Documentation reviewed
Mistral Search Index
Checked
Recheck due
Mistral Search Index is a public-preview managed backend for searchable document chunks. It supports vector, keyword and hybrid retrieval with typed-field filters. Search Toolkit supplies ingestion and retrieval components and can use other storage backends.
- Availability
- Preview. Public preview documented October 30. This is retrieval infrastructure, not a ready-made document assistant or a promise that every workspace has permission.
- How to use it
- API; Library / CLI. Workspace API and Python pipeline tooling. The hosted backend is separate from the Search Toolkit package and its self-operated Postgres/Vespa options.
- Processing and data
- Vendor cloud. Mistral hosts indexed chunks and fields. Extraction, embedding and any answer-generation provider are separate processing decisions. Hosting location, deletion behavior and access controls need account-specific verification.
- Costs and limits
- Pricing not verified; check storage, query, embedding and extraction charges
- Outputs
- Retrieved passages and metadata for an application. A search hit is not a checked answer; test relevance, stale-document handling and source permissions on your own collection.
Requires a workspace API key and create_search_indexes permission. Index creation and schema changes are asynchronous. Preview APIs can change. Storage, query, embedding and extraction pricing, quotas, regional availability and retention terms were not established by this check; verify them before uploading data. No index or API request was created.