How can we help?

Go to Documents and drag files into the upload zone, or click it to browse. We support PDF (scans are read by OCR), DOCX, XLSX, CSV, HTML and TXT files up to 100 MB each. A file of another type is refused with a message saying why.

A domain pack sets the terminology and document types for your industry. Extraction templates and rule packs are not preloaded: we configure them with you during onboarding, for your own requirements.

Go to Admin > Users and choose Invite Member. Enter their full email address and a role (Admin, Member or Viewer). They receive an email with a secure link that is valid for 7 days.

A set of requirements for one supplier or subject. DataStruct locates the evidence for each requirement in your documents and records what it found and what it could not find. A named reviewer confirms, rejects or marks each finding for follow-up. DataStruct records no verdict of its own.

Open Evidence Programmes, enter the supplier or subject and create the programme. Add your own requirements if you need them, then choose Run programme. Runs happen in the background, so you can leave the page; earlier runs stay in the run history.

Yes. Every run exports as DOCX, XLSX, CSV or JSON from the programme page, with each finding, its cited sources and the reviewer's decision.

Search & Ask uses hybrid retrieval (lexical + semantic) combined with AI to answer questions. Switch between "Search" mode for ranked passages and "Ask" mode for AI-generated answers with evidence citations.

DataStruct links answer claims to cited passages where available, checks citations against the retrieved sources, and flags claims a verification pass does not support. If evidence is insufficient, the response says so and the answer is routed to the Review Queue.

Extraction datasets as CSV, XLSX or JSON (started from Datasets, downloaded from Exports), document reports as HTML, DOCX or JSON (from Exports), and Evidence Programme runs as DOCX, XLSX, CSV or JSON (from the programme page). PDF export is not supported.

An extraction template lists the fields to pull from documents (for example "Certificate number" or "Expiry date"), each with a type, an optional hint and a confidence threshold. Run a template against documents from the Datasets page; each run links to a results page showing every value with the passage it came from. Values below the threshold go to the Review Queue.

Rule packs are sets of deterministic checks (keyword presence and absence, regex match, numeric thresholds, missing-disclosure checks) that run against your documents. Each rule reports its result with the passage that triggered it. None are preloaded: they are configured with you for your own requirements.

There is no rule editor in the app yet. Email info@datastructai.com with the checks you need and we will set up the rule pack for your workspace.

Go to Settings > Security to change your password. You'll need to enter your current password and a new password (minimum 8 characters).

Not yet. Until it is, use a strong password you use nowhere else. If a device is lost, choose Sign out everywhere in Settings > Security: it also revokes your API keys.

Workspaces are isolated in the application layer: each tenant-scoped route checks your workspace membership and permission before returning anything. Role-based access control (RBAC) governs access through granular permission types, and sensitive actions are recorded in the audit log. See /security for what we do and do not claim.

Contact support@datastructai.com to close your pilot workspace. We export your data on request and run a hard purge of the workspace when you ask us to, then confirm completion. Deletion is not automatic.