Skip to main content

Modaic, a python framework for packaging and reproducing DSPy programs

Project description

Docs PyPI

Modaic 🐙

Modular + Mosaic, a Python framework for composing and maintaining DSPy applications.

Key Features

  • Hub Support: Load and share precompiled DSPY programs from Modaic Hub
  • Program Framework: Precompiled and auto-loading DSPY programs
  • Automated LM Judge Alignment: Continuously align your LM judges to your preferences while staying at the pareto frontier!

Never lose your progress again. Save everything you need to compare and reproduce optimization runs with GEPA, MIPROv2, etc. - architecture, hyperparameters, precompiled prompts, predictions, git commits, and even datasets - in 5 minutes. Modaic is free for personal use and academic projects, and it's easy to get started.

Installation

Using uv (recommended)

uv add modaic

Optional (for hub operations):

export MODAIC_TOKEN="<your-token>"

Using pip

Please note that you will not be able to push DSPY programs to the Modaic Hub with pip.

pip install modaic

Quick Start

Creating a Simple Program

from modaic import PrecompiledProgram, PrecompiledConfig

class WeatherConfig(PrecompiledConfig):
    weather: str = "sunny"

class WeatherProgram(PrecompiledProgram):
    config: WeatherConfig

    def __init__(self, config: WeatherConfig, **kwargs):
        super().__init__(config, **kwargs)

    def forward(self, query: str) -> str:
        return f"The weather in {query} is {self.config.weather}."

weather_program = WeatherProgram(WeatherConfig())
print(weather_program(query="Tokyo"))
weather_program.push_to_hub("me/my-weather-program")

Save and load locally:

weather_program.save_precompiled("./my-weather")

from modaic import AutoProgram, AutoConfig

cfg = AutoConfig.from_precompiled("./my-weather", local=True)
loaded = AutoProgram.from_precompiled("./my-weather", local=True)
print(loaded(query="Kyoto"))

from hub:

from modaic import AutoProgram, AutoConfig

loaded = AutoProgram.from_precompiled("me/my-weather-program", rev="v2.0.0")
print(loaded(query="Kyoto"))

Architecture

Program Types

  1. PrecompiledProgram: Statically defined programs with explicit configuration
  2. AutoProgram: Dynamically loaded programs from Modaic Hub or local repositories

Support

For issues and questions:

  • GitHub Issues: https://github.com/modaic-ai/modaic/issues
  • Docs: https://docs.modaic.dev

Development

install development dependencies:

uv sync --all-extras

Project details


Release history Release notifications | RSS feed

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

modaic-0.13.0b1.tar.gz (52.4 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

modaic-0.13.0b1-py3-none-any.whl (56.0 kB view details)

Uploaded Python 3

File details

Details for the file modaic-0.13.0b1.tar.gz.

File metadata

  • Download URL: modaic-0.13.0b1.tar.gz
  • Upload date:
  • Size: 52.4 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.7

File hashes

Hashes for modaic-0.13.0b1.tar.gz
Algorithm Hash digest
SHA256 bb2580e843fedc89812b507763efda89d36fb3dfee757db83d02c1f5cf90d6f1
MD5 69eb39e726c1f16db07ad4d0959e02f5
BLAKE2b-256 b2640e5579cb5f5ca33f5e02879cb846a29d0984254b16ff378d2bfb9b4183a6

See more details on using hashes here.

Provenance

The following attestation bundles were made for modaic-0.13.0b1.tar.gz:

Publisher: python-publish.yml on modaic-ai/modaic

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file modaic-0.13.0b1-py3-none-any.whl.

File metadata

  • Download URL: modaic-0.13.0b1-py3-none-any.whl
  • Upload date:
  • Size: 56.0 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.7

File hashes

Hashes for modaic-0.13.0b1-py3-none-any.whl
Algorithm Hash digest
SHA256 7c9b10d6a619429031b0ca8b4bb82e6dc5dc47c2304e816ce116f5a76b84c068
MD5 7c5cdbc26d2654b0988d4affcc7b1dca
BLAKE2b-256 a8f80f908c7b6ef44627618b3f4514aef583e32a534a44e8ca44bfd93ed68509

See more details on using hashes here.

Provenance

The following attestation bundles were made for modaic-0.13.0b1-py3-none-any.whl:

Publisher: python-publish.yml on modaic-ai/modaic

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page