Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/jxnl/kura/llms.txt

Use this file to discover all available pages before exploring further.

​
Message

Represents a single message within a conversation.
​
created_at
datetime
required
Timestamp when the message was created
​
role
Literal['user', 'assistant']
required
Role of the message sender
​
content
str
required
Text content of the message

​
Example

from kura.types import Message
from datetime import datetime

message = Message(
    created_at=datetime.now(),
    role="user",
    content="How do I implement a binary search tree?"
)

​
Conversation

Represents a complete conversation with metadata.
​
chat_id
str
required
Unique identifier for the conversation
​
created_at
datetime
required
Timestamp when the conversation was created
​
messages
list[Message]
required
List of messages in the conversation
​
metadata
metadata_dict
required

​
Class Methods

​
from_hf_dataset()

Load conversations from a Hugging Face dataset. 
@classmethod
def from_hf_dataset(
    cls,
    dataset_name: str,
    split: str = "train",
    max_conversations: Union[int, None] = None,
    chat_id_fn=lambda x: x["chat_id"],
    created_at_fn=lambda x: x["created_at"],
    messages_fn=lambda x: x["messages"],
    metadata_fn=lambda x: {},
) -> list["Conversation"]
Parameters:
  • dataset_name (str): Name of the HuggingFace dataset
  • split (str): Dataset split to load, defaults to “train”
  • max_conversations (Union[int, None]): Maximum number of conversations to load
  • chat_id_fn (Callable): Function to extract chat_id from dataset item
  • created_at_fn (Callable): Function to extract created_at from dataset item
  • messages_fn (Callable): Function to extract messages from dataset item
  • metadata_fn (Callable): Function to extract metadata from dataset item
Example: 
from kura.types import Conversation

conversations = Conversation.from_hf_dataset(
    "allenai/WildChat-nontoxic",
    split="train",
    max_conversations=1000,
    metadata_fn=lambda x: {
        "model": x["model"],
        "toxic": x["toxic"],
    }
)

​
from_claude_conversation_dump()

Load conversations from a Claude conversation export file. 
@classmethod
def from_claude_conversation_dump(
    cls,
    file_path: str,
    metadata_fn: Callable[[dict], metadata_dict] = lambda x: {},
) -> list["Conversation"]
Parameters:
  • file_path (str): Path to the Claude conversation dump JSON file
  • metadata_fn (Callable): Function to extract metadata from conversation dict
Example: 
from kura.types import Conversation

conversations = Conversation.from_claude_conversation_dump(
    "conversations.json",
    metadata_fn=lambda x: {"project": x.get("project_uuid")}
)

​
generate_conversation_dump()

Save a list of conversations to a JSON file. 
@classmethod
def generate_conversation_dump(
    cls, conversations: list["Conversation"], file_path: str
) -> None
Parameters:
  • conversations (list[Conversation]): List of conversations to save
  • file_path (str): Output file path
Example: 
from kura.types import Conversation

Conversation.generate_conversation_dump(conversations, "backup.json")

​
from_conversation_dump()

Load conversations from a previously saved conversation dump. 
@classmethod
def from_conversation_dump(cls, file_path: str) -> list["Conversation"]
Parameters:
  • file_path (str): Path to the conversation dump JSON file
Example: 
from kura.types import Conversation

conversations = Conversation.from_conversation_dump("backup.json")

​
Complete Example

from kura.types import Conversation, Message
from datetime import datetime

conversation = Conversation(
    chat_id="chat-123",
    created_at=datetime.now(),
    messages=[
        Message(
            created_at=datetime.now(),
            role="user",
            content="What is recursion?"
        ),
        Message(
            created_at=datetime.now(),
            role="assistant",
            content="Recursion is when a function calls itself."
        )
    ],
    metadata={
        "model": "claude-3-sonnet",
        "token_count": 150,
        "tags": ["programming", "computer-science"]
    }
)