Sujal Magar 
1. Introduction
ChromaDB is a open-source vector database designed for storing and querying embeddings.
It is widely used in AI applications like semantic search and Retrieval-Augmented Generation (RAG).
2. What are Embeddings?
Embeddings are numerical vector representations of data (text, images, audio).
Example:
Text -> [0.12, 0.98, 0.44, …]
They allow similarity comparison using distance metrics.
3. Key Features
- Vector similarity search
- Metadata filtering
- Persistent storage
- Scalable indexing
- AI framework integrations
4. Architecture
Components:
- Client
- Collections
- Embedding functions
- Storage engine
5. Creating a Client
import chromadb
client = chromadb.PersistentClient(path="./db")6. Collections
Equivalent to tables.
collection = client.get_or_create_collection("docs")7. Adding Data
collection.add(
documents=["AI is powerful"],
ids=["1"],
embeddings=[[0.1, 0.2, 0.3]]
)8. Querying
collection.query(
query_embeddings=[[0.1, 0.2, 0.3]],
n_results=2
)Returns closest vectors.
9. Distance Metrics
- Cosine similarity
- Euclidean distance
- Dot product
10. Metadata Filtering
collection.query(
query_embeddings=[[...]],
where={"topic": "AI"}
)11. Indexing
Uses ANN (Approximate Nearest Neighbor):
- HNSW algorithm
12. Persistence
Stores embeddings on disk for reuse.
13. Integrations
- LangChain
- LlamaIndex
- Haystack
- Hugging Face
- OpenAI embeddings
14. Use Cases
- Semantic search
- Chatbots
- Recommendation systems
- Document retrieval
15. Advantages
- Optimized for vectors
- Fast similarity search
- AI-native design
16. Limitations
- Not for transactional data
- Requires embeddings pipeline