npm package discovery and stats viewer.

Discover Tips

  • General search

    [free text search, go nuts!]

  • Package details

    pkg:[package-name]

  • User packages

    @[username]

Sponsor

Optimize Toolset

I’ve always been into building performant and accessible sites, but lately I’ve been taking it extremely seriously. So much so that I’ve been building a tool to help me optimize and monitor the sites that I build to make sure that I’m making an attempt to offer the best experience to those who visit them. If you’re into performant, accessible and SEO friendly sites, you might like it too! You can check it out at Optimize Toolset.

About

Hi, 👋, I’m Ryan Hefner  and I built this site for me, and you! The goal of this site was to provide an easy way for me to check the stats on my npm packages, both for prioritizing issues and updates, and to give me a little kick in the pants to keep up on stuff.

As I was building it, I realized that I was actually using the tool to build the tool, and figured I might as well put this out there and hopefully others will find it to be a fast and useful way to search and browse npm packages as I have.

If you’re interested in other things I’m working on, follow me on Twitter or check out the open source projects I’ve been publishing on GitHub.

I am also working on a Twitter bot for this site to tweet the most popular, newest, random packages from npm. Please follow that account now and it will start sending out packages soon–ish.

Open Software & Tools

This site wouldn’t be possible without the immense generosity and tireless efforts from the people who make contributions to the world and share their work via open source initiatives. Thank you 🙏

© 2026 – Pkg Stats / Ryan Hefner

@kedem/okcode

v1.0.2

Published

Agentless workspace tools: find, read and edit code by name in local or remote workspaces — indexed with okdb (symbols, full-text, embeddings), content never copied.

Readme

okcode

Agentless workspace tools for code: find, read and edit code by name in local or remote workspaces. Part of the ok family — okjs (UI), okdb (data), okcode (working on code).

  • Find — symbols by name, full-text over names/doc prose and over file content, semantic search (embeddings), references, outlines, dependency packages.
  • Read — a symbol, a range, a whole file; every read hands back an at token (the file's hash).
  • Edit — by symbol or range, guarded by at: a stale edit is refused, never rebased. Candidates are validated locally before the write (a change that introduces a new diagnostic is refused); the write is an atomic compare-and-swap on the target.
  • Anywhere — okcode never opens a connection. You hand it an access facade (list / stat / read / commit). okcode ships a local filesystem facade and a shell facade that speaks bash/PowerShell over any run(script, stdin) — so ssh, docker exec, kubectl exec reach a remote workspace with nothing installed on it — plus two read-only ones over the host's own functions or an in-memory map.
  • No copies — the index (symbols, full-text postings, vectors) lives in okdb; file content is never stored. okdb reads it through the facade when indexing and when showing a result (resolved fields).

Status: 1.0.0. okcode 1.0 is a native (Rust) implementation behind the same JavaScript API as 0.2; it requires @kedem/okdb ^3 and talks to okdb through okdb's stable C bridge, sharing the host's okdb instance instead of opening its own. Upgrading from 0.2: docs/UPGRADING.md. What changed and by how much: CHANGELOG.md.

Requirements

  • Node.js ≥ 20 (for the library and the okcode command), or no Node at all for the standalone native okcode binary.
  • @kedem/okdb ^3 (a normal dependency; npm dedupes it with your own). A host that passes its own db must pass an okdb v3 instance; a v2 db is refused (OKCODE_OKDB_V3_REQUIRED) — v2 hosts stay on okcode 0.2.
  • linux-x64 with glibc 2.28+ (Debian 11+, Ubuntu 20.04+, RHEL 8+), the one platform 1.0.0 bundles a native addon for. Elsewhere require('@kedem/okcode') throws OKCODE_UNSUPPORTED_PLATFORM. See docs/NATIVE.md.

Quick look

const okcode = require('@kedem/okcode');

const oc = await okcode.open({
    path: '/var/lib/okcode', // okcode's own okdb store (or pass `db`)
    embedders: { qwen: { type: 'ollama', model: 'qwen3-embedding:0.6b' } },
    active: 'qwen',
});

// A local folder
await oc.addWorkspace('app', { access: okcode.access.localFs('/work/app') });

// A huge repo you only ever ask by name: skip the (bigger) content index
await oc.addWorkspace('monorepo', { access: okcode.access.localFs('/work/monorepo'), fts: 'symbols' });

// A remote folder — okcode supplies the scripts, you supply the transport
await oc.addWorkspace('prod-api', {
    access: okcode.access.shell({
        dialect: 'bash',
        root: '/srv/api',
        run: (script, stdin) => mySsh.run(script, stdin),
    }),
});

const ws = oc.workspace('app');
await ws.sync();
const hits = await ws.find('retryWithBackoff');
const { body, at } = await ws.read('retryWithBackoff');
await ws.edit('retryWithBackoff', newBody, { at });

// Semantic search; or with a vector the host already embedded (no embed call),
// once its identity — [type, endpoint, model, dims] — matches the profile's
const { identity } = oc.embedders().find((e) => e.active);
await ws.ask('where do we retry failed requests');
await ws.ask({ text: 'where do we retry failed requests', vector, identity });

// Listings from the index alone — no file content is read
await ws.files({ dir: 'src' }); // sizes, lines, languages, symbol counts
await ws.symbols('src/http.js'); // the stored symbol table of one file

await oc.close();

mySsh, newBody and vector are the host's: a transport, the replacement text, and the question embedded in the profile's space (a Float32Array or a number array).

CLI — run it in a folder; the index lives in .okcode/ there (workspace default), created and kept current automatically:

okcode find retryWithBackoff
okcode read retryWithBackoff            # prints the code; at=<hash> on stderr
okcode edit retryWithBackoff --file body.txt --at <hash>
okcode status
okcode find retry --store /data/okcode --id api   # a shared store holding many workspaces

Licensing. Without a license, okdb's free tier covers about three workspaces (5 envs, 2 embeddings pipelines per env). A license goes into okcode's store with okcode.open({ path, license }) (idempotent — pass it on every open), okcode --license FILE, or OKDB_LICENSE_FILE=FILE for any process; status().license shows what is in effect. A host that passes its own db licenses it itself. See DESIGN §5.

API

require('@kedem/okcode') is { open, access, analysis }. Behaviour and result shapes: DESIGN §9.

okcode.open(options) → oc

| option | meaning | | ----------- | ----------------------------------------------------------------------------------------------------------- | | path | okcode's own okdb store, opened (and closed on close()) by okcode. Pass path or db, not both | | db | an open okdb v3 instance the host owns; okcode never closes it | | role | with path: okdb's process roles for that store, { processors, engines, compaction } | | license | with path: an okdb license (text, or { blob, activation }), installed first (DESIGN §5) | | embedders | named embedder profiles (DESIGN §8); active names the one ask() uses | | workspace | defaults for workspaces that don't say, e.g. { fts: 'symbols' } | | log | a logger ({ info, warn, error, debug }) or one function; okcode's warnings go to warn |

oc — addWorkspace(id, { access, index, fts, options }) (resolves to the workspace), workspace(id), workspaces(), removeWorkspace(id), sync(id?, { force }), status(id?), reset(id, { scope }), embedders(), active, addEmbedder(name, cfg), useEmbedder(name), removeEmbedder(name), removeOrphaned(id?), compare(id, query, names?), close(); db (the okdb instance) and meta (okcode's own okcode env). addWorkspace's options tune one workspace: skip (directory names to prune; default okcode.access.DEFAULT_SKIP), textMax, libraryFiles, libraryParseFiles, libraryScans.

A workspace (ws) — properties id, access, index ('full' | 'none'), readOnly, store; methods:

| group | methods | | ------- | ----------------------------------------------------------------------------------------------------------------------------- | | look up | find(q, { kind, path, limit, offset }), ask(q, { profile, path, limit }), refs(q, { limit, offset }), importers(file), mentions(text), grep(pattern, opts), glob(patterns), outline(file), structure({ dir }), files({ dir, lang, stat }), symbols(file, { kind, stat }), packages({ refresh }), package(name), isLibraryPath(rel) | | read | read(symbol \| file#name \| file \| file:a-b), readFile(file, { maxLines }), readAt(file, a, b, at?) — results carry body and at | | edit | edit(target, body, { at }), editBatch({ at, edits: [{ target, body }] }), writeAt(file, a, b, body, { at }), writeWholeFile(file, text, { create \| expectedHash }), syntaxCheck(file) (node --check on the target, through the facade's exec) | | manage | sync({ force }), refresh(), flush(), stats(), scanWarnings(), close() |

ws.store (null when index: 'none') is the workspace's okdb side: envName, env (the okdb env), ask(q, { profile, limit }) (one profile's raw vector hits), search(q, limit, mode) (the symbol full-text index), profiles(), and settle({ profile }), which resolves once that profile's embeddings have caught up.

okcode.access — localFs(root), shell({ root, run, dialect }) (with shell.quote(s) and shell.localRunner(dialect), a run for a local shell), fn({ list, read, stat? }), memory(map), and the DEFAULT_SKIP / PRUNE_DOT_DIRS lists. Contract: docs/ACCESS.md.

okcode.analysis — the pure functions on text the workspace is built on: language detection (langOf, extOf, isTextual, isSecret), symbol extraction (extract, parseSource), validation (validateSource, compareDiagnostics), document sections (regions, sectionOf), text/EOL/diff utilities (decode, normaliseEol, hunksFor, …), template balance, convention colours, okjs, and dependency surfaces (scanPackages, readPackages). See DESIGN §2. A host adds its own analysers with registerExtension({ id, match, analyze, version }) and registerExtractor(lang, fn): extract() and the workspaces' indexing use them for the textual paths they claim (a new version re-analyses those files on the next scan); every other path keeps the built-in analysis.

Model-facing tools — require('@kedem/okcode/lib/tools').createTools({ workspace: async (hint) => ws, workspaces }) → { schemas(), render(name, args), TOOLS } for code_map, code_find, code_grep, code_glob, code_outline, code_read, code_package, code_ask, code_edit, code_write. render resolves to text and never throws.

CLI

okcode <command> [args] [--root DIR] [--id NAME] [--store DIR] [--json]; okcode help <command> describes one command.

  • look up: structure, find, refs, importers, grep, glob, outline, read, read-at, packages, ask
  • edit: edit, edit-batch, write, check
  • manage: status, sync [--force] [--no-wait], reset --scope vectors|fts|all, embedders, workspaces, remove-workspace, stats
  • flags: --root, --store, --id, --fts full|symbols, --embedder-config FILE, --license FILE, --json
  • environment: OKCODE_VERBOSE=1 routes okdb's log to stderr; OKDB_LICENSE_FILE installs a license on every open.

The same commands run without Node from the standalone binary in the package (docs/NATIVE.md).

Docs

  • docs/DESIGN.md — what okcode is, its layers, storage, embedders, management surface, and the decisions behind them.
  • docs/ACCESS.md — the access facade contract (implement it to reach a new kind of workspace).
  • docs/NATIVE.md — the native addon, the standalone binary, and how okcode reaches okdb.
  • docs/UPGRADING.md — 0.2 → 1.0 for hosts.
  • CHANGELOG.md — what changed in each release.

License

MIT — see LICENSE. okcode runs on okdb, which has its own license: its free tier covers okcode for about three workspaces (embeddings included); larger hosts install an okdb license (OKDB_LICENSE_FILE, or okcode.open({ license })).