Skip to main content

VectorQueryPipeline

Description: Pipeline for converting text to Vector index queries and executing vector queries.

VectorQuerier Agent

Overview​

VectorQuerier is an intelligent retrieval agent designed to process natural language (English) queries, translate them into Elasticsearch-compatible queries, and retrieve relevant data from a Vector Store backed by Elasticsearch (ES). It acts as a bridge between user intent and structured vector-based or indexed data, ensuring accurate, secure, and filtered retrieval with consistently formatted results.

Core Responsibilities​

  • Interpret user English queries
  • Convert natural language intent into Elasticsearch DSL queries
  • Execute searches against a Vector Store (Elasticsearch)
  • Apply dataset-level security and user-selected filters
  • Normalize and format retrieved results for downstream agents or UI consumption

Supported Data Set​

  • Vector Store Dataset
    • Supports metadata and field-level filtering
    • Supports multiple vector document simultaneous retrieval

Filters Applied (Security & User-Controlled)​

Dataset-Level Filters​

  • If the selected Vector Store dataset defines filterable fields (e.g., department, region, tenant_id):
    • These fields are automatically detected by VectorQuerier.
    • They are exposed to the UI as selectable filters.

UI Filter Interaction​

  • Each filter field is displayed as a checkbox list.
  • Users can:
    • Select one or more filter values
    • Combine multiple filters

Filtered Retrieval Logic​

  • If no filters are selected:
    • Data is retrieved based on semantic relevance only.
  • If one or more filters are selected:
    • Elasticsearch query includes additional bool.filter clauses.
    • Ensures:
      • Data-level security
      • Tenant or role-based isolation
      • Context-aware retrieval

When to Use VectorQuerier​

Use VectorQuerier when:

  • Users submit queries in natural language (English) and expect relevant data without knowing query syntax.
  • The underlying data is stored in a Vector Store backed by Elasticsearch.
  • Dataset-level security controls or field-based filtering must be enforced during retrieval.
  • Users need the ability to apply selectable filters (for example, via checkboxes) to refine results securely.
  • Retrieved data must be structured, normalized, and consistently formatted for:
    • Downstream agents
    • UI components
    • Reports or templates
    • Automated workflows

Summary​

VectorQuerier is used when users query data using natural language, the data resides in an Elasticsearch-backed vector store, semantic relevance is required, dataset-level security or filtering must be enforced, and the results need to be structured and formatted for downstream agents or UI rendering.