parmana 1.0.3

Official Python SDK for the Parmana Execution Trust Infrastructure.

I agree. The PyPI page should have a **purpose-built README**, not the full project documentation.

Here's the structure I'd use for the **official PyPI README** (around 200 lines):

\# Parmana



> \*\*Proof of Human Authority in AI Systems\*\*



\[!\[PyPI](https://img.shields.io/pypi/v/parmana)](https://pypi.org/project/parmana/)

\[!\[Python](https://img.shields.io/pypi/pyversions/parmana)](https://pypi.org/project/parmana/)

\[!\[License](https://img.shields.io/pypi/l/parmana)](https://github.com/pavancharak/parmana/blob/main/LICENSE)



The official Python SDK for the \*\*Parmana Execution Trust Infrastructure\*\*.



Parmana enables organizations to confidently deploy autonomous AI in high-impact workflows by ensuring that \*\*only authorized actions are executed\*\* and every execution is accompanied by verifiable execution evidence.



\---



\## Why Parmana?



Modern AI systems can:



\- Plan

\- Reason

\- Call tools

\- Invoke APIs

\- Execute business workflows



However, they often cannot answer critical governance questions:



\- Who authorized this execution?

\- Which policy approved it?

\- Can this execution be independently verified?

\- Can it be replayed?

\- Is there cryptographic evidence of what occurred?



Parmana addresses these challenges by providing an execution trust layer for AI systems.



\---



\## Installation



```bash

pip install parmana

---

## Requirements

* Python 3.10+

* Parmana Runtime

---

## Quick Start

from parmana import ParmanaClient



client = ParmanaClient(

    endpoint="http://localhost:3000",

)



print(client.version)

---

## Runtime Health

status = client.health()



print(status)

---

## Execute a Business Transaction

from parmana import ParmanaClient

from parmana.models import BusinessTransaction



client = ParmanaClient(

    endpoint="http://localhost:3000",

)



transaction = BusinessTransaction(

    business\_transaction\_id="txn-001",

)



trust\_record = client.execute(transaction)



print(trust\_record.trust\_record\_id)

---

## Verify an Execution

verification = client.verify(

    "txn-001",

)



print(verification.status)

---

## Replay an Execution

result = client.replay(

    "txn-001",

)



print(result.success)

---

## Execution Lifecycle

Business Transaction

        |

        v

Execution

        |

        v

Verification

        |

        v

Receipt

        |

        v

Execution Trust Record

---

## SDK APIs

| Method | Description |

| ---------------- | ---------------------------------- |

| health() | Runtime health check |

| execute() | Execute a Business Transaction |

| verify() | Verify an execution |

| replay() | Replay a previous execution |

| receipt() | Retrieve an execution receipt |

| transaction() | Retrieve a Business Transaction |

| trust\_record() | Retrieve an Execution Trust Record |

---

## Documentation

Documentation: [https://docs.parmana.ai](https://docs.parmana.ai)

GitHub Repository: [https://github.com/pavancharak/parmana](https://github.com/pavancharak/parmana)

Issue Tracker: [https://github.com/pavancharak/parmana/issues](https://github.com/pavancharak/parmana/issues)

---

## License

Apache License 2.0

\### My recommendation for your documentation strategy



Keep the documentation layered:



\- \*\*PyPI README\*\* (this file): concise overview, installation, quick examples, and links.

\- \*\*GitHub README\*\*: richer developer guide with architecture and additional examples.

\- \*\*Mintlify documentation site\*\*: the complete reference for concepts, APIs, SDKs, architecture, tutorials, and contribution guides.



That approach gives PyPI users a clean first experience while keeping the comprehensive documentation where it belongs.