AI
AI Search

Documentation

Complete guide to installing and configuring AI Search

Module Overview

AI Search for OpenCart is an extension that replaces the standard store search with semantic AI-powered search. The module uses vector embeddings to understand query context and find relevant products.

Key Features

  • Semantic search (understands context, not just keywords)
  • Automatic typo correction and fuzzy matching
  • Synonym support without additional dictionaries
  • Search by product attributes, filters, and options
  • Query autocomplete (Business+ plans)
  • Mixed results: products, categories, pages
  • Fallback to standard search on failures
  • Automatic re-indexing on product changes

System Requirements

SaaS Version

  • OpenCart 4.0 or higher
  • PHP 8.0 or higher
  • MySQL 5.7+ or MariaDB 10.3+
  • 50MB free space
  • Internet connection (for API requests)

Self-hosted Version

All SaaS version requirements, plus:

  • VPS/dedicated server (minimum 2 CPU, 4GB RAM)
  • Docker and Docker Compose
  • 10GB free space (for Ollama models)
  • GPU recommended (CUDA or ROCm)

Architecture

Vector Index

The oc_ai_embeddings table stores vector representations of products, categories, and pages.

  • › Model: multilingual-e5-large-instruct (1024d)
  • › Search: Cosine similarity in PHP
  • › Cache: file-based (OpenCart Cache)

Trigram Index

The oc_ai_trigrams table for fuzzy autocomplete.

  • › Method: 3-character tokens
  • › Re-ranking: via levenshtein()
  • › Speed: <20ms

Installation

  1. 1
    Download — go to Dashboard → Downloads and download aisearch.ocmod.zip.
  2. 2
    Upload — in your OpenCart admin go to Extensions → Installer, click Upload and select the ZIP file.
  3. 3
    Enable — go to Extensions → Extensions → Modules, find AI Search and click Install, then Edit.
  4. 4
    Enter License Key — paste your key from the dashboard into the License Key field and save.
  5. 5
    Index products — open the Indexer tab and click Start Indexing. Done.

Configuration

The General tab controls core behaviour: mode, API connection, embedding model, and search parameters.

AI Search — General settings tab

Mode

SaaS — uses our cloud API (no server needed, requires license key). Self-hosted — runs Ollama on your own VPS.

Embedding Model

Choose from the table: multilingual-e5-large-instruct (best quality, 100+ languages), nomic-embed-text-v1.5 (fastest, English only), and others. Changing the model requires re-indexing.

Min Query Length

Minimum characters before AI search activates. Recommended: 3.

Fallback to LIKE

If AI search is unavailable, automatically falls back to OpenCart's standard LIKE search. Keep enabled.

Indexing

The Indexer tab lets you control what content gets indexed and how. Select fields carefully — more data improves accuracy, but unnecessary technical fields add noise.

AI Search — Indexer settings: What to Index

Use Field Order (drag to reorder) to prioritise what gets included first — if the total text exceeds the model's token limit, lower-priority fields are dropped. Use Re-index (Farm Queue) for large catalogs to offload embedding generation to the GPU farm.

AI Search — Field Order and Farm Queue

After indexing, switch to the Index tab to verify all products are indexed. You can filter by truncated items to check if any descriptions were cut off.

AI Search — Index tab showing indexed products list

API

API reference coming soon.

Troubleshooting

Troubleshooting guide coming soon.