v1.4.0

Streaming updates

We've done some notable improvements on streaming:

• Better streaming performance on concurrent requests
• Support streaming from multiple components - docs
• Using on_pipeline_end callback - docs
• Accessing pipeline intermediate outputs from streaming helpers - docs
• Support for hybrid streaming on async pipelines with sync-only components - docs

What's Changed

• Add support to automatic "hybrid" streaming by @mpangrazzi in #182
• Docs update by @mpangrazzi in #183
• chore: Use builtins for list and dict where possible by @sjrl in #186

Full Changelog: v1.3.0...v1.4.0

v1.3.0

What's Changed

• Add include_outputs_from support on pipelines execution by @mpangrazzi in #180
• Change Hayhooks docs branding by @bilgeyucel in #181

Full Changelog: v1.2.0...v1.3.0

v1.2.0

New features

We now support streaming from multiple components - See related docs for more details!

---

What's Changed

• Support streaming from multiple pipeline components by @mpangrazzi in #178

Full Changelog: v1.1.0...v1.2.0

v1.1.0

What's Changed

• Hayhooks documentation website by @mpangrazzi in #167
• fix docs to use deepset/hayhooks:main Docker image by @anakin87 in #169
• docs: add warning about OpenAIChatGenerator when using Open WebUI events by @anakin87 in #170
• chore: Remove email address from pyproject.toml by @julian-risch in #171
• Update haystack-ai dep version by @mpangrazzi in #172
• chore: Update license.md with additional legal information by @julian-risch in #173
• Ensure UTF-8 is properly supported by @mpangrazzi in #175
• Fix docs links in README.md by @sjrl in #177
• feat: When using streaming_generator return final pipeline result as a final streaming chunk by @sjrl in #162

New Contributors

• @julian-risch made their first contribution in #171
• @sjrl made their first contribution in #177

Full Changelog: v1.0.1...v1.1.0

v1.0.1

What's Changed

• Fix CLI command for upload files managing correctly file objs by @mpangrazzi in #165

Full Changelog: v1.0.0...v1.0.1

v1.0.0

New features

• We have rewritten YAML support for deploying YAML-serialized Haystack agents / pipelines.

Breaking changes

• We've removed hayhooks pipeline deploy command, in favour of hayhooks pipeline deploy-files (for PipelineWrapper-based deployments) and hayhooks pipeline deploy-yaml for deploying YAML serialized pipelines.

Extra

• We made linting and type checking more consistent with Haystack's one.
• We did some iteration on improving the Docker image (reduced size)

---

What's Changed

• Docker image improvements by @mpangrazzi in #153
• Docker image improvements by @mpangrazzi in #154
• Raise a PipelineWrapperError if run_api or run_api_async have no return type by @mpangrazzi in #152
• chore: adopt ruff rules from Haystack and enforce them by @anakin87 in #155
• chore: set required_variables in Prompt Builders to reduce noise in logs by @anakin87 in #158
• ci: make mypy run on PRs + fix some types by @anakin87 in #157
• Make mypy ignore errors in tests.* modules by @mpangrazzi in #160
• Breaking: Remove legacy hayhooks pipeline deploy ; introduce hayhooks pipeline deploy-yaml to deploy YAML pipelines with required inputs/outputs fields by @mpangrazzi in #161

Full Changelog: v0.10.1...v1.0.0

v0.10.1

What's Changed

• chore: fix typos in README by @anakin87 in #147
• Fix RAG example by @mpangrazzi in #148
• Force media_type on Form to make Swagger use multipart/form-data content-type by @mpangrazzi in #149
• Add a /status endpoint to MCP server by @mpangrazzi in #150
• Refactored imports to lower import time and speedup CLI by @mpangrazzi in #151

Full Changelog: v0.10.0...v0.10.1

v0.10.0

Agents support

Agents are now first class citizens on Hayhooks. Check how easy is to deploy an async Agent!

Support of open-webui events

Now you can trigger some open-webui events to make UX better when deploying pipelines or agents.

Check the full code example!

Added hooks for intercepting Tool calls

We've also added hooks to intercept tool calls (so you can send events also from there!).

.

Check the full code example!

Add example index

Now you can have a better view of existing examples here.

MCP

We've added the support of Streamable HTTP transport to Hayhooks MCP Server.

---

What's Changed

• Add Streamable HTTP MCP transport by @mpangrazzi in #128
• feat: Read hayhooks version locally from package metadata by @vblagoje in #127
• Better Agent and open-webui events support by @mpangrazzi in #132
• Updated documentation for open-webui events integration by @mpangrazzi in #137
• chore: fix examples in README by @anakin87 in #138
• Add specific README for events example by @mpangrazzi in #141

Full Changelog: v0.9.1...v0.10.0

v0.9.1

What's Changed

• Rename HAYHOOKS_ADDITIONAL_PYTHONPATH to HAYHOOKS_ADDITIONAL_PYTHON_PATH by @mpangrazzi in #126
• Fix request model creation and MCP execution for pipeline wrappers with only run_api_async implemented by @mpangrazzi in #125

Full Changelog: v0.9.0...v0.9.1

v0.9.0

Async support for pipeline wrappers

Now you can implement run_api_async and run_chat_completion_async methods in your pipeline wrapper.

You can now also make use of async_streaming_generator if you need to stream pipeline output to from async methods.

For example:

from pathlib import Path
from typing import AsyncGenerator, List
from haystack import AsyncPipeline
from haystack.dataclasses import ChatMessage
from hayhooks import (
    get_last_user_message,
    BasePipelineWrapper,
    async_streaming_generator,
)

SYSTEM_MESSAGE = "You are a helpful assistant that can answer questions about the world."


class PipelineWrapper(BasePipelineWrapper):
    def setup(self) -> None:
        pipeline_yaml = (Path(__file__).parent / "question_answer.yml").read_text()
        self.pipeline = AsyncPipeline.loads(pipeline_yaml)

    async def run_api_async(self, question: str) -> str:
        result = await self.pipeline.run_async(
            {
                "prompt_builder": {
                    "template": [
                        ChatMessage.from_system(SYSTEM_MESSAGE),
                        ChatMessage.from_user(question),
                    ]
                }
            }
        )
        return result["llm"]["replies"][0].text

    async def run_chat_completion_async(self, model: str, messages: List[dict], body: dict) -> AsyncGenerator:
        question = get_last_user_message(messages)

        return async_streaming_generator(
            pipeline=self.pipeline,
            pipeline_run_args={
                "prompt_builder": {
                    "template": [
                        ChatMessage.from_system(SYSTEM_MESSAGE),
                        ChatMessage.from_user(question),
                    ]
                },
            },
        )

Full example is available here.

---

What's Changed

• Better example video resolution on Hayhooks Core MCP Tools docs by @mpangrazzi in #121
• Add support for async pipeline wrapper methods by @mpangrazzi in #122

Full Changelog: v0.8.0...v0.9.0