Bring your own model
Import a local GGUF or download a pinned public Hugging Face GGUF. Compatibility depends on the model and your hardware.
Features / VEYLUM
A little light. A lot of possibility.
Local, capable, considered
Built around your day.
Optional OpenAI API
Use your own protected API key for personal text chat. Review each cloud answer before context is sent. API usage is billed separately.
ARC models
Install a pinned, checksum-verified model and run it on your computer. Hardware recommendations help you choose a practical tier.
Conversations and branches
Stream answers, stop generation, regenerate, and branch a conversation. Try another direction without losing the original.
Local memory
Keep useful context locally. Review, edit, or remove memories and choose the scope that can use them.
Bots and projects
Give a bot a role and keep project context together. Attach a folder explicitly to enable bounded workspace tools.
Coder and project agents
Opt into a Development bot for reviewed file changes, a fixed offline Node test profile, and structured local Git add/commit. This preview has no unrestricted shell or Git push.
Tasks and approvals
Local background tasks preserve their state and support pause, resume, and cancellation. Waiting approvals remain visible; restart does not grant new authority.
Backups
Create a local .veylum archive and optionally protect it with a passphrase. Model weights are separate. The live database and rollback copies remain unencrypted.
Veyling
A prism companion with factual activity states, text equivalents, and reduced motion. No hidden reasoning is displayed.
Desktop Halo
An optional, compact floating Veyling shows factual activity and approval notifications. Review opens the desktop approval flow; hiding it stays in your control.
A deliberate preview
Know what to expect.
Mobile companion
Paired personal text chat is available on the same LAN. Broader project, bot, task, and approval surfaces are in final validation and are not yet advertised as verified here.
Current limits
Windows is the verified target. Mac runtime verification is pending. Image input is unavailable. Optional OpenAI API personal chat is available with your own paid API key and per-answer review. There is no automatic updater or broader remote mobile networking.