Conversational Product Search Agent with Query Rewriting

Does a small local language model improve product search? This project builds a product-search agent with txtai: the txtai Agent rewrites the query, finds brand and colour filters, searches a txtai hybrid index through SQL with bind parameters, drops a filter when fewer than 10 hits come back, and reranks the top 10 titles. The model (qwen2.5:3b) runs in local Ollama on a laptop with RTX 2050 4 GB (Ryzen 7, 8 GB RAM).

NOTE: Research and teaching demo over a public catalogue slice: read-only search, no prices and no purchases. Relevance labels cover only some products per query; an unjudged product scores 0 even if it would suit the shopper, which can understate query rewriting.

Category: Department of Computer Science and Engineering / Agent Development

Agent Development Project AG20 from ProTechSC Academy, with graphical abstract, video demonstration and project guidance.