Quickstart

AI memory system — episodic, semantic, and working memory with pluggable backends and consolidation.

Maturity

Alpha (0.1.x) — public APIs may change before 1.0.

Install

uv add lexigram-ai-memory

Minimal Usage

import asyncio
from datetime import UTC, datetime

from lexigram import Application
from lexigram.ai.memory import MemoryModule, MemoryConfig
from lexigram.contracts.ai.memory import MemoryEntry, MemoryQuery


async def main():
    config = MemoryConfig()
    async with Application.boot(
        name="memory-demo",
        modules=[MemoryModule.configure(config)],
    ) as app:
        store = await app.container.resolve(MemoryStoreProtocol)

        entry = MemoryEntry(
            id="1",
            content="The capital of France is Paris.",
            role="assistant",
            timestamp=datetime.now(UTC),
            importance=0.9,
        )
        await store.store(entry)

        results = await store.retrieve(
            MemoryQuery(query="france capital", top_k=5)
        )
        for r in results:
            print(f"[{r.score:.2f}] {r.entry.content}")


asyncio.run(main())

What Just Happened

1. MemoryModule.configure(config) wired MemoryProvider into the container.
2. MemoryStoreProtocol resolved the default InMemoryMemoryBackend.
3. store.store() saved an entry with importance metadata.
4. store.retrieve() searched entries by query and returned scored results.

Next Steps

• Guide — three memory tiers, consolidation, end-to-end usage
• How-Tos — common memory patterns
• Configuration — all config keys