Scenario, an enterprise AI platform for generative creative assets, migrated from OpenSearch on AWS to Meilisearch to handle tens of millions of indexed assets and 35 million monthly searches per month. The switch was driven by the need for hybrid search combining keyword, vector, and multi-modal (image) queries using Scenario's own custom embeddings. After evaluating Pinecone and Qdrant, Meilisearch was chosen based on a Hugging Face founders' recommendation. Results include near-zero search-related support tickets, 80% of searches being complex vector queries, and an 'Exploration' feature powered by visual similarity search that became a key engagement driver. The next phase involves AI agents that autonomously surface relevant assets and recommend the best-fit generative model for a given task.

5m read timeFrom meilisearch.com
Post cover image
Table of contents
The challenge: scaling search for the AI eraThe solution: a modern hybrid search engineImplementation: building a scalable search infrastructureThe result: scaling volumes with minimal maintenanceConclusion: pushing the boundaries of AI search
753 Impressions