Implementation:Explodinggradients Ragas LocalJSONLBackend Class

Knowledge Sources	Explodinggradients_Ragas
Domains	Storage, Backend
Last Updated	2026-02-10 00:00 GMT

Overview

File-based storage backend using JSONL format for local persistence with data type preservation and datetime handling.

Description

LocalJSONLBackend implements BaseBackend to store data as JSONL (JSON Lines) files on the local filesystem. Unlike the CSV backend, it preserves nested structures and supports datetime serialization/deserialization. Each line in the file is a valid JSON object.

Usage

Use this backend when you need persistent local storage with support for complex nested data structures and datetime fields.

Code Reference

Source Location

• Repository: Explodinggradients_Ragas
• File: src/ragas/backends/local_jsonl.py
• Lines: 13-206

Signature

class LocalJSONLBackend(BaseBackend):
    def __init__(self, root_dir: str) -> None:
        ...

Import

from ragas.backends.local_jsonl import LocalJSONLBackend

I/O Contract

Inputs

Name	Type	Required	Description
root_dir	str	Yes	Root directory for JSONL storage
name	str	Yes	Name of the dataset or experiment
data	List[Dict[str, Any]]	Yes	Records to save as JSONL lines

Outputs

Name	Type	Description
load returns	List[Dict[str, Any]]	Records with datetime fields deserialized
list returns	List[str]	Sorted list of JSONL file names (without extension)

Usage Examples

from ragas.backends.local_jsonl import LocalJSONLBackend
from datetime import datetime

backend = LocalJSONLBackend(root_dir="./my_data")

# Save a dataset with nested data and datetime
backend.save_dataset("eval_results", [
    {
        "question": "What is AI?",
        "contexts": ["AI is a field of computer science.", "AI simulates human intelligence."],
        "score": 0.85,
        "evaluated_at": datetime.now(),
    },
])

# Load the dataset (datetime is preserved)
data = backend.load_dataset("eval_results")
print(data[0]["evaluated_at"])  # datetime object