Docs · Alice
The alice researcher.
Alice acquires, cleans, and digests web content — text, images, video, and 3D models — into datasets, with safety designed in from the first fetch. She runs as an extension on the godspeed engine, sells standalone, and is included free in every conjure subscription. Product page: /products/alice.
The pipeline
- Acquire — fetch pages and media across the modalities: text, images, video, models. Sources are declared up front; alice works from what you allow, not from whatever she stumbles into.
- Clean — strip boilerplate, deduplicate, normalize, and scrub fetched material into usable form.
- Digest — structure the cleaned material: extract, summarize, label, and organize it into corpora ready for search or training.
- Export — datasets leave alice in formats your tools consume, with provenance kept.
The safety model
Alice is security-first by construction, not by configuration:
- Allowlist-first — fetching starts from allowed sources; the unsafe default is no fetch at all.
- Hard content filters — abusive and illegal content classes are blocked in the acquisition path itself, before storage, always on.
- A paper trail — what was fetched, from where, when, and what happened to it: recorded. Datasets you can answer questions about later.
In the BigBrain line
Alice is the acquisition layer of the AI products: conjure queries her live at generation time to study references for what it is asked to build (which is why conjure bundles her), and our own training corpora are prepared with the same pipeline customers get.
Licensing & activation
- Alice activates with her own AL-XXXX-XXXX-XXXX-XXXX key — whether bought standalone or received free inside a conjure subscription. Any supported OS, transferable between machines.
Status & roadmap
Coming soon; founders subscriptions are open. The acquisition, safety, cleaning, and digestion stack is built and passing its quality gates; packaged builds publish once the workstation surfaces settle. Subscribers’ keys activate the first published build.