Common use cases
More like this A user reads a ticket and wants to find similar ones. Seed by UID → embedding similarity → return top-k. Graph filter: same product. Threshold: 0.65+.
Related items panel A device page shows related KB articles, incidents, and known issues. Use the graph to collect candidates (articles linked to this device type) and re-rank by embedding similarity to the device description.
Duplicate detection Before creating a new ticket, check if a near-duplicate already exists. Seed by query text → embedding similarity → return results above 0.85. Surface to the user before they submit.
RAG candidate generation
Before sending context to an LLM, retrieve the most relevant chunks. Use StartAtSimilarTextAsync with the user's query as seed. Top-k = 8 is a common starting point for RAG.
var candidates = await Q().StartAtSimilarTextAsync(userQuery, count: 8, nodeTypes: ["Ticket", "Article"]);