Vespa

Build and operate Vespa applications with the Search Toolkit plugin. Define schemas through Python migrations, configure ranking and query profiles, deploy with the mistral-vespa CLI, and manage production clusters.

Note

The Retrieval, Retrievers, and Filters pages are written for a Mistral Search Index. A Vespa store supports search, but it differs in two ways:

No retrieve. Query with search, and tune ranking with query profiles.
No filters. The Vespa backend ignores VectorSearchQuery.filter, so a filtered search returns unfiltered results and doesn't raise an error. To scope queries, use the yql_filter of the schema and the exclude_ids and exclude_source_ids parameters. For yql_filter, see Migration helpers.

Getting started

Follow these guides to build and deploy a Vespa application:

Guide	Purpose
Anatomy of an Application	Understand application packages, schemas, fields, ranking profiles, query profiles, and migrations
Manage Schema	Create and evolve schemas using Python migrations
Manage Ranking	Configure ranking at query time without schema changes
Local Development	Full local development loop: create, deploy, feed, query, iterate
Deploy and Operate	Production configuration, health checks, monitoring
CLI Reference	`mistral-vespa` command reference

Use Vespa as a search backend

To use Vespa in a Search Toolkit pipeline, get a store with app.get_search_index(config, collection_name=...). Pass it to Pipeline(stores=...) for ingestion, and query it with search. For the connection settings, see Deploy and operate.

Installation

uv add "mistralai-search-toolkit[vespa]"

uv add "mistralai-search-toolkit[vespa]"