Vector Store

After we turn text into embeddings (those special number lists), we need a place to keep them. This is where vector databases come in. Think of a vector database as a huge library where instead of books, we store vectors. These databases are smart because they can look inside these vectors to find meanings, similarities, or differences. This helps us search for information based on its meaning, not just keywords.

Why Vector Databases?

Vector databases are important for a few reasons:

• Semantic Search: They let us search by meaning. For example, if you search for "ancient civilizations," it can find documents related to "Egyptians" or "Mayans," even if those exact words aren't used.

• Finding Relationships: They help us see how different pieces of information are related. This could mean finding documents that talk about the same thing in different words.

• Understanding Context: They're good at understanding the context or the deeper meaning of the text, which is great for answering complex questions or making recommendations.

How Vector Stores Help LLMs

Vector stores act as smart helpers for LLMs by doing a few important things:

• Providing Context: They bring forward relevant information as context. This means if someone asks a question, the vector store finds the most related information to help the LLM give a better answer.

• Enhancing Responses: With the right context, LLMs can create answers that are not just accurate but also fit well with the topic or question. This is like having a conversation where the other person really "gets" what you're talking about.

Example: Using Chroma for Storing Embeddings

When we've got our document text broken down into smaller chunks and turned into embeddings, we can store them in a vector database like Chroma. Chroma is one of the databases we mentioned that's good for this job. When setting this up, it's important to watch out for circular icons – these help guide us in making the right connections between the chunked documents, their embeddings, and the Chroma database. This step ensures that all our prepared and processed text is stored correctly and ready for us to search through and analyze, making our language model even more powerful and insightful.

Conclusion

Here we have provided an example case with Chroma but there are other vector stores as well. To know more about VectorStores and their configuration refer their component documentation here.