Code Sharing Between Wrappers

Hayhooks provides two ways to organize and share code in pipeline wrappers:

1. Relative Imports - Import from sibling modules within the same pipeline folder
2. Shared Python Path - Share code across multiple pipeline wrappers

Relative Imports (Recommended)

Pipeline wrappers are loaded as Python packages, enabling you to use relative imports to organize your code into multiple files within the same pipeline folder.

Structure

my_pipeline/
├── pipeline_wrapper.py    # Main wrapper
├── utils.py               # Helper functions
├── prompts.py             # Prompt templates
└── config.py              # Configuration

Usage

Use Python's relative import syntax (from .module import ...):

# pipeline_wrapper.py
from haystack import Pipeline
from hayhooks import BasePipelineWrapper

# Relative imports from sibling modules
from .utils import process_text, format_response
from .prompts import SYSTEM_PROMPT
from .config import DEFAULT_MODEL

class PipelineWrapper(BasePipelineWrapper):
    def setup(self) -> None:
        self.pipeline = Pipeline()
        # ... setup using imported utilities

    def run_api(self, query: str) -> str:
        processed = process_text(query)
        result = self.pipeline.run({"prompt": {"query": processed}})
        return format_response(result)

Benefits

• No configuration needed - Works out of the box
• Clean organization - Split large wrappers into logical modules
• IDE support - Full autocomplete and type checking
• Tracing compatibility - Works with Phoenix/OpenInference and other tracing libraries

Ruff Linting

If your project uses ruff with the flake8-tidy-imports plugin, you may need to disable the TID252 rule which bans relative imports. Add this comment at the top of your pipeline_wrapper.py:

# ruff: noqa: TID252

Or configure it in your pyproject.toml:

[tool.ruff.lint.flake8-tidy-imports]
ban-relative-imports = "parents"  # Allow sibling relative imports

Example

See examples/pipeline_wrappers/relative_imports for a complete working example.

---

Shared Python Path

For sharing code across multiple pipeline wrappers, add a common folder to the Hayhooks Python path.

Configuration

Set HAYHOOKS_ADDITIONAL_PYTHON_PATH to point to your shared code directory:

Environment Variable.env FileCLI Flag

export HAYHOOKS_ADDITIONAL_PYTHON_PATH='./common'
hayhooks run

# .env
HAYHOOKS_ADDITIONAL_PYTHON_PATH=./common

hayhooks run --additional-python-path ./common

Usage

Once configured, import shared code in your wrappers:

# In your pipeline_wrapper.py
from my_custom_lib import sum_two_numbers

class PipelineWrapper(BasePipelineWrapper):
    def run_api(self, a: int, b: int) -> int:
        return sum_two_numbers(a, b)

Example

See examples/shared_code_between_wrappers for a complete working example.

---

Choosing the Right Approach

Use Case	Recommended Approach
Splitting a large wrapper into modules	Relative Imports
Helpers specific to one pipeline	Relative Imports
Sharing utilities across many pipelines	Shared Python Path
Company-wide libraries	Shared Python Path