Feature Request: Consider adding in memory index for local mode

[rahulgoel]

If i am trying to chat with a one off github repo, running docker feels a bit heavy. A local python file for embeddings, a local HF model for calculating the vectors might also be a reasonable thing to do.

[iuliaturc]

Great suggestion Rahul! Here are some more detailed instructions for new contributors:

• We should integrate Langchain's InMemoryVectorStore.

Steps:

1. Implement a subclass of VectorStore.
2. Add an "inmemory" value to the --vector-store-provider flag.
3. Update build_vector_store_from_args to instantiate it appropriately.
4. Update the README.

[iuliaturc]

Looks like LanceDB also has an in-memory option: https://lancedb.github.io/lancedb/basic/