Elasticsearch: search input suggestion [performance issue]

We use Web2py framework and our final goal is to have a fulltext suggestion in the search field of our website: https://www.becomix.me
After many attempts we managed to develop the suggestion during the typing, but is really slow and we have difficulty understanding the cause.

We use a remote Elasticsearch cluster https://cloud.elastic.co/
- Do We need a mapping for search? (Now we use the auto-generated mapping)
- Or we simply need a second cluster? (recommended for production?).

Thank you.