runnable_family's Introduction

Runnable Family

A python library implementing a family of Runnables in langchain like loopback, self-consistent, self-refine, and self-translate.

Requirements

Python 3.10 or higher

See the pyproject.toml file for the required packages.

Installation

First, you should create a virtual environment before installing the package.

On Unix or MacOS:

python -m venv venv
source venv/bin/activate

On Windows:

python -m venv venv
venv\Scripts\activate

After creating the virtual environment, you can install the package using pip:

pip install git+https://github.com/hmasdev/runnable_family

or you can clone the repository and install the package using pip:

git clone https://github.com/hmasdev/runnable_family
cd runnable_family
pip install .

Usage

Just import your favorite classes.

from runnable_family.basic import RunnableConstant, RunnableAdd, RunnablePartialLambda, RunnableLog
from runnable_family.loopback import RunnableLoopback
from runnable_family.gacha import RunnableGacha
from runnable_family.random import RunnableRandomBranch
from runnable_family.runnable_diff import RunnableDiff
from runnable_family.self_consistent import RunnableSelfConsistent
from runnable_family.self_refine import RunnableSelfRefine
from runnable_family.self_translate import RunnableSelfTranslate

They are inherited from the Runnable class, so you can use them as a Runnable, that is, you can use them as chain components.

See basic-examples.ipynb to see how to use them.

Development

Fork the repository: https://github.com/hmasdev/runnable_family

Clone the repository

git clone https://github.com/{YOURE_NAME}/runnable_family
cd runnable_family

Create a virtual environment

python -m venv venv
source venv/bin/activate

Install the required packages
```
pip install -e .[dev]
```
Checkout your working branch
```
git checkout -b your-working-branch
```
Make your changes

Test your changes

pytest
flake8 runnable_family tests
mypy runnable_family tests

Commit your changes

git add .
git commit -m "Your commit message"

Push your changes
```
git push origin your-working-branch
```
Create a pull request: https://github.com/hmasdev/runnable_family/compare

License

MIT

Author

hmasdev

runnable_family's People

Contributors

Watchers

runnable_family's Issues

Scheduled Tests Failed

https://github.com/hmasdev/runnable_family/actions/runs/10189680109

`RunnableLoopback` is too slow.

I have used the following script performance.py to compare RunanbleLoopback and a equivalent graph of langgraph:

from contextlib import contextmanager
import time
from langgraph.graph import Graph, END
from langchain_core.runnables import RunnableLambda, RunnablePassthrough
from runnable_family import RunnableLoopback

countdown = RunnableLambda(lambda n: n-1)

loopback_chain = RunnableLoopback(countdown, RunnableLambda(lambda n: n > 1), RunnablePassthrough())  # noqa

workflow = Graph()
workflow.add_node('countdown', countdown.invoke)
workflow.add_conditional_edges('countdown', lambda n: "continue" if n > 1 else 'end', {'continue': 'countdown', 'end': END})  # noqa
workflow.set_entry_point('countdown')
loopback_graph_compiled = workflow.compile()

@contextmanager
def measure_time():
    start = time.time()
    yield
    print(time.time() - start)

with measure_time():
    loopback_chain.invoke(100)
with measure_time():
    loopback_graph_compiled.invoke(100, {"recursion_limit": 1000})

The results of 3 trials as follows:

$ python performance.py 2>/dev/null
1.2169935703277588
0.6739192008972168
$ python performance.py 2>/dev/null
1.1129276752471924
0.7416691780090332
$ python performance.py 2>/dev/null
1.158017873764038
0.7393209934234619

Recommend Projects