OpenAI integration with Elasticsearch

[Aleksa1996]

Hello i have integrated Elasticsearch with swirl via search provider and its working flawlessly. Now i would like to use RAG response together with Elasticsearch search provider. Is it possible ? How can i do it ? I followed AI Guide but it's not working. I have set OPENAI_API_KEY as environment variable through docker compose. I also tried to add that variable manually inside container .env, but still does not work. Thanks in advance!

Swirl image: swirlai/swirl-search:latest

Logs i've managed to capture:

logs/celery-worker.log -> [2024-07-16 04:33:13,691: ERROR/ForkPoolWorker-6] federate[]: GenAIGPT_10_9: name 'llm' is not defined

[erikspears]

Hi @Aleksa1996 ,

Happy to try and help get you going with this. If I read the above correctly, you have an Elastic SearchProvider configured, enabled, and working, which is great! In order to include those results in RAG processing, there are a couple of next steps.

1. Do go ahead and setup your OpenAI key and the necessary MSAL values to support RAG in Docker (which you mentioned using) via the export commands outlined in the Quick Start Guide.
2. You then need to add configuration to the page_fetch_config_json section of your Elastic SearchProvider, as outlined in the AI Guide. Try the basic configuration first, like so:

"page_fetch_config_json": {
        "cache": "false",
        "headers": {
            "User-Agent": "Swirlbot/1.0 (+http://swirl.today)"
        },
        "timeout": 10
}, 

Be sure to PUT that change to the SearchProvider and double-check that it saved. With those steps accomplished, refresh your Galaxy UI, and try a fresh search that gets results from your Elastic source. Then, toggle on the "Generate AI Response" option just under the search box. If you don't get a RAG response in a minute or less, grab any errors that have gone by in your Docker app window, and post them here for us to take a look at.

Many thanks,
--Erik