Does Building Your Own Index Work?

[DennisFaucher]

I ran this on a new, clean, Debian VM as well as a running bare metal Debian server. Both failed with errors like this:

inv index-wikipedia-dump --workdir ./workdir --language en

WARNING:root:Skipping batch because the embedding server returned error HTTPConnectionPool(host='localhost', port=6338): Max retries exceeded with url: /embed (Caused by NewConnectionError('<urllib3.connection.HTTPConnection object at 0x7f21d180b810>: Failed to establish a new connection: [Errno 111] Connection refused'))

[s-jse]

This requires configuring an embedding endpoint first. For example, by running a TEI docker using the included script on a supported VM:

WikiChat/tasks/docker_tasks.py

Line 219 in ce8eeec

def start_embedding_docker_container(