SitongGong's picture
merge code dirs 1031 (part 3)
4c30587 verified
Raw History Blame Contribute Delete
1.38 kB
"""
Data processing module for Agentic Memory.
This module provides data processors for different datasets with a unified interface.
The core abstraction is chunk-based processing where long context is split into
chunks for sequential reading and memory operations.
Usage:
from src.data_processing import get_processor
# Get processor for a specific dataset
processor = get_processor("locomo", chunk_mode="turn-pair")
# Process a single sample
sample = processor.process(raw_data)
# Access processed data
chunks = sample.chunks # List of text chunks
qa_list = sample.qa_list # List of QA items
"""
from .base import (
DataProcessor,
DataSample,
ChunkMode,
MultiDatasetProcessor,
get_processor,
register_processor,
list_processors,
)
from .locomo import LoCoMoProcessor
from .longmemeval import LongMemEvalProcessor
from .hotpotqa import HotpotQAProcessor
from .alfworld import ALFWorldOfflineDataset, chunk_trajectories_by_tokens
__all__ = [
# Base classes
"DataProcessor",
"DataSample",
"ChunkMode",
"MultiDatasetProcessor",
# Registry functions
"get_processor",
"register_processor",
"list_processors",
# Concrete processors
"LoCoMoProcessor",
"LongMemEvalProcessor",
"HotpotQAProcessor",
"ALFWorldOfflineDataset",
"chunk_trajectories_by_tokens",
]