2. Configuration

After installation, point LevelApp at your LLM endpoints and set up credentials.

1. Create a `.env` file in your repo root:

# .env
OPENAI_API_KEY=sk-...
IONOS_API_KEY=eyJ...
IONOS_ENDPOINT=https://inference.de-txl.ionos.com/models
GOOGLE_APPLICATION_CREDENTIALS=./googlecreds.json

2. Verify env vars are loaded

# Python check
python - << 'EOF'
import os, dotenv
dotenv.load_dotenv()
print("OpenAI:", bool(os.getenv("OPENAI_API_KEY")))
print("Ionos:", bool(os.getenv("IONOS_API_KEY")))
EOF

3. Model-specific config

By default, the FastAPI app will register both openai and ionos providers if keys are present. You can override or add new evaluators in app.py:

from level_core.evaluators.service import EvaluationService
from level_core.evaluators.schemas import EvaluationConfig
 
eval_svc.set_config("openai", EvaluationConfig(
  api_key=os.getenv("OPENAI_API_KEY"),
  model_id="gpt-4o-mini"
))

4. Firestore Connection Methods

LevelApp supports two ways to connect to a Firestore database. Choose one of the following methods:

Option 1: Service Account JSON Key

Add the path to your service account JSON key in the .env file:

GOOGLE_APPLICATION_CREDENTIALS=./googlecreds.json

Option 2: Google Cloud SDK Default Credentials

Authenticate locally using the Google Cloud SDK and specify your project ID in the YAML config:

gcloud auth application-default login

database:
  type: firestore
  project_id: your-project-id

Important: You must configure either the Service Account JSON key or the Google Cloud SDK method. If both or neither are provided, LevelApp will raise an error.

Installation First Evaluation