AI Usage & Interaction System

Purpose

This app provides the system for tracking all AI interactions throughout a Django project.

Installation

Simply add the ai application to your INSTALLED_APPS:

INSTALLED_APPS = [
    ...
    'django_spire.ai',
    ...
]

Warning

A properly configured Dandy install is required — Spire uses the Dandy LLM/recorder libraries under the hood.

Usage

Given the probabilistic nature of AI, every interaction — especially ones involving users — should be tracked.

Below we make a simple interaction with the LLM and have the ai app track it:

from dandy import Bot
from dandy.intel import BaseIntel

from django_spire.ai.decorators import log_ai_interaction_from_recorder


class HorseIntel(BaseIntel):
    first_name: str
    breed: str
    color: str
    has_cone_taped_to_head: bool


@log_ai_interaction_from_recorder(actor='Anonymous User')
def generate_horse_intel(user_input: str) -> HorseIntel:
    bot = Bot()
    return bot.llm.prompt_to_intel(
        prompt=user_input,
        intel_class=HorseIntel,
    )


horse_intel = generate_horse_intel('Make me a magical horse that grants wishes!')

The decorator records the interaction (module, callable, user or actor) and the full Dandy recorder trace. Pass user= when the acting user is known — the chat and SMS pipelines do this automatically:

@log_ai_interaction_from_recorder(user=request.user)
def my_ai_work(request, user_input: str) -> None:
    ...

Warning

The log_ai_interaction_from_recorder decorator is designed to be used with the Dandy intelligence library and will not track other LLM libraries properly.

Admin

You can view AI interactions and daily usage in the Django admin under the django_spire_ai app.