EcomBot Docs

Semantic Search Engine

Discover how EcomBot uses vector search (pgvector) to understand your customers' intent and dynamically filter your products.

EcomBot goes beyond simple keyword search. It features a search and orchestration engine designed to replicate the reasoning of a human salesperson.

Powered by our PostgreSQL and pgvector infrastructure, every product in your WooCommerce catalog is transformed into a data vector (Embedding).

  • Intent Understanding: The chatbot grasps abstract concepts. If a customer searches for "furniture to put the TV on", the AI will suggest "TV Stands", even if the word "TV" isn't in the product title.
  • Typo-Tolerance: Spelling mistakes no longer prevent your products from being discovered.
  • Semantic Expansion: The engine automatically expands queries to cover more synonyms, drastically reducing "no products found" dead ends.

Orchestration & Dynamic Filters

When faced with a large catalog, EcomBot's AI orchestrates queries to refine results:

  1. Multi-Criteria Search: Your customers can combine complex criteria (e.g., "Blue sofa, under €500, on sale").
  2. Hybrid Filtering & Anti-Noise (TF-IDF): The engine detects "semantic noise" (e.g., a search for "relax" matching a sofa whose description says "Relax: No"). It cross-references word frequency (TF-IDF) to isolate true discriminants and prevent false positives (Negative Proof).
  3. Smart Refinement Filters: The AI analyzes search results in real time to extract not only your structured attributes, but also discriminating keywords hidden in product names (e.g., "velvet", "convertible"). It instantly suggests these filters as clickable buttons to help the customer narrow down their choice.
  4. Agentic Orchestration: The AI autonomously decides when to consult the Knowledge Base (to answer a policy question) and when to query the product catalog.

This level of artificial intelligence ensures the agent always provides the most relevant recommendations, maximizing conversions.