This page gives you a fast, practical overview of ChromaDB and how it supports modern AI search.
You’ll learn what the engine does, where it fits in an LLM stack, and how to start building with it today.
In brief, it’s an open-source vector database that runs in-process.
Because of that design, developers can prototype quickly, keep everything local, and avoid extra services.
Moreover, the store keeps both embeddings and metadata together, which makes retrieval simple and predictable.
The diagram below shows a typical RAG flow from text to embeddings and then to nearest-neighbour results.
It clarifies how documents move through ingestion, indexing, and querying.

If you want a full LLM workflow, connect this store to a framework that manages chunking, prompts, and evaluation.
For example, read our guide on
LangChain vector stores for RAG.
In addition, the official repository provides reference code and persistence tips.

Scroll to the flashcards for a structured recap.
Then, try the links below for deeper examples and next steps.
For best results, start small, measure recall and latency, then tune query parameters.
Next, add persistence before you ship anything important.
Finally, test your prompts and chunking so retrieval stays relevant.
Explore our hands-on tutorial: LangChain vector stores for RAG.
For code samples and releases, visit the
ChromaDB GitHub repository.