Hayhooks

Serve Haystack Pipelines and Agents as production APIs:REST, OpenAI-compatible chat, MCP, A2A, and durable executions that survive restarts.

Documentation · Quick start · Durable execution · Examples

---

Write a small Python wrapper around your Pipeline or Agent, and Hayhooks turns it into typed endpoints, streaming chat backends, and agent-to-agent tools. When the work outlives a request, a restart, or a deploy, run it as a durable execution: checkpointed in Redis, resumable on any replica, and able to wait for a human.

📦 REST APIs from Python Every wrapper or YAML pipeline becomes typed, validated endpoints with OpenAPI docs, file uploads, and custom routes.

♻️ Durable executions Checkpoints, retries, human approval, cancellation, and restart recovery for long-running Pipelines and Agents. Learn more

💬 OpenAI-compatible chat Streaming /chat/completions and /responses backends for Open WebUI, or an embedded Chainlit UI with --with-chainlit.

🛠️ MCP server Each pipeline becomes an MCP tool in Cursor, Claude Desktop, or any MCP client, which can also deploy and run pipelines through chat.

🤝 A2A protocol Publish pipelines as A2A agents (guide) with auto-generated agent cards, so other agents can delegate tasks to them.

📈 Tracing and dashboard OpenTelemetry spans (hayhooks[tracing]) for deploy, run, and durable attempts, with trace/span inspection in hayhooks dashboard --traces. Authenticated local monitoring is enabled by default.

Quick start

1. Install Hayhooks

# Install Hayhooks
pip install hayhooks

2. Start Hayhooks

hayhooks run

3. Create a simple agent

Create a minimal agent wrapper with streaming chat support and a simple HTTP POST API:

from typing import AsyncGenerator
from haystack.components.agents import Agent
from haystack.dataclasses import ChatMessage
from haystack.tools import Tool
from haystack.components.generators.chat import OpenAIChatGenerator
from hayhooks import BasePipelineWrapper, async_streaming_generator

# Define a Haystack Tool that provides weather information for a given location.
def weather_function(location):
    return f"The weather in {location} is sunny."

weather_tool = Tool(
    name="weather_tool",
    description="Provides weather information for a given location.",
    parameters={
        "type": "object",
        "properties": {"location": {"type": "string"}},
        "required": ["location"],
    },
    function=weather_function,
)

class PipelineWrapper(BasePipelineWrapper):
    def setup(self) -> None:
        self.agent = Agent(
            chat_generator=OpenAIChatGenerator(model="gpt-4o-mini"),
            system_prompt="You're a helpful agent",
            tools=[weather_tool],
        )