testset_generation.py

from ragas.llms.haystack_wrapper import HaystackLLMWrapper
from ragas.embeddings.haystack_wrapper import HaystackEmbeddingsWrapper
from haystack.components.generators.openai import OpenAIGenerator
from haystack.components.embedders import OpenAITextEmbedder
from haystack.utils import Secret
import json
import os
from dotenv import load_dotenv

load_dotenv()  # This loads variables from .env into the environment
api_key = Secret.from_token(os.getenv("NEBIUS_API_KEY"))

llm = HaystackLLMWrapper(OpenAIGenerator(
        api_base_url="https://api.studio.nebius.com/v1/",
        model="meta-llama/Llama-3.3-70B-Instruct",
        api_key=api_key
    ))

embedding = HaystackEmbeddingsWrapper(OpenAITextEmbedder(
    api_base_url="https://api.studio.nebius.com/v1/",
    model="BAAI/bge-en-icl",
    api_key=api_key,
))
from ragas.testset import TestsetGenerator
from langchain_core.documents import Document as LCDocument

file_path="ltu_programme_data.json"
with open(file_path, 'r', encoding='utf-8') as f:
    data = json.load(f)
print(f"Successfully loaded {len(data)} records from {file_path}")
lcdocs = []
for i, item in enumerate(data):        
    lcdocs.append(LCDocument(item['content']))
# If no documents are found, provide a helpful message
if not lcdocs:
    print("No documents found in the Qdrant store. Make sure documents are indexed first.")


generator = TestsetGenerator(llm=llm, embedding_model=embedding)
dataset = generator.generate_with_langchain_docs(lcdocs, testset_size=10)
# Save the generated test samples to a JSON file
dataset.to_jsonl("testset.jsonl")

print(f"Saved {len(dataset)} test samples to testset.jsonl")