Skip to main content

​
Curated template library

Olira maintains a library of curated view templates aimed at real clinical and product workflows: symptom-oriented snapshots, adherence-focused views, protocol- or disease-pack-oriented layouts, and similar packaged layouts. Each catalog entry defines an ordered set of blocks, default cadence, and metadata, such as protocol or disease pack hints, so teams can recognize fit quickly. In organization settings you pick a template from that library and either activate it as-is or open it in the builder to tune blocks before patients see it. Templates marked for default enablement can also seed new patients when your organization has not forked a custom version yet. What you enable drives rendered views under Patient state. At configuration time there is no parallel ad hoc document per patient.
Organization-owned templates keep a stable lineage id across edits so you can track a single “concept” through drafts and published versions; the catalog remains the shared starting point for new organizations and new use cases.

​
Template builder & block configuration

Every organization-scoped view is an editable template: a named view key plus an ordered list of blocks. You can start from scratch (empty draft) or duplicate an existing template (from the catalog or your own) and adjust blocks, prompts, and inputs from there. Templates move through draft, inactive (finalized but not serving patients), and active (wired into platform config and generating). Saving an edit typically creates a new versioned document; activating a version updates which template patients resolve to and kicks off background work to align existing patients’ view snapshots with the new block list.

​
Per-block configuration

Each block has a stable block_id and display name. Generation behavior is driven by execution mode, inputs, and mode-specific fields:
ModeRole
PipelineDeterministic ETL over materialized patient state. Requires a configured list of ETL pipeline operations; may also bind inputs from state. Skips when required inputs are missing.
AgentLanguage-model-driven block. Requires a prompt that instructs the model; typically uses resolved inputs and skips when those inputs have no usable content. Suited to narrative synthesis when facts are present.
  • Inputs: dot-path references into materialized state (each marked required or optional). The orchestrator resolves these before generation.
  • Content shape: structured vs unstructured influences how results are written and validated; optional JSON output_schema constrains structured payloads.
  • Cadence: base_cadence_days sets the minimum gap between successful runs for that block so views do not regenerate continuously.
  • Confidence: optional specs so block outputs can carry scored confidence alongside content.
Changing paths, modes, or block order after go-live affects agents and Console readers that depend on a stable shape (validate in draft, then activate when ready). New or expanded views may show empty blocks until scheduled generation or propagation catches up for each patient.

​
Agentic configuration (assistant)

This feature is in beta. Behavior and availability may change as we expand support.
In the Console Template Builder, you can open a chat session with Olira’s configuration assistant instead of only clicking through forms. The flow is designed around your current draft: the UI saves work in progress, then the assistant receives the draft template context, including view key and block list, so you can describe outcomes in natural language (“surface labs and toxicities together”, “add a short adherence narrative”) and iterate on the same template interactively.
Always review the resulting block list, inputs, and prompts before activating a template for patients.