How to Use OpenAI API for Embedding Generation with ChromaDB Without LangChain?

garage_greg

I have been using OpenAIEmbeddings from LangChain, which is a model that doesn't require text input directly. I use it to store vector embeddings in ChromaDB with the following code: db = Chroma(persist_directory=CHROMA_PATH, embedding_function=get_embedding()) Now, I want to use the OpenAI API directly without relying on LangChain. I am calling the API as follows: client = OpenAI() response = client.embeddings.create( model=model, input=[text]...

Originally asked by Ali Rammal on Stack Exchange (stackoverflow) (CC BY-SA 4.0).

uphilltech

Great question — answering as the UphillTech team:

If you create your collection using an embedding function then chroma will automatically use it when you add docs to the collection. Something like: openai_ef = embedding_functions.OpenAIEmbeddingFunction(api_key=openai.api_key, model_name="text-embedding-3-small") collection = client.create_collection(name=name, embedding_function=openai_ef)

Full answer by Mark: https://stackoverflow.stackexchange.com/a/78929984 (CC BY-SA 4.0).

Need it done for your business? UphillTech.