Finish chroma onboard

[emrgnt-cmplxty]

No description provided.

[bhavitsharma]

lgtm, taking a deeper look now

[bhavitsharma]

How are these IDs generated?

[emrgnt-cmplxty]

rn for symbols this returns symbol.dotpath. This unpacks the complicated symbol into a simple string like 'MyPath.MyClass.my_method'

[bhavitsharma]

Will this still be used somewhere?

[emrgnt-cmplxty]

I think it's good to have the option of using a simple JSON database for now, but we can remove at some point if we are not using it.

[bhavitsharma]

anyways, can you confirm if the chroma database will be persisted to disk or something by default? I think if chroma is running via a docker-compose image then it probably will (they provide that option). I think by default it uses the in-memory DB, probably worth checking

[emrgnt-cmplxty]

anyways, can you confirm if the chroma database will be persisted to disk or something by default? I think if chroma is running via a docker-compose image then it probably will (they provide that option). I think by default it uses the in-memory DB, probably worth checking

The constructor allows either in-mem or persistent creation. See this snippet -

class ChromaVectorDatabase(VectorDatabaseProvider, Generic[K, V]):
    """Concrete class to provide a vector database that uses Chroma."""

    def __init__(self, collection_name: str, persist_directory: Optional[str] = None):
        self._setup_chroma_client(persist_directory)
        self._collection = self.client.get_or_create_collection(collection_name)

    def _setup_chroma_client(self, persist_directory: Optional[str] = None):
        """Setup the Chroma client, here we attempt to contain the Chroma dependency."""
        try:
            import chromadb
            from chromadb.config import Settings
        except ImportError as e:
            raise ImportError(
                "Please install Chroma Python client first: " "`pip install chromadb`"
            ) from e
        if persist_directory:
            self.client = chromadb.Client(
                Settings(
                    chroma_db_impl="duckdb+parquet",
                    persist_directory=persist_directory,
                )
            )
        else:
            # A single instance client which terminates at session end
            self.client = chromadb.Client()

[bhavitsharma]

this is good that we're passing the embeddings to chroma.
We need to keep in mind not to exceed Openai's rate limits while creating embeddings. (not related to this PR since you're passing embeddings to chroma directly)