Function _pack_single_community_describe

_cluster_utils.py:351–403

(
    entitys,
    relations,
    max_token_size: int = 12000,
    global_config: dict = {},
)

    def perform_clustering(self, embeddings: np.ndarray, **kwargs) -> List[List[int]]:
        pass
def _pack_single_community_describe(
    entitys,
    relations,
    max_token_size: int = 12000,
    global_config: dict = {},
) -> str:
   
    node_fields = ["id", "entity", "type", "description", "degree"]
    edge_fields = ["id", "source", "target", "description", "rank"]
    nodes_list_data = [
        [
            i,
            entity.get("entity_name"),
            entity.get("entity_type", "UNKNOWN"),
            entity.get("description", "UNKNOWN"),
            entity.get("degree",1)
        ]
        for i,entity in enumerate(entitys)
    ]
    nodes_list_data = sorted(nodes_list_data, key=lambda x: x[-1], reverse=True)
    nodes_may_truncate_list_data = truncate_list_by_token_size(
        nodes_list_data, key=lambda x: x[3], max_token_size=max_token_size // 2
    )
    edges_list_data = [
        [
            i,
            relation.get("src_tgt"),
            relation.get("tgt_src"),
            relation.get("description", "UNKNOWN"),
        ]
        for i, relation in enumerate(relations.values())
    ]
    edges_list_data = sorted(edges_list_data, key=lambda x: x[-1], reverse=True)
    edges_may_truncate_list_data = truncate_list_by_token_size(
        edges_list_data, key=lambda x: x[3], max_token_size=max_token_size // 2
    )

    truncated = len(nodes_list_data) > len(nodes_may_truncate_list_data) or len(
        edges_list_data
    ) > len(edges_may_truncate_list_data)

  
    nodes_describe = list_of_list_to_csv([node_fields] + nodes_may_truncate_list_data)
    edges_describe = list_of_list_to_csv([edge_fields] + edges_may_truncate_list_data)
    return f"""
-----Entities-----
```csv
{nodes_describe}
```
-----Relationships-----
```csv
{edges_describe}
```"""
def process_cluster( 
    use_llm_func, embeddings_func, \
    clusters,label,nodes,community_report_prompt,\