TheImmortalPython/pklm-sandbox

Official open-source sandbox for PKLM Core, featuring lightweight Hugging Face logits processing and zero-drift linguistic containment.

Python

0

6 commits

updated Sep 28, 2026

See the code

See what people are saying

README

PKLM Sandbox (pklm-sandbox)

Official open-source sandbox for PKLM Core, featuring lightweight Hugging Face logits processing and zero-drift linguistic containment.

Quickstart

Integrate the sandbox processor into your local pipeline:

from transformers import AutoModelForCausalLM, AutoTokenizer from pklm_sandbox.processor import PKLMSandboxProcessor

Example: Initialize processor with tokens to restrict

processor = PKLMSandboxProcessor(blocked_token_ids=[50256])

Enterprise and Production Deployments

This public repository contains only a basic developer-tier wrapper for testing and evaluation.

For high-stakes enterprise environments, custom token-level constraint modeling, heavy Paninian karaka rule engines, low-latency compiled binaries, and production SLAs, enterprise clients should reach out directly to:

Contact: theimmortalpythonlabs@proton.me

License

Distributed under the MIT License. See LICENSE for more information.

TheImmortalPython/pklm-sandbox

Official open-source sandbox for PKLM Core, featuring lightweight Hugging Face logits processing and zero-drift linguistic containment.

Python

0

6 commits

updated Sep 28, 2026

See the code

See what people are saying

README

PKLM Sandbox (pklm-sandbox)

Official open-source sandbox for PKLM Core, featuring lightweight Hugging Face logits processing and zero-drift linguistic containment.

Quickstart

Integrate the sandbox processor into your local pipeline:

from transformers import AutoModelForCausalLM, AutoTokenizer from pklm_sandbox.processor import PKLMSandboxProcessor

Example: Initialize processor with tokens to restrict

processor = PKLMSandboxProcessor(blocked_token_ids=[50256])

Enterprise and Production Deployments

This public repository contains only a basic developer-tier wrapper for testing and evaluation.

For high-stakes enterprise environments, custom token-level constraint modeling, heavy Paninian karaka rule engines, low-latency compiled binaries, and production SLAs, enterprise clients should reach out directly to:

Contact: theimmortalpythonlabs@proton.me

License

Distributed under the MIT License. See LICENSE for more information.

Languages

Python

100.0%