koine-sdk 2.0.3

Python SDK for Koine gateway

koine-sdk

Python SDK for Koine — the HTTP gateway for Claude Code CLI.

Running the Gateway

docker run -d -p 3100:3100 \
  -e CLAUDE_CODE_GATEWAY_API_KEY=your-key \
  -e CLAUDE_CODE_OAUTH_TOKEN=your-token \
  ghcr.io/pattern-zones-co/koine:latest

See Docker Deployment for version pinning and production setup.

Installation

uv pip install koine-sdk
# or: pip install koine-sdk

Quick Start

import asyncio
from koine_sdk import KoineConfig, create_koine

config = KoineConfig(
    base_url="http://localhost:3100",
    auth_key="your-api-key",
    timeout=300.0,
)

async def main():
    koine = create_koine(config)
    result = await koine.generate_text(prompt="Hello, how are you?")
    print(result.text)

asyncio.run(main())

Features

• Text Generation — koine.generate_text() for simple prompts
• Streaming — koine.stream_text() with async iterators
• Structured Output — koine.generate_object() with Pydantic schema validation
• Type Safety — Full type hints for all requests and responses
• Error Handling — KoineError class with error codes

API

Client Factory

koine = create_koine(config)

Creates a client instance with the given configuration. The config is validated once at creation time.

Methods

Method	Description
koine.generate_text(*, prompt, system?, session_id?)	Generate text from a prompt
koine.stream_text(*, prompt, system?, session_id?)	Stream text via Server-Sent Events
koine.generate_object(*, prompt, schema, system?, session_id?)	Extract structured data using a Pydantic model

Types

Type	Description
KoineConfig	Client configuration (base_url, auth_key, timeout, model)
KoineClient	Client interface returned by create_koine()
GenerateTextResult	Text generation response with usage stats
GenerateObjectResult[T]	Object extraction response (generic over schema)
StreamTextResult	Streaming result with async iterators and futures
KoineUsage	Token usage information
KoineError	Error class with code and raw_text

Documentation

See the SDK Guide for:

• Configuration options
• Streaming examples
• Structured output with Pydantic
• Error handling
• Multi-turn conversations

Examples

Runnable examples are available in the examples/ directory. Run from the SDK directory:

cd packages/sdks/python
uv pip install -e ".[dev]"
uv run python examples/hello.py           # Basic text generation
uv run python examples/extract_recipe.py  # Structured output with Pydantic
uv run python examples/stream.py          # Real-time streaming
uv run python examples/conversation.py    # Multi-turn sessions

License

Dual-licensed under AGPL-3.0 or commercial license.