Crate meme 

meme

Long-term memory for AI agents.

A Rust implementation of a production-grade memory pipeline:

1. Semantic Structured Compression — dialogues → compact memory entries
2. Lifecycle Reconciliation — LLM-driven ADD/UPDATE/DELETE/NOOP
3. Intent-Aware Retrieval Planning — multi-view hybrid retrieval

Memory is persistent across sessions — the vector store is stored on disk.

Quick Start

use meme::{Meme, MemeBuilder};

let meme = MemeBuilder::new()
    .api_key("sk-...")
    .model("gpt-4.1-mini")
    .build()
    .await?;

// Dialogue-based ingestion
meme.add_dialogue("Alice", "Let's meet at 2pm tomorrow", None).await?;
meme.finalize().await?;

// Direct fact ingestion (skips dialogue windowing)
meme.add("Alice prefers coffee over tea").await?;

// CRUD
let results = meme.search("Alice meeting").await?;
let answer = meme.ask("When will Alice meet?").await?;

Modules

config

Configuration system with TOML file + environment variable support.

embedding

Embedding model abstraction — unified interface for API and local ONNX providers.

error

Unified error types for the meme library.

http

Shared HTTP client with production-ready defaults.

llm

LLM client abstraction — OpenAI-compatible async interface.

model

Data models for the meme memory system.

pipeline

Three-stage memory pipeline: compression, synthesis, and retrieval.

store

Storage layer — LanceDB vector store with multi-view indexing and history tracking.

Structs

Meme

The main entry point for the meme memory system.

MemeBuilder

Builder for constructing a Meme instance.