General Engine

The General Engine is the default path for production agents that need governed durable user context without an industry-specific schema. Typical uses include long-lived assistants, coding agents, productivity copilots, SaaS personalization, and internal agents where preferences, goals, procedures, relationships, or user facts can change over time. It is not intended to replace complete transcripts, static document retrieval, or an application database. Short-lived agents whose entire required context already fits in one prompt may not need durable memory yet.

Memory categories

Category	Example
`preference`	User prefers concise answers with Python examples.
`fact`	User works in a B2B SaaS support team.
`goal`	User is preparing to launch an AI product.
`procedure`	User wants weekly summaries every Friday.
`relationship`	User reports to the Head of Product.
`expertise`	User is comfortable with FastAPI and PostgreSQL.

Write

add() authenticates the tenant, resolves external_user_id, runs the quality gate, queues extraction, resolves conflicts, stores memories, and indexes vectors — then returns a job result. Extraction is async.

result = mem.add(
    messages=[
        {
            "role": "user",
            "content": "Remember that I prefer concise technical explanations with Python examples.",
        }
    ],
    external_user_id="user_123",
)

Possible outcomes:

Status	What to do
`queued`	Continue normally
`passthrough`	Skip memory context, still call the LLM
`L1`	Per-user rate limit — retry later
`L2`	Low-quality input — don’t retry the same message
`L3`	Near-duplicate — avoid repeated writes
`L4`	Budget block — wait for reset or upgrade

Blocking add() should not affect LLM calls.

Retrieve

Call get() before the model call. MemoryOS ranks memories by semantic relevance, importance, recency, and hot-tier signals, then returns system_prompt_addition ready to prepend.

memories = mem.get(
    query="answer this user's integration question",
    external_user_id="user_123",
    limit=5,
    context_max_tokens=500,
)

system_prompt = "You are a helpful assistant."
if memories.has_context:
    system_prompt = f"{system_prompt}\n\n{memories.system_prompt_addition}"

Use a query that describes the current task. "answer this user's question" retrieves better than "preferences" or "memory".

When to switch to a domain schema

The General Engine is the right default. Move to a domain schema when your product needs structured domain memory:

EdTech — student profiles, weak topics, exams, forgetting stages
Support — customer issue history, support type, sentiment, escalation risk

Getting Started

Core Memory

Domain Memory

Reliability

Legal and Trust

Memory categories

Write

Retrieve

When to switch to a domain schema

​Memory categories

​Write

​Retrieve

​When to switch to a domain schema

​Related pages

Memory categories

Write

Retrieve

When to switch to a domain schema

Related pages