PostgreSQL + pgvector backend for the Mnemosyne agentic memory library.
This library provides a Mnemosyne.GraphBackend implementation backed by
PostgreSQL with pgvector for vector similarity search. It stores memory
graph nodes, edges, and metadata using Ecto schemas and supports
multi-tenant isolation via configurable table prefixes.
Setup
# Set as default backend in your supervision tree
{Mnemosyne.Supervisor,
config: config,
llm: MyApp.LLM,
embedding: MyApp.Embedding,
backend: {MnemosynePostgres.Backend, repo: MyApp.Repo}}
# Open repos -- repo_id is injected from the first argument
Mnemosyne.open_repo("my-project")
Mnemosyne.open_repo("my-project", tenant_id: "org-123")