Embeddings

An embedding model converts a piece of text into a list of numbers (a vector), typically hundreds or thousands of dimensions long. The key property is that texts with similar meaning end up as vectors that are mathematically close together, even if they share no exact words — 'reset my password' and 'I forgot my login' land near each other, while 'reset my password' and 'store opening hours' land far apart. This is what powers semantic search: instead of matching keywords, the system finds the stored content whose vector is nearest to the query's vector.

Embeddings are the foundation of retrieval-augmented generation (RAG): a company's documents, FAQs, or product catalog are converted to embeddings and stored in a vector database, so when a customer asks a question in Arabic or English, the system retrieves the most relevant passages before the LLM generates an answer — grounding responses in real company content instead of the model guessing. Quality of Arabic embeddings varies significantly between providers, since many embedding models are trained predominantly on English text, which is a key evaluation point before building any Arabic-language search or RAG system.