leocrm/app/plugins/builtins/agent_memory/models.py

"""AgentMemory model for persistent agent memory with pgvector embeddings."""

from __future__ import annotations

import uuid
from typing import Any

from sqlalchemy import Index, String, Text
from sqlalchemy.dialects.postgresql import JSONB
from sqlalchemy.dialects.postgresql import UUID as PGUUID
from sqlalchemy.orm import Mapped, mapped_column

from app.core.db import Base, TenantMixin
from app.models.owned_mixin import OwnedMixin


class AgentMemory(Base, TenantMixin, OwnedMixin):
    """Persistent agent memory with semantic search via pgvector.

    Stores agent memories (facts, context, learned patterns) with
    vector embeddings for semantic retrieval.
    """

    __tablename__ = "agent_memories"
    __table_args__ = (
        Index("ix_agent_memories_tenant_agent", "tenant_id", "agent_id"),
        Index("ix_agent_memories_tenant_type", "tenant_id", "memory_type"),
    )

    id: Mapped[uuid.UUID] = mapped_column(
        PGUUID(as_uuid=True), primary_key=True, default=uuid.uuid4
    )
    agent_id: Mapped[uuid.UUID] = mapped_column(
        PGUUID(as_uuid=True), nullable=False, index=True
    )
    memory_type: Mapped[str] = mapped_column(
        String(50), nullable=False, default="fact"
    )
    content: Mapped[str] = mapped_column(Text, nullable=False)
    metadata_: Mapped[dict[str, Any] | None] = mapped_column(
        "metadata", JSONB, nullable=True
    )
    # embedding column is managed via raw SQL (pgvector extension)
    # embedding vector(768) — see migration 0001_initial.sql