9 ms·
You can use the Wikipedia embedding's released by Cohere to build something pretty easily : https://huggingface.co/Cohere https://huggingface.co/Cohere. If you
by manibatra 3y ago
You can use the Wikipedia embedding's released by Cohere to build something pretty easily : https://huggingface.co/Cohere https://huggingface.co/Cohere.
If you want a completely offline version you'd be running one of the open source LLMs locally. Otherwise put the embeddings in a VectorDB, query it for the context and send it to one of the completion APIs available (OpenAI, etc)