Build a small relevance set before changing search settings
For an illustrative replacement-parts catalog, collect a correct part name, a realistic misspelling, a customer's alternate term, an exact part code, and a query with no suitable result. Record which item should appear and which similar item must not be mistaken for it.
Typesense's relevance guide describes several controls that affect matching and ranking. A result appearing somewhere is a different observation from the correct result ranking first. Keep the indexed sample and query set fixed while investigating a specific problem.
| Query category | Expected behavior | Problem to detect |
|---|---|---|
| Correct product name | Relevant part ranks clearly | A popular but unrelated item dominates |
| Misspelled name | The intended part remains discoverable | Overly broad correction changes meaning |
| Approved alternate term | Relevant vocabulary is expanded | A synonym makes unrelated parts equivalent |
| Exact part code | The intended identifier is respected | A nearby code is treated as interchangeable |
| Unavailable part | The limitation is understandable | A weak match is presented as an exact answer |
Tune typo tolerance by field meaning
Typesense supports typo-tolerance parameters and per-field settings. Names and descriptions can benefit from forgiving matching, while identifiers may require stricter treatment. Use the parameter guidance for your installed version instead of applying one sensitivity to every field.
For the sample catalog, a misspelled descriptive word may reasonably find the intended replacement part. Changing a digit in a part code may refer to a different item. Decide what the search interface should communicate when a query resembles an identifier but no exact item exists.
Use synonyms for vocabulary, not for every weak result
Typesense supports one-way and multi-way synonyms. The direction matters: broadening a category term into related named items is different from declaring all the terms interchangeable. Synonyms do not expand structured filters or exact phrase searches in the same way as ordinary query text.
For the catalog example, let a person responsible for the product vocabulary approve the intended relationship. An informal customer term may mean several different parts. Add a counterexample to the relevance set before deploying a broad mapping, so the improvement for one query does not silently damage another.
Inspect token dropping and ranking before adding rules
Typesense can drop query tokens when too few results match, controlled by drop_tokens_threshold. Other parameters influence prefix matching and field relevance. Inspect these mechanisms when a long query returns a result that matches only part of the intended meaning.
For a search such as a particular size of replacement filter, returning the correct product family with the wrong size can look plausible. Compare the actual request parameters with the expected result. A new synonym will not repair a missing indexed field or a configuration that relaxed the query more than the product should allow.
Check the synonym API when upgrading versions
Typesense v30 introduced synonym sets and changed their API routes and required key actions. The current documentation explains migration and how sets are associated with collections or search requests. Existing code written for collection-specific synonyms needs a version-aware review.
Before an upgrade, record the server version, client version, synonym definitions, and the way the application sends queries. Follow the documented backup and migration procedure for your deployment. Rerun the fixed relevance set afterward, including a check that the intended synonym set is actually being used.
Measure useful results and keep a rollback route
For each proposed change, compare the accepted queries and counterexamples side by side. Keep a note of which result moved, why that matters, and which configuration changed. A small approved improvement is easier to maintain than several interacting adjustments with no recorded purpose.
Use real search feedback to expand the evaluation set over time. Separate missing inventory, unclear terminology, and matching configuration as different causes of failure. Typesense is useful when the team can maintain that connection between customer language, indexed data, and a search result the customer can confidently act on.
Sources and editorial method
Official sources linked in this guide were checked on October 27, 2026. Product features, interfaces, and plan limits can change. The examples are proposed workflows, not reports of a hands-on benchmark or customer results. Confirm the documentation for your product version before implementing a workflow.