Search Intelligence — LLM Provider
- Community
See It In Action
AI-Powered SEO Metadata Drafting for Magento
The LLM Provider is the bridge between Search Intelligence Core and any OpenAI-compatible language model. It builds deterministic context packets from your catalog and advisory queue, sends them to the LLM (GPT-4o mini by default), and returns structured JSON proposals for meta titles, meta descriptions, and other SEO fields.
Everything runs with dry run enabled by default — no live API calls, no catalog writes, no surprises. You see the packet structure, token estimate, and proposal format before committing a single API credit.
- OpenAI-compatible endpoint with configurable model slug
- Deterministic packet builder: site, bucket, entity, and action types
- Daily call budget with budget-remaining counter
- Strict JSON proposal validator (field, proposed_value, reason, confidence, evidence_ids)
- Dry-run preview page — see packet structure without making API calls
- Advisory only by default: proposals require admin approval before any writes
Safe Token Budget Management
Every outbound request is constrained by a configurable daily call budget. The connector tracks calls-today and budget-remaining in real time, preventing runaway API spend during aggressive advisory cycles or automated optimization runs.
The packet shrinker trims context deterministically when token estimates exceed the configured budget — ensuring the LLM always receives a well-formed, in-budget packet regardless of catalog size.
- Daily call budget enforced per store configuration
- Deterministic packet shrinker for oversized contexts
- Token estimate shown before any live call
- Packet hash for deduplication across advisory runs
- Last call timestamp with UTC display
Requirements
- Magento 2.4.4+ · PHP 8.1+
toweringmedia/module-search-intelligence-coretoweringmedia/module-ai-core2.1+- OpenAI API key (or compatible endpoint)
- Composer:
toweringmedia/module-search-intelligence-llm