Domain Packs
Liya Engine organises AI intelligence into domain packs — vertical contexts each with their own intents, knowledge source types, and configuration.
A domain pack defines a vertical context for AI execution. Each API call targets a specific domain and intent within it. The domain determines which intents are available, what knowledge sources are relevant, and how the engine scopes retrieval and persona behaviour.
Built-in domains
| Domain | Key | Status | Intents |
|---|---|---|---|
| Hiring | hiring | Generally available | 12 |
| Fintech | fintech | Early access | Coming soon |
| Healthcare | healthcare | Early access | Coming soon |
| EHS | ehs | Early access | Coming soon |
Custom domains
On Growth and Enterprise plans, you can define your own domain packs with custom intents, prompt templates, and knowledge source categories. See Custom Domain Packs.
API structure
Domain packs map directly to API paths:
For example:
Domain permissions
Your tenant's enabled_domains list controls which domains you can call. Within each domain, domain_permissions controls which specific intents are enabled. Both are configurable from the dashboard packs page or via the Packs API.
Knowledge sources
Each domain has a set of source types — categories for documents ingested into the knowledge base. Source types are namespaced as domain:type (e.g. hiring:resume, fintech:compliance_doc). When RAG is enabled, retrieval is automatically scoped to the domain making the request.
| Domain | Source types |
|---|---|
hiring | resume, job_description, policy, profile, session_summary |
fintech | transaction_record, compliance_doc, regulatory_filing, risk_model, product_policy |
healthcare | clinical_guideline, patient_record, drug_reference, protocol, discharge_summary |
ehs | safety_data_sheet, incident_report, audit_template, regulation, hazard_assessment |