AI: Intelligence with a defined role.
The AI domain covers the language models and retrieval tools behind assistants, chatbots, AI search, document analysis and other intelligent product features.
AI is a component inside a wider system. A model needs instructions, business rules, data sources, an interface, integrations and monitoring around it.
AI is most useful when it solves a real problem. The aim is to give it a defined role inside a controlled workflow, not simply to add it.
Gemini Pro
Google's Gemini family of large language models, used through an API.
Models differ in strengths, cost and capabilities, so the model is matched to the use case.
Anthropic Claude
Anthropic's Claude family of large language models, used through an API.
Having more than one model family available means the model can be matched to the task.
OpenAI
OpenAI's models, used through an API.
Having more than one model family available means the model can be matched to the task.
pgvector
An extension that adds vector similarity search to PostgreSQL.
The data that powers AI search can live in the same database as the rest of the product.
LangChain & RAG
RAG (retrieval-augmented generation) gives a model relevant passages from your own content before it answers. LangChain is a framework for composing AI workflows of that kind.
Grounding answers in the business's own information can make them more relevant than a model's general knowledge alone.