@lore-context/sdk-middleware

SDK middleware for Lore Context — wraps OpenAI / Anthropic / OpenRouter SDK clients to auto-inject memory pre-LLM-call and auto-retain post-LLM-call without requiring the AI to call an MCP tool.

Install

pnpm add @lore-context/sdk-middleware @lore-context/sdk openai

Quick start

import OpenAI from "openai";
import { LoreContextApiClient } from "@lore-context/sdk";
import { withMemoryHooks } from "@lore-context/sdk-middleware";

const lore = new LoreContextApiClient({ token: process.env.LORE_API_KEY! });
const openai = new OpenAI({ apiKey: process.env.OPENAI_API_KEY });

const client = withMemoryHooks(openai, lore, { mode: "hybrid" });

const response = await client.chat.completions.create({
  model: "gpt-4o",
  messages: [{ role: "user", content: "What was that Postgres dump command?" }],
});

Modes

| Mode | inject | retain | tools exposed | |---|---|---|---| | hybrid (default) | ✅ | ✅ | ✅ | | context | ✅ | ✅ | ❌ | | tools | ❌ | ❌ | ✅ |

See docs/sdk/middleware-modes.md in the lore-cloud repo for full mode semantics.

License

MIT — see LICENSE.