Evaluate Pinecone as optional alternative vector store

[tomdyson]

Try out Pinecone as an optional alternative to FAISS. Expected pros: smaller container, lower memory use. Expected cons: slower indexing and querying because of network latency, cost for large document stores.

[tomdyson]

From https://langchain.readthedocs.io/en/latest/ecosystem/pinecone.html:

1. Add client as dependency: pip install pinecone-client (add to README and Dockerfile)
2. Add import to microllama.py: from langchain.vectorstores import Pinecone

From Pinecone notebook example:

import pinecone 

# initialize pinecone
pinecone.init(
    api_key="YOUR_API_KEY",  # find at app.pinecone.io
    environment="YOUR_ENV"  # next to api key in console
)

pinecone_index_name = "langchain-demo"

Then something like

def build_pinecone_index():
    return Pinecone.from_documents(
        get_text_chunks(), OpenAIEmbeddings(), index_name=PINECONE_INDEX_NAME
    )

def get_pinecone_index(index_name=PINECONE_INDEX_NAME):
    return Pinecone.from_existing_index(index_name, OpenAIEmbeddings())

[tomdyson]

Initial impressions are the Pinecone is noticeably slower than FAISS. I'd need to refactor get_index to support multiple index types. The check for an existing index (pinecone.describe_index("index-name")) is ~0.7s.