writerslogic-warm
v0.7.0
Published
WritersLogic Adaptive Retention & Memory (WARM): local document search (BM25 + embedding cosine) and vector-symbolic associative memory.
Maintainers
Readme
writerslogic-warm: embedded WARM for Node.js
Local document search and agent memory backed by the WARM Rust engine. The package selects a native binding for your operating system and architecture. It does not start a server or make service calls for ordinary document search.
The primary JavaScript client is the separate portable warm-sdk HTTP package. Choose this
writerslogic-warm package when the engine should run inside Node. For a published engine
release, install it explicitly:
npm install [email protected]Check the available npm versions. To build from the repository root with Rust installed:
npm ci --omit=optional
npm run buildThe native binding is distributed through an optional platform package, so keep npm's optional dependencies enabled. The release targets are macOS (arm64, x64), Linux (arm64, x64; glibc and musl), and Windows (arm64, x64).
Package names
Existing embedded consumers keep writerslogic-warm, /semantic, /memory and /vectra
imports, store directories, and embedding/security configuration. The portable warm-sdk
exports WarmClient and uses a separately running HTTP server; adopting it requires an
explicit API migration. The separately versioned MCP launcher remains writerslogic-warm-mcp.
Search documents
const { WarmMemorySystem } = require('writerslogic-warm')
const { DocumentMemory } = require('writerslogic-warm/semantic')
async function main() {
const memory = new DocumentMemory(new WarmMemorySystem(4096, './memory-store'))
await memory.memorize({
id: 'backups',
text: 'Restore deleted documents from verified backups.',
sourceUri: 'operations.md',
})
const [hit] = await memory.search('restore backups')
console.log(hit.documentId, hit.sourceUri, hit.text)
await memory.flush()
}
main().catch(error => { console.error(error); process.exitCode = 1 })Search uses local BM25 by default. You can supply embeddings for hybrid search; the optional
@huggingface/transformers integration loads a model from a local directory. CommonJS, ESM, and
TypeScript declarations are included. See the
Node integration guide
and API documentation.
MCP clients
writerslogic-warm-mcp is a separate
launcher for the MCP server. Install or run that package if you need a server for an MCP client:
npx -y writerslogic-warm-mcpThis writerslogic-warm package embeds the engine in a Node.js application. It does not expose
the MCP command.
