DAX LLM Backend

A FastAPI-based backend service that provides LLM-powered DAX query assistance using Anthropic's API.

Prerequisites

• Python 3.12+
• Environment variables setup (see Configuration section)

Installation

1. Clone the repository
2. Install dependencies:

pip install -r requirements.txt

Configuration

Create a .env file in the root directory with the following variables:

ANTHROPIC_API_KEY=your_api_key_here
# Add any other required environment variables

Running the Application

Local Development

uvicorn src.main:app --reload --host 127.0.0.1

Production Deployment (Azure)

Configure the app service with startup.sh as the startup command.

Project Structure

Data

Various scripts utilized to pull DAX function docs, overview of the dax language, etc - all stored in the src/data folder.

LLM

• anthropic_helpers.py: Making API calls and basic data input/output formatting.
• embeddings.py: Simple rag lookup model to add function information to context.
• schemas.py: Expected API input schema.
• system_prompt.py: Building up the system prompt, combination of defined text and appended dynamic values.

Evaluation

All stored in src/evaluate folder - allows for running a test suite where a given question and expected answer can be tested against an LLM DAX query result.

Fast API Sepcifics

Mostly in startup folder - setting up app, routes, throttling, etc.