Voiceflow Adds Query Rewriting to Improve Knowledge Base Search.

Voiceflow announced on LinkedIn that it has added query rewriting to its agent Knowledge Base system. The feature addresses a persistent challenge in conversational AI: users rarely phrase questions the way documentation is written.

Query rewriting uses a language model to transform ambiguous or conversational user inputs into precise search queries before they reach the Knowledge Base. A user asking "it won't let me log in" gets rewritten into a troubleshooting query; "how much is it?" becomes a pricing search. The rewriting happens based on instructions you provide, giving developers control over how queries are transformed.

The feature sits between the user's message and the Knowledge Base lookup. When enabled, the model takes the user's last message and rewrites it according to your instructions before searching. This preprocessing step aims to improve retrieval accuracy for questions that don't match the literal phrasing of your content.

Developers can enable query rewriting through System tools, then Knowledge base, then Advanced query settings. The official documentation provides configuration details.

The update follows Voiceflow's recent rollout of its Core model, which the company optimized for tool calling and instruction following in agent workflows. Query rewriting appears designed to work with that model's strengths, though it likely functions with other supported LLM providers including Anthropic, OpenAI, and Gemini.

Knowledge Base retrieval quality directly impacts agent usefulness. Even sophisticated semantic search can struggle when users ask colloquial questions or use different terminology than your documentation. Query rewriting attempts to bridge that gap by normalizing inputs before search.

The feature gives Voiceflow users another lever for tuning agent behavior without rebuilding content or retraining embeddings. For teams building customer support or information retrieval agents, it may reduce the need to anticipate every possible phrasing in documentation.

Voiceflow continues to expand its agent development platform with features targeting production deployments. The company recently added global guardrails to prevent agent drift and introduced an environment system for safer deployments across development, staging, and production.