@squidcloud/client
    Preparing search index...

    Interface AiKnowledgeBaseGraphSearchOptions

    Fine-grained options for searchMode: 'graph' retrieval (GraphRAG). The graph channel is always RRF-fused with a standard hybrid search; themes-level "global" answering has moved to the globalSummary op of the queryKnowledgeGraph surface.

    interface AiKnowledgeBaseGraphSearchOptions {
        includeGraphContext?: boolean;
        maxHops?: number;
        seedLimit?: number;
    }
    Index

    Properties

    includeGraphContext?: boolean

    When true, attach the traversed AiKnowledgeBaseGraphContext to the response. Default false.

    maxHops?: number

    Number of hops to expand from each seed (traversal depth + 1). Default 2, max 3.

    seedLimit?: number

    Entities seeded via vector search before graph expansion. Default 8, max 25.