Skip to main content
Version: 2.0

Configure Queries

When you add data to a corpus, the Query tab lets you experiment with different semantic search, summarization, and chat options. Whether you want to retrieve relevant information, generate summaries grounded in facts with Retrieval-Augmented Generation (RAG), or engage in conversational interactions, explore these options and configurations to find the optimal settings for your applications.

Search your Data​

The Query tab enables you to perform searches against your data while testing different query configuration settings for retrieval and generation.

Query tab

The Pretty Request and Request Code tabs allow you to view and copy the request code used for executing queries. This enables easy integration of query configurations into applications.

A magnifying glass icon allows you to locate specific controls. Use this icon to search and instantly locate a retrieval or generation setting. This is useful for navigating complex query configurations quickly.

Configure retrieval​

The Retrieval section provides different configuration settings to tailor the query experience to your specific needs. For example, configuring filters, reranking results, and refining output context.

Enter your query​

Enter a natural language search query in the Query field. You can either use the default retrieval and generation settings or experiment with different configurations.

Configure search filters​

By default, no filter is applied to your query. Select Filter to

  • Enter a custom filter expression
  • Select predefined filter attributes to refine your search results.
  • View Available filter attributes and Syntax examples.

Configure filters

Hybrid search improves retrieval accuracy by balancing neural search with keyword-based search using the Lambda option.

  • Enable hybrid search to activate hybrid retrieval.
  • Adjust the Lambda range from 0 (pure neural search) to 1 (pure keyword-based search).
  • The recommended range is 0.005 to 0.0075.
  • By default, Lambda is set to 0.5.

Configure hybrid search

Reranking lets you rerank orders of search results also use our available rerankers:

  • Multilingual Ranker – Optimized for multiple languages.
  • User-Defined Function (UDF) Reranker – Customizable ranking function.
  • Maximum Marginal Relevance (MMR) Reranker – Reduces bias by applying a diversity factor.

See Reranking more advanced details about our reranking capabilities.

Configure reranking

Configure relevance tuning​

Fine-tune search relevance by adjusting custom dimension weights.

Enable relevance tuning to control how search results rank based on predefined relevance factors.

See Custom Dimensions for more details.

Configure result context​

Control how much contextual information appears before and after matched search results:

  • Sentences Before: Default: 2
  • Sentences After: Default: 2
  • Characters Before: Default: 0
  • Characters After: Default: 0

If you use character-based limits, Vectara ensures that full sentences display to preserve meaning.

Configure result context

Configure generation​

The Generation section shows the LLM and prompt template used and lets you configure the Language and Summarization options for the query:

Configure generation options

Summarization uses Retrieval-Augmented Generation (RAG) to provide concise summaries in response to your queries. This can be particularly useful when you need an overview of the relevant information within your data.

Configure evaluation​

The Factual Consistency Score automatically evaluates and detects hallucinations in generated output.

  • Score range: 0.0 - 1.0
  • Higher score: More factually accurate responses
  • Lower score: Greater probability of hallucinations
  • Enable calculation of Factual Consistency Score: to monitor and improve response reliability.

Configure evaluation

Configure chat​

Chat enables you to engage in conversational interactions with your data. This can be particularly useful when you need to ask follow-up questions, clarify information, or explore your data in a more interactive manner.

Select Enable Chat to retain the conversational context across your queries.

Save query to history​

Set this toggle on to generate a query history from your query. Vectara enables you to generate query histories from your queries. These are analyses of your queries that give you visibility into how Vectara executed your query through specific steps such as retrieval, reranking, and generation. This is important for troubleshooting issues, inspecting past queries, and optimizing configurations.

View the history of the query​

  1. Ask questions of your data with Send query.
  2. Select the Query history tab to reveal the query history for the most recently executed query.
  3. View the complete breakdown of how the query was executed.
  4. Use this information to optimize your configuration and submit a new query.

Query tab

Query results​

After executing a query, the results appear in the output panel, which includes the following tabs:

  • Pretty Response: Formatted view of the query response.
  • Response JSON: Raw JSON representation of the query output.
  • Query History: Displays the query history for the most recently executed query.
  • Errors: Displays encountered errors (if any).