQL server host |
| `VECTOR_MEMORY_DB_PORT` | `5432` | PostgreSQL server port |
| `VECTOR_MEMORY_DB_NAME` | `vector_memory` | Database name |
| `VECTOR_MEMORY_DB_USER` | `aister` | Database user |
| `EMBEDDING_SERVICE_URL` | `http://127.0.0.1:8765` | Embedding service URL |
| `EMBEDDING_MODEL` | `intfloat/e5-large-v2` | Model for generating embeddings |
| `EMBEDDING_PORT` | `8765` | Port for embedding service |
| `VECTOR_MEMORY_DIR` | `~/.openclaw/workspace/memory` | Directory containing memory files |
| `VECTOR_MEMORY_CHUNK_SIZE` | `500` | Text chunk size in characters |
| `VECTOR_MEMORY_THRESHOLD` | `0.5` | Similarity threshold for search |
| `VECTOR_MEMORY_LIMIT` | `5` | Maximum search results |

## Features

- **Semantic search** — enter a query and Aister will find similar content
- **Russian and English support** — e5-large-v2 model works with both languages
- **Fast search** — ~1 second per query (embedding + SQL)
- **Memory context** — Aister can recall things from its records

## Usage

### Search

```
/search_memory <query>
```

Examples:
```
/search_memory my communication style
/search_memory what I did today
/search_memory Moltbook settings
```

### Reindex

```
/reindex_memory
```

This reads all memory files (MEMORY.md, IDENTITY.md, USER.md, etc.) and updates the vector database.

## How it works

1. When Aister remembers something, it splits the text into chunks
2. Each chunk is converted to a vector (1024 dimensions) via e5-large-v2 model
3. Vectors are stored in PostgreSQL with pgvector ex...

# Vector Memory for Aister

Vector memory for semantic search instead of grep! 🧠

## Description

Vector Memory for Aister — smart search system using PostgreSQL + pgvector + e5-large-v2. Finds information by meaning, not just keywords.

**Key features:**
- ✅ **Semantic search** — enter a query, find relevant content
- ✅ **Russian and English support** — e5-large-v2 understands both languages
- ✅ **Fast search** — ~1 second per query
- ✅ **Memory context** — recall things from records
- ✅ **Auto-save** — thoughts are automatically saved to vector memory

## Warnings

> **Important before installation:**
> - **Network:** First run will download e5-large-v2 model (~1.3GB) from HuggingFace
> - **Privileges:** Requires root for apt/dnf and PostgreSQL superuser
> - **Security:** Configure your own passwords, don't use examples

## How it works

1. **Indexing** — text is split into 500-character chunks
2. **Vectorization** — each chunk is converted to a vector (1024 dimensions) via e5-large-v2
3. **Storage** — vectors are stored in PostgreSQL with pgvector extension
4. **Search** — query is vectorized and similarity is found via cosine distance

## Usage

### Installation

Full instructions in [INSTALL.md](INSTALL.md).

**Option A: Docker (Recommended for isolation)** — see SKILL.md for docker-compose setup.

**Option B: Quick start (bare metal):**
```bash
# 1. Create venv and install dependencies
python3 -m venv ~/.openclaw/workspace/vector_memory_venv
source ~/.openclaw/workspace/vector_memory_venv/bin/activate
pip install flask psycopg2-binary requests sentence-transformers numpy

# 2. Configure environment variables (including DB password!)
mkdir -p ~/.config/vector-memory
cat > ~/.config/vector-memory/env << 'EOF'
export VECTOR_MEMORY_DB_PASSWORD="YOUR_SECURE_PASSWORD"
EOF
chmod 600 ~/.config/vector-memory/env

# 3. Start embedding service (first run downloads ~1.3GB)
source ~/.config/vector-memory/env
~/.openclaw/workspace/vector_memory_venv/bin/python3 ~/.ope...