Test and evaluate the Ask-Astro after hybrid search

[sunank200]

Test and evaluate the Ask-Astro results after following:

• hybrid search
• increase the number of documents we retrieve per rewording to 5
• when joining the docs together, we should rerank using something like cohere reranker

Depends on: #147

[vatsrahul1001]

Results after hybrid search implementation Total test - 27 Pass - 13 Fail - 14

Results sheet

As per results responses are very inconsistent and for the same questions opposite answer

Raised below issues

1. https://github.com/astronomer/ask-astro/issues/167
2. https://github.com/astronomer/ask-astro/issues/168
3. https://github.com/astronomer/ask-astro/issues/169

[sunank200]

Conclusion from testing by @vatsrahul1001 @pankajkoti :

1. We should consider rolling back the hybrid approach as both the sources and responses have degraded in quality The results are in the link. Hence, rolling back the deployment to the state with airflow docs.

[vatsrahul1001]

Closing this tasks as per above results we are not going with hybrid for 28th release