guillotina-nuclia 1.0.1

nuclia support for guillotina

This package lets your Guillotina instance talk to Nuclia. Below is the only thing you have to do: drop the utility in app_settings and export your credentials.

Install

pip install guillotina_nuclia

Configure the Nuclia utility

Add this to the settings you pass when starting Guillotina.

import os

app_settings = {
    "load_utilities": {
        "nuclia": {
            # interface that the utility will provide
            "provides": "guillotina_nuclia.utility.INucliaUtility",
            # dotted path to the factory
            "factory": "guillotina_nuclia.utility.NucliaUtility",
            # parameters forwarded to the factory
            "settings": {
                # which LLM model Nuclia will use for generative answers
                "generative_model": os.environ.get("GENERATIVE_MODEL", "chatgpt4o"),
                # your personal Nuclia key (required)
                "nua_key": os.environ.get("NUA_KEY"),
                # hard limit for tokens (optional)
                "max_tokens": os.environ.get("MAX_TOKENS"),
            },
        }
    }
}

Set the environment variables

Environment variables

Variable	Required	Example value	Description
NUA_KEY	Yes	nua_pk_live_…	API token generated in the Nuclia dashboard.
GENERATIVE_MODEL	No	chatgpt4o (default)	LLM used for ask / predict.
MAX_TOKENS	No	2048	Maximum tokens per answer.

Export them before launching Guillotina:

export NUA_KEY=\"nua_pk_live_your_token_here\"
export GENERATIVE_MODEL=\"chatgpt4o\"   # optional
export MAX_TOKENS=2048                 # optional

Done!

Start Guillotina as usual—INucliaUtility is now available everywhere and the built-in routes (@NucliaAsk, @NucliaSearch, etc.) will automatically work.

❤️ Happy coding!

1.0.1 (2025-07-28)

• Fixing KBID auth [nilbacardit26]

• Adding @NucliaAskStream endpoint [nilbacardit26]

1.0.0 (2025-07-23)

• Initial release [nilbacardit26]