agentics-py 0.3.3a1

Agentics is a Python framework that provides structured, scalable, and semantically grounded agentic computation.

Agentics

Transduction is all you need

Agentics is a Python framework for structured, scalable, and semantically grounded agentic computation.
Build AI-powered pipelines as typed data transformations—combining Pydantic schemas, LLM-powered transduction, and async execution.

---

✨ Why Agentics

Most "agent frameworks" let untyped text flow through a pipeline. Agentics flips that: types are the interface. Workflows are expressed as transformations between structured states, with predictable schemas and composable operators.

---

🚀 Key Features

• Typed agentic computation: Define workflows over structured types using standard Pydantic models.
• Logical transduction (<<): Transform data between types using LLMs (few-shot examples, tools, memory).
• Async mapping & reduction: Scale out with amap and areduce over datasets.
• Batch execution & retry: Built-in batching, retries, and graceful fallbacks.
• Tool support (MCP): Integrate external tools via MCP.

---

📦 Getting Started

Quickstart:

Install Agentics in your current env, set up your environment variable, and run your first logical transduction:

uv pip install agentics-py

set up your .env using the required parameters for your LLM provider of choice. Use .env_sample as a reference.

Find out more 👉 Getting Started: docs/getting_started.md

Examples

Run scripts in the examples/ folder (via uv):

uv run python examples/hello_world.py

---

🧪 Example Usage

from pydantic import BaseModel, Field
from agentics.core.transducible_functions import transducible, Transduce

class ProductDescription(BaseModel):
    name: str
    features: str
    price: float

class ViralTweet(BaseModel):
    tweet: str = Field(..., description="Engaging tweet under 280 characters")
    hashtags: list[str] = Field(..., description="3-5 relevant hashtags")
    hook: str = Field(..., description="Attention-grabbing opening line")

@transducible()
async def generate_viral_tweet(product: ProductDescription) -> ViralTweet:
    """Transform boring product descriptions into viral social media content."""
    return Transduce(product)

# Transform a product into viral content
product = ProductDescription(
    name="Agentics Framework",
    features="Type-safe AI workflows with LLM-powered transductions",
    price=0.0  # Open source!
)

tweet = await generate_viral_tweet(product)
print(f"🔥 {tweet.tweet}")
print(f"📱 {' '.join(tweet.hashtags)}")

Output:

🔥 Stop wrestling with unstructured LLM outputs! 🎯 Agentics gives you type-safe AI workflows that just work. Build production-ready agents in minutes, not weeks. And it's FREE! 🚀
📱 #AI #OpenSource #Python #LLM #DevTools

---

📘 Documentation and Notebooks

Complete documentation available here

Notebook	Description
agentics.ipynb	Core Agentics concepts: typed states, operators, and workflow structure
atypes.ipynb	Working with ATypes: schema composition, merging, and type-driven design patterns
logical_transduction_algebra.ipynb	Logical Transduction Algebra: principles and examples behind <<
map_reduce.ipynb	Scale out workflows with amap / areduce (MapReduce-style execution)
synthetic_data_generation.ipynb	Generate structured synthetic datasets using typed transductions
transducible_functions.ipynb	Build reusable @transducible functions, explanations, and transduction control

✅ Tests

Run all tests:

uv run pytest

---

📄 License

Apache 2.0

---

👥 Authors

Project Lead and Main Contributor

• Alfio Massimiliano Gliozzo (IBM Research) — [e-mail]

Core Contributors

• Junkyu Lee (IBM) — [e-mail]
• Nahuel Defosse (IBM) — [e-mail]
• Naweed Aghmad Khan (IBM) — [e-mail]

Community Contributors

• Christodoulos Constantinides (IBM) — [e-mail]
• Nandana Mihindukulasooriya (IBM) — [e-mail]
• Mustafa Eyceoz (Red Hat) — [e-mail]
• Gaetano Rossiello (IBM) — [e-mail]
• Agostino Capponi (Columbia University) — [e-mail]
• Chunghyun Han (Columbia University) — [e-mail]
• Abhinav Goel (Columbia University) [e-mail]
• Chaitya Shan (Columbia University) — [e-mail]
• Brian Zi Qi Zhu (Columbia University) — [e-mail]

---

🧠 Conceptual Overview

Most “agent frameworks” let untyped text flow through a pipeline. Agentics flips that: types are the interface. Workflows are expressed as transformations between structured states, with predictable schemas and composable operators.

Because every step is a typed transformation, you can compose workflows safely (merge and compose types/instances, chain transductions, and reuse @transducible functions) without losing semantic structure.

Agentics makes it natural to scale out: apply transformations over collections with async amap, and aggregate results with areduce.

Agentics models workflows as transformations between typed states.

Core operations:

• amap(func): apply an async function over each state
• areduce(func): reduce a list of states into a single value
• <<: logical transduction from source to target Agentics
• &: merge Pydantic types / instances
• @: compose Pydantic types / instances

📜 Reference

Agentics implements Logical Transduction Algebra, described in:

• Alfio Gliozzo, Naweed Khan, Christodoulos Constantinides, Nandana Mihindukulasooriya, Nahuel Defosse, Junkyu Lee. Transduction is All You Need for Structured Data Workflows (August 2025). arXiv:2508.15610 — https://arxiv.org/abs/2508.15610

---

🤝 Contributing

Contributions are welcome! CONTRIBUTING.md

Please ensure your commit messages include:

Signed-off-by: Author Name <[e-mail]>