Skip to main content
Researcher Center
SupportGet Early Access
Menu

Lab admin demo: the pipeline behind the answers

Before a researcher gets a cited answer, a lab admin turns raw files into a governed, citeable knowledge base. These walkthroughs show that pipeline on the real product UI — from upload and security scanning through ingestion, embedding, and structured sources.

Get Early AccessSee the researcher demo

The data pipeline

The data pipelineAdmin walkthrough
  1. Upload research files — PDFs, Office docs, notes, code and simulation decks.
  2. Each file is malware-scanned before anything else touches it.
  3. Ingestion extracts text; embedding makes it searchable — each stage shows its status.

Every upload clears security scanning, then ingestion, then embedding — visible per file.

Text description of this walkthrough
  1. A lab admin opens the Documents view for their workspace.
  2. The list shows the workspace's uploaded files, each with its own processing state.
  3. One file, autoclave-protocol.md, has finished: its document status reads ready and its embedding status reads ready, so it is searchable and citeable.
  4. A second file, reagent-safety.pdf, is still moving through the pipeline: its document status reads processing and its embedding status is still pending.
  5. No stage is hidden — an admin can see exactly where any file is before anyone asks a question of it.

Data structuring

Data structuringAdmin walkthrough
  1. A document carries its type, source, and confidentiality — captured, never invented.
  2. Its page and section structure is what later answers cite back to.
  3. Nothing is summarized or altered; the original passages remain the source of truth.

A raw file becomes a structured, citeable document with type, source, and page/section context.

Text description of this walkthrough
  1. A lab admin opens the detail view for a single processed document, autoclave-protocol.md.
  2. The page is headed by the original filename — the file is never renamed or rewritten.
  3. Alongside it sits the document's captured metadata: its type, its source, and its confidentiality marking.
  4. Below that is the structure extracted from the file — the page and section context that later answers cite back to.

Governance and provider choice

Governance and provider choiceAdmin walkthrough
  1. Every workspace shows its own AI usage — the basis for hard spend caps.
  2. The AI provider is the admin's call: platform-managed, your own API key, or on-prem/local.
  3. Where the research data and the model run stays a governance decision, not a default.

Per-workspace AI usage plus the provider setup — managed, your own key, or fully local.

Text description of this walkthrough
  1. A lab admin opens the organization settings page for their workspace.
  2. The Workspace AI usage panel is shown first: the number of AI events, the total tokens used — 1,500 in this recording — and the estimated cost, USD 2.50.
  3. Those figures are estimates for visibility, and they are the basis for the workspace's hard AI spend cap.
  4. The view then moves to the AI provider setup panel, where the admin chooses how AI runs: platform-managed, the lab's own provider API key, or a fully local, self-hosted model.

Governance the whole time

Get Early AccessReview Security