Kytchen Python SDK

Python client for Kytchen - BYOLLM Context Orchestration.

Installation

pip install kytchen

Quick Start

import asyncio
from kytchen_sdk import KytchenClient, Budget

async def main():
    # Initialize client
    client = KytchenClient(
        api_key="kyt_sk_...",
        base_url="http://localhost:8000",  # For self-hosted
    )

    # Upload a dataset
    dataset = await client.datasets.create("my-data", "document.txt")

    # Query with your own LLM key (BYOLLM)
    result = await client.query(
        query="What is the main topic?",
        dataset_ids=[dataset.id],
        provider="anthropic",
        provider_api_key="sk-ant-...",  # Your Anthropic key
        budget=Budget(max_iterations=10, max_cost_usd=0.50),
    )

    print(result.answer)

    await client.close()

asyncio.run(main())

Features

• Async-first: Built on httpx for efficient async HTTP
• BYOLLM: Bring your own LLM API key (Anthropic, OpenAI)
• Streaming: Real-time query progress via SSE
• Self-host ready: Works with Docker self-hosted instances
• Type-safe: Full type hints and dataclasses

API Reference

KytchenClient

client = KytchenClient(
    api_key="kyt_sk_...",
    base_url="https://api.kytchen.dev",  # Optional
    timeout=60.0,  # Optional
)

Datasets

# Upload
dataset = await client.datasets.create("name", "path/to/file.txt")

# List
datasets = await client.datasets.list()

# Get
dataset = await client.datasets.get("dataset-id")

# Delete
await client.datasets.delete("dataset-id")

Query

# Standard query
result = await client.query(
    query="Your question",
    dataset_ids=["id1", "id2"],
    budget=Budget(max_cost_usd=0.50),
    provider="anthropic",
    provider_api_key="your-llm-key",
)

# Streaming query
async for event in client.query_stream(...):
    print(event.type, event.data)

Self-Hosting

For self-hosted Kytchen instances:

client = KytchenClient(
    api_key="kyt_sk_selfhost_dev_key_12345",
    base_url="http://localhost:8000",
)

See the Docker self-host guide for setup instructions.

License

MIT