Complete ChromaDB Tutorial for AI Developers

steps

1. Install Required Packages: Install chromadb, langchain, openai, and related dependencies.

   pip install chromadb langchain openai tiktoken

2. Verify Installations: Ensure that the installed versions meet your requirements.

   pip show chromadb
   # Output example:
   # Name: chromadb
   # Version: 0.3.21

3. Set Up Environment Variables: Set your API key for OpenAI to use embeddings later.

   pythonCopy

   import os
   os.environ["OPENAI_API_KEY"] = "your_openai_api_key"

steps

1. Load Documents: Use the LangChain DirectoryLoader to load text files or .pdf documents.

   pythonCopy

   from langchain.document_loaders import DirectoryLoader, TextLoader
   
   loader = DirectoryLoader("data_folder", glob="*.txt", loader_cls=TextLoader)
   documents = loader.load()

2. Split Data: Use RecursiveCharacterTextSplitter to divide documents into chunks while maintaining context.

   pythonCopy

   from langchain.text_splitter import RecursiveCharacterTextSplitter
   
   splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=200)
   chunks = splitter.split_documents(documents)
   print(len(chunks))  # Number of document chunks

steps

1. Initialize OpenAI Embeddings: Load the default embedding model using LangChain’s OpenAIEmbeddings.

   pythonCopy

   from langchain.embeddings import OpenAIEmbeddings
   
   embedding = OpenAIEmbeddings()

2. Store Embeddings in ChromaDB: Use the from_documents function from LangChain’s Chroma integration.

   pythonCopy

   from langchain.vectorstores import Chroma
   
   db = Chroma.from_documents(
       chunks, 
       embedding, 
       persist_directory="chroma_db"
   )
   db.persist()

3. Reload Database: After saving, reload the database for later use.

   pythonCopy

   db = Chroma(
       persist_directory="chroma_db", 
       embedding_function=embedding
   )

steps

1. Query the Database: Use the similarity_search method to retrieve relevant chunks.

   pythonCopy

   query = "What are the benefits of using vector databases?"
   results = db.similarity_search(query, k=2)  # Retrieves top-2 results
   print(results)

2. Integrate with LangChain: To enhance responses, integrate ChromaDB with LangChain's RetrievalQA.

   pythonCopy

   from langchain.chains import RetrievalQA
   from langchain.llms import OpenAI
   
   retriever = db.as_retriever(search_kwargs={"k": 2})
   llm = OpenAI(temperature=0)
   qa_chain = RetrievalQA.from_chain_type(
       llm=llm, 
       chain_type="stuff", 
       retriever=retriever, 
       return_source_documents=True
   )
   
   question = "What is ChromaDB used for?"
   response = qa_chain.run(question)
   print(response)