Consumer and product data covers real-world things and organisations: the food and cosmetics on shop shelves, and the companies behind them. Much of it is built collaboratively, gathered and checked by volunteers in the way an open encyclopaedia is, which makes it openly reusable in a way most commercial product databases are not. For a RAG system, the setup that answers questions by retrieving relevant records and then generating a reply from them, this category is what lets you answer grounded, factual questions about specific products and firms rather than talking in generalities.
Choosing within this category is really about which real-world thing you are answering about. Open Food Facts and its beauty and products siblings are the place to go for packaged goods, ingredients, and labelling. OpenCorporates is the reference for company registrations and structure across many jurisdictions. OpenSanctions is narrower and pointed, aimed at screening people and organisations against sanctions and watchlists. These solve different problems, so let the question pick the source.
The main thing to watch with crowdsourced product data is uneven coverage and quality: some entries are thorough, others are a barcode and little else, and popular regions are far better covered than others. Treat missing or partial records as normal and design retrieval to cope. Licences here can be more particular than usual, since some of these projects use share-alike terms that require you to release derived data under the same licence, which matters if you plan to build a commercial product, so read the terms before you commit. The sources below span food and consumer goods through to company and sanctions data.