Introduction to Kura

import asyncio
from kura.types import Conversation
from kura.summarisation import SummaryModel, summarise_conversations
from kura.cluster import ClusterDescriptionModel, generate_base_clusters_from_conversation_summaries
from kura.meta_cluster import MetaClusterModel, reduce_clusters_from_base_clusters
from kura.dimensionality import HDBUMAP, reduce_dimensionality_from_clusters
from kura.visualization import visualise_pipeline_results
from kura.checkpoints import JSONLCheckpointManager
from rich.console import Console

async def main():
    # Load conversations from HuggingFace
    conversations = Conversation.from_hf_dataset(
        "ivanleomk/synthetic-gemini-conversations",
        split="train"
    )
    
    # Configure pipeline
    console = Console()
    checkpoint_manager = JSONLCheckpointManager("./checkpoints", enabled=True)
    
    # Process conversations
    summaries = await summarise_conversations(
        conversations,
        model=SummaryModel(console=console),
        checkpoint_manager=checkpoint_manager
    )
    
    clusters = await generate_base_clusters_from_conversation_summaries(
        summaries,
        model=ClusterDescriptionModel(console=console),
        checkpoint_manager=checkpoint_manager
    )
    
    reduced_clusters = await reduce_clusters_from_base_clusters(
        clusters,
        model=MetaClusterModel(console=console),
        checkpoint_manager=checkpoint_manager
    )
    
    projected_clusters = await reduce_dimensionality_from_clusters(
        reduced_clusters,
        model=HDBUMAP(),
        checkpoint_manager=checkpoint_manager
    )
    
    # Visualize results
    visualise_pipeline_results(projected_clusters, style="rich")

if __name__ == "__main__":
    asyncio.run(main())