JSPM

idkitx is a flexible, URL-safe ID generator designed for performance, compactness, and clarity.

It gives you short, secure IDs that can be random, deterministic, or sequential in either Base62 or Crockford's Base32 character sets. Inspired by the best parts of nanoid, ulid, ksuid, and hashids, it provides a simple API for generating IDs that are both human-readable and database-friendly.

import { generateId } from 'idkitx';

generateId();                               // e.g., 'A9bzK2XcPv' (random)
generateId({ length: 8 });                  // e.g., '4MfbTz1q' (shorter random ID)
generateId({ input: 'user:42' });           // deterministic — same input = same ID
generateId({ sequential: true });           // e.g., '0v4mG6CZKuSk31H8', sortable by timestamp
generateId({ sequential: true, length: 20 }); // longer sequential ID (must be ≥ 16)
generateId({ alphabet: 'crockford' }); // e.g., '2Z5G1K9DMQ', Base32 Crockford encoding, fewer similar chars, but less entropy per char

✅ Customizable entropy via length
✅ Compact, URL-friendly, human-readable (Base62 or Crockford's Base32)
✅ Deterministic mode for de-duping/idempotency (hash of input)
✅ Sequential mode (timestamp + randomness) for insert performance
✅ Random mode (short but globally unique)
✅ Fully typed, fast, 100% test coverage

🚀 Why idkitx?

idkitx is for developers who want:

• Shorter IDs than UUIDs (as few as 8–12 characters)
• Better insert/index performance than ULID or UUID in some databases
• More flexibility than libraries that only do random or sequential
• Postgres-friendly primary keys

It uses Node’s crypto.randomBytes() under the hood, just like uuid, nanoid, and ulid. While this provides strong randomness for uniqueness and distribution, idkitx is not a cryptography library and should not be used to generate secrets or session tokens.

🤔 When not to use idkitx

While idkitx is great for compact, flexible ID generation, it’s not ideal for every situation:

• ❌ You need universally recognized ID formats (e.g. for APIs, OAuth, logs) — use uuid
• ❌ You need monotonicity across distributed systems — use ulid with a monotonic factory
• ❌ You require binary-compatible formats (e.g. 128-bit UUIDs stored in binary columns) — use uuid or native database GUID types
• ❌ You need audited cryptographic guarantees — use crypto.randomUUID() or a dedicated crypto library

🔍 How it compares

⚠️ ULID is sortable within a process, but only guaranteed monotonic with a factory. UUIDv4 is cryptographically secure, but 36 characters long and not sortable.

Installation

npm i idkitx

API Reference

generateId(options?: GenerateIdOptions): string

Generates a Base62-encoded or Crockford Base32-encoded ID string.

Options:

📏 Entropy and Choosing a Length

Each character in a Base62-encoded ID contributes approximately 5.95 bits of entropy. Use this to estimate how long your IDs should be based on how many you’ll generate.

If you're using crockford encoding, the entropy per character is slightly lower (5 bits), so adjust your length accordingly.

💡 Recommendation

• Use 10–12 characters for most app-level IDs
• Use 16+ characters for large-scale, high-throughput systems
• Deterministic IDs should be long enough to minimize hash collisions
• Sequential IDs must be ≥16 to make room for the timestamp prefix

idkitx uses Node’s crypto.randomBytes() for high-entropy randomness. But even good randomness can collide — make sure your length matches your scale.

Contributing

• ⭐ Star this repo if you like it!
• 🐛 Open an issue for bugs or suggestions.
• 🤝 Submit a PR to main — all tests must pass.

Related Projects

• autorel: Automate semantic releases based on conventional commits.
• hoare: A fast, defensive test runner for JS/TS.
• jsout: A minimal logger for JS/TS, syslog-style.
• brek: Typed config loader for dynamic, secret-based configs.
• pgsmith: A SQL builder for parameterized queries in PostgreSQL.

Made with 💜 by @mhweiner