memeloop
v0.3.4
Published
MemeLoop core runtime: agent framework, storage, sync engine, tools, IM adapter.
Downloads
2,151
Readme
memeloop
Core runtime for MemeLoop: a local-first, multi-agent framework that runs across Desktop, CLI, Mobile, and Cloud.
What is this package?
memeloop provides the shared runtime used by every MemeLoop host:
- Agent loop framework (
loopAPI/loops) – ReAct tool-loop and sub-agent orchestration primitives. - Loop registry & profiles – register built-in loops, tool plugins, prompts, and LLM providers, then create a runner for a profile.
- Storage & sync engine – SQLite-backed persistence with TiddlyWiki integration and conflict-aware sync.
- LLM providers – unified provider entry point over the
aiSDK (OpenAI, Anthropic, Google, Groq, Ollama, etc.). - libp2p networking – peer discovery, relay, and encrypted channels between nodes.
- IM adapter – bridge agents to chat-style interaction.
Install
pnpm add memeloop
# or
npm install memeloopUsage
import { createAgentLoopRunner, registerBuiltinLoops, registerBuiltinToolPlugins } from "memeloop";
registerBuiltinLoops();
registerBuiltinToolPlugins();
const runner = createAgentLoopRunner({ profileId: "general-assistant" });For LLM provider setup:
import { createLanguageModel } from "memeloop/llm-providers";
const model = createLanguageModel({
provider: "openai",
modelId: "gpt-4o",
apiKey: process.env.OPENAI_API_KEY,
});Exports
memeloop– core runtimememeloop/loop-api– loop primitives and script context typesmemeloop/llm-providers– unified LLM provider factory
Development
pnpm install
pnpm --filter memeloop build
pnpm --filter memeloop testLicense
MIT
