FAQ
Frequently asked questions
Everything you need to know about NexDok. Can’t find an answer? Contact our team.
Platform basics
What is NexDok?
NexDok is a next-generation document management and intelligence platform designed for organizations that need to securely store, organize, version, and collaborate on documents. It combines traditional DMS capabilities with AI-powered search and question-answering, approval workflows, and multi-tenant workspace isolation — all accessible via a modern web interface and a developer-friendly REST API.
Who is NexDok designed for?
NexDok is built for teams and organizations of any size that manage structured documents — from contracts, policies, and procedures to technical manuals and compliance records. It is especially valuable for legal teams, compliance departments, quality management, HR, and any business unit that needs version control, approval workflows, and intelligent document retrieval.
What types of documents can I upload to NexDok?
NexDok supports a wide range of file formats including PDF, DOCX, XLSX, PPTX, DOC, XLS, PPT, CSV, TXT, Markdown, HTML, JSON, and XML files. Each uploaded file is stored securely and can be versioned, indexed for AI search, and organized into collections.
What is a Workspace in NexDok?
A Workspace is an isolated environment where your team stores and manages documents. Each workspace has its own members, document types, collections, security zones, and settings. Data is strictly isolated between workspaces — users in one workspace cannot access documents from another. A single user can belong to multiple workspaces and switch between them seamlessly.
How does document versioning work?
Every document in NexDok maintains a complete version history. When you upload a new file to an existing document, a new version is created automatically with an incremented version number. Each version is immutable — once created, it cannot be modified. You can view, download, and compare any historical version. The system tracks a head version (latest) and a published version (the one visible to consumers), which may be different.
What is the difference between head version and published version?
The head version is always the most recently uploaded version of a document. The published version is the one that has been explicitly approved and published by an administrator. You can upload multiple draft versions while the current published version remains unchanged. This allows teams to iterate on document content before making it officially available.
How do Document Types work?
Document Types are templates that define the structure and behavior of documents within a workspace. Each type specifies a custom metadata schema (with required and optional fields), allowed workflow states, transition rules between states, retention policies, and AI indexing configuration. For example, you might create a Contract type with fields like counterparty, effective date, and value, and a Policy type with department and review frequency fields.
What are Collections and how do I use them?
Collections are flexible groupings for organizing documents. A document can belong to multiple collections simultaneously. Collections support three visibility levels: private (only the owner sees it), restricted (specific users or groups have access), and tenant-public (visible to all workspace members). Collections can also be linked to Security Zones for additional permission control.
Can I organize documents with custom metadata?
Yes. Each Document Type defines a metadata schema using JSON Schema. When creating or editing a document, users fill in the metadata fields defined by its type. You can then filter and search documents by any metadata field. For example, filter all contracts where counterparty is Acme Corp or all policies expiring this quarter.
What is the document lifecycle in NexDok?
Documents follow a clear lifecycle: Draft, Published, and Archived. New documents start as drafts. An admin or owner can publish a specific version, making it the official record. Documents can be archived when no longer active (read-only, irreversible). Draft documents can be soft-deleted. At each stage, NexDok records who made the change, when, and why — providing a complete audit trail.
Does NexDok support bulk document uploads?
Yes. NexDok provides a bulk upload feature that allows you to upload multiple documents at once. You can select a document type and assign metadata during the bulk upload process, streamlining the onboarding of large document sets.
Can I download documents from NexDok?
Yes. NexDok generates secure, time-limited signed URLs for document downloads. These URLs expire after 10 minutes and contain no credentials. Any workspace member can download the current head version; downloading historical versions requires admin or owner role.
Is there an API available for integration?
Yes. NexDok exposes a comprehensive REST API that covers all platform functionality: document CRUD, version management, file uploads, search, AI queries, approval workflows, member management, and more. The API supports both JWT authentication (for user-facing apps) and API Key authentication (for server-to-server integrations), with configurable scopes and rate limits.
What happens when I delete a document?
Deletion in NexDok is a soft delete — the document is marked as deleted but not physically removed from storage. Only documents in draft state can be deleted. Published or archived documents must first be archived. This design prevents accidental data loss and supports compliance requirements for document retention.
Does NexDok have a dashboard?
Yes. The Dashboard provides an at-a-glance overview of your workspace including document counts by status, recent activity, pending approvals assigned to you, storage usage, AI indexing progress, and member activity — all updated in real-time.
Search & AI
How does AI-powered search work in NexDok?
NexDok uses semantic search powered by OpenAI embeddings (text-embedding-3-small, 1536 dimensions). When a document is indexed, its content is extracted, split into chunks, and each chunk is converted into a vector embedding. When you search, your query is also converted into an embedding and matched against all document chunks using cosine similarity via pgvector. This means you can find documents based on meaning, not just exact keyword matches.
What is the Ask feature?
The Ask feature is a Retrieval-Augmented Generation (RAG) system. You type a natural language question, NexDok retrieves the most relevant document chunks via semantic search, and then sends them to a large language model (GPT-4o-mini) along with your question. The AI generates a grounded answer citing the specific documents and passages it used. This is especially powerful for quickly finding information across large document libraries.
Which file types support AI indexing?
AI indexing is supported for PDF, DOCX, CSV, TXT, Markdown (.md), HTML, JSON, XML, and RDF/XML. These formats can be automatically extracted, chunked, and embedded. Other file types (like images, spreadsheets without text, or presentations) are stored and versioned normally but are not indexed for AI search.
How is document content extracted for AI?
NexDok uses different extraction strategies depending on the file type: PDF files are processed with specialized text extraction (per page), DOCX files extract paragraphs and tables, and plain text formats (TXT, MD, HTML, JSON, XML) are decoded directly. The extracted text is then split into chunks using a sliding window approach (512 words with 64-word overlap) to maintain context across chunk boundaries.
What does the AI indexing status mean?
Each document version has an indexing status: Created (just uploaded), Uploading (file transfer in progress), Uploaded (ready for indexing), Queued (waiting for the indexing worker), Indexing (being processed), Indexed (fully searchable with embeddings), Text Only (text extracted but no embeddings), or Failed (an error occurred during processing). You can retry failed indexing at any time.
Can I search within a specific collection or document type?
Yes. Both the Search and Ask features support filtering by collection, document type, workflow state, and AI indexing status. This allows you to scope your queries to specific subsets of your document library — for example, searching only within published contracts or asking questions only about HR policies.
Is AI indexing automatic?
AI indexing is triggered automatically when a document version is published. For unpublished versions, you can manually trigger indexing from the document detail page. The indexing process runs asynchronously via a background worker, so it does not block your workflow. The document detail page polls for status updates and shows real-time progress.
How accurate is the AI search?
Semantic search finds documents based on meaning rather than exact keywords, which significantly improves recall for natural language queries. The accuracy depends on the quality and structure of your documents. Well-structured documents with clear headings and consistent terminology produce the best results. The Ask feature includes source citations so you can always verify the AI answer against the original document.
Are there rate limits on AI features?
Yes. To ensure fair usage and control costs, NexDok applies rate limits to Search (60 queries/minute per user, 2,000/day), Ask (10 queries/minute per user, 300/day), and indexing (5 triggers/minute per user). Limits can be adjusted per workspace. Rate limit status is returned in API responses so applications can handle throttling gracefully.
Can I use the AI features via the API?
Yes. The Search endpoint (POST /api/v1/query) and Ask endpoint (POST /api/v1/ask) are available via both JWT and API Key authentication. API Keys require the query scope for Search and the ask scope for Ask.
Does NexDok store my AI queries?
NexDok tracks AI usage metrics (query counts per month) for usage reporting. The content of your queries and the AI responses are not stored persistently. All AI processing happens in real-time and results are returned directly to the caller.
What AI model does NexDok use?
NexDok uses OpenAI text-embedding-3-small model (1536 dimensions) for document embeddings and semantic search, and GPT-4o-mini for the Ask (RAG) feature. The models are configurable at the platform level for future upgrades.
Access & security
How does authentication work in NexDok?
NexDok uses industry-standard authentication with JWT tokens signed using ES256 (asymmetric cryptography). Tokens are validated against public JWKS endpoints, ensuring they cannot be forged. Token expiration, audience, and issuer are all verified on every request. The platform supports email/password login with secure password reset flows.
What user roles are available in a workspace?
NexDok provides four workspace-level roles in a strict hierarchy: Viewer (read-only access to documents and collections), Editor (can create and edit documents, upload versions, and request approvals), Admin (full document management, member management, publish/archive, security configuration), and Owner (same as Admin, with the added protection that the last owner cannot be removed or demoted).
How is data isolated between workspaces?
Data isolation is enforced at every level. Every database table includes a tenant_id column, and all queries filter by tenant. Row-Level Security (RLS) policies are active on all tables to prevent cross-tenant access even at the database level. File storage uses tenant-scoped paths. The API validates workspace membership on every request before returning any data.
What are Security Zones?
Security Zones are permission boundaries within a workspace. They allow you to create logical areas (e.g., Legal Department, Executive, Public) and control which users or groups can access documents and collections within each zone. Zones support user-level and group-level permission grants with viewer, editor, and manager levels.
What are Access Groups?
Access Groups are named sets of users within a workspace. Instead of granting permissions to individual users one by one, you can create a group (e.g., Finance Team, External Auditors), add members to it, and then grant the group access to collections or security zones. When a user is added to or removed from the group, their access updates automatically.
How does the approval workflow work?
NexDok supports an optional document approval workflow. An editor can submit a specific document version for approval, selecting one or more approvers from the workspace members. Each approver independently reviews and approves or rejects. A single rejection closes the request immediately. When all approvers approve, the document becomes eligible for publication. Approval is version-specific — uploading a new version does not affect a pending approval on an older version.
Can I control who can publish documents?
Yes. Only users with Admin or Owner role can publish documents. Additionally, when an approval workflow is active for a document version, publication is blocked until all designated approvers have approved. This dual gate (role + approval) ensures that documents go through proper review before becoming official.
How do API Keys work?
API Keys enable server-to-server integration without requiring user JWT tokens. Each key is scoped to a specific workspace and has configurable permissions (upload, query, ask). Keys are stored as SHA-256 hashes — the raw key is shown once at creation and never stored. Keys can have expiration dates and can be revoked at any time.
Does NexDok maintain an audit trail?
Yes. NexDok maintains multiple layers of audit information: an events table that records all document lifecycle events (created, published, archived, deleted) with actor, timestamp, and payload; a workflow history table that tracks all state transitions with before/after snapshots; and a governance audit log for account-level operations (member changes, tenant management). All audit writes are transactional — they occur atomically with the operations they record.
What happens when a workspace member is disabled?
When an admin disables a workspace member, that user immediately receives a 403 error on all API calls to that workspace. Their data is preserved, and they can be re-enabled at any time without losing their role or history. The last active owner of a workspace cannot be disabled, preventing accidental lockout.
How does the member invitation flow work?
When a new member is invited to a workspace, they receive a pending status. On their first authenticated API call (typically when they log in and the app loads), their status is automatically activated. This lazy-activation approach means no separate accept invitation step is required — the member simply logs in and starts working.
Can account-level roles access workspace documents?
No. Account-level roles (Billing Owner, Billing Admin, Account Admin) are strictly separated from workspace-level roles. A Billing Owner can manage subscriptions, create workspaces, and suspend tenants, but they cannot access any documents unless they are also explicitly added as a member of that workspace. This separation ensures billing administrators do not automatically gain access to sensitive document content.
Does NexDok support document-level permissions?
Yes. Beyond workspace roles and collection visibility, NexDok supports document-level Access Control Lists (ACLs). Permissions can be granted to specific users, roles, or publicly within a workspace. Available permission types include read, write, manage, and publish. When ACL entries exist on a document, only grantees (and workspace admins) can access it.
How are files stored securely?
All document files are stored in cloud object storage using tenant-scoped paths that prevent cross-tenant access. Files are never served directly — the backend generates time-limited signed URLs (10-minute expiry) for downloads. The service credentials that access storage are never exposed to the frontend or API consumers.
Still have questions?
If your question isn’t listed here, contact us or email support@nexdok.com.