Product · Insights & AI
Configurable analysis, from one-off runs to scheduled agents
Insights run block chains against your stakeholder/system requirements via the analysis service. Beyond that, AI assists in almost every corner of processing: rewriting, batch editing, duplicate/conflict handling, and natural-language search.
Overview, Insights
Insight = saved configuration with target (stakeholder/system/both), language, context flags, and block list (JSON).
InsightResult = a run with status and resultJSON that the UI can visualize.
InsightAgent = schedule (scheduleTime, enabled, nextRunAt). The scheduler creates new results when the time arrives, skips if a run is already active, and can skip if nothing relevant changed since lastRunAt.
Blocks execute in topological order; each block calls /api/insights/generate-block with data from earlier blocks.
AI across the platform
Templates and AI-assisted scaffolding
/insights/generate-blocks suggests blocks from a description and domain. /insights/customize-template adapts existing templates with natural language. Easier to start from 'we want to find gaps in area X' without writing JSON.
Rewriting and generation
rewrite-systemkrav / rewrite-intressentkrav from an instruction, suggest-divide-intressentkrav to multiple items, create-intressentkrav-from-instruction, area suggestions, system requirements from a description.
Smart batch editing
The asynchronous job smart-edit-batch runs one instruction across many requirements, saving time on large reformulations.
Context and query builder
Upload PDF/Word/txt → reusable project context string. AI-generated test items and AI search for requirements in V&V plans. Natural language → saved filter query (/queries/ai-build).