aister-vector-memory
by alekhm
Search & ResearchBrief
Vector memory for Aister — search by meaning, not by grep!
Description
What This Skill Does
Adds semantic search to Aister's memory using PostgreSQL with pgvector and the e5-large-v2 embedding model. Queries match by meaning rather than exact keywords. Supports both English and Russian.
Finds relevant memory entries even when the query shares no words with the stored text, unlike grep or keyword search.
When to Use It
- Finding past notes when you remember the topic but not the exact wording
- Searching stored preferences and settings by description
- Recalling what was written about a project without knowing precise terms
- Looking up identity or behavior notes by concept
- Retrieving context from memory files before starting a new conversation
Example Workflow
User asks: /search_memory my communication style
- 1
Converts the query to a 1024-dimension vector using e5-large-v2
- 2
Runs cosine similarity search against stored vectors in PostgreSQL
- 3
Filters results above the 0.5 similarity threshold
- 4
Returns up to 5 matching memory chunks with source context
Excerpts from MEMORY.md or USER.md describing the user's communication preferences, ranked by semantic similarity
Requirements
- PostgreSQL 16 with pgvector extension (requires root or sudo to install)
- VECTOR_MEMORY_DB_PASSWORD environment variable set to the PostgreSQL user password
Installation
clawhub install alekhm/aister-vector-memoryRun the CLI command, or paste the prompt into Claude Code, Cursor, Codex or whichever coding agent you use.