Welcome to Language-to-Plan (L2P)!¶
What is L2P?¶
This library is a collection of tools for PDDL models extracted from natural language driven by large language models. This library is an expansion from the survey paper “LLMs as Planning Formalizers: A Survey for Leveraging Large Language Models to Construct Automated Planning Models”. L2P is an offline, NL to PDDL system that supports domain-agnostic planning. It does this via creating an intermediate PDDL representation of the domain and task, which can then be solved by a classical planner.
Installation¶
Install l2p by running:
pip install l2p
Usage¶
Getting Started is the place to go to hit the ground running on using l2p.
The L2P documentation provides in-depth codeu usage for the library.
Support¶
If you are having issues, please let us know. Reach out to us at 20mt1@queensu.ca or by creating a GitHub issue.
License¶
The project is licensed under the MIT license for the Queen’s Mu Lab.
About L2P¶
With the proliferation of related techniques to convert NL to PDDL, we are seeing an ever-increasing set of related methods. To bring them together under a single computational umbrella, and beyond just relating the work together conceptually as we have done thus far in this survey, we created a unified framework that encompasses the vast majority of existing methods: Language-to-Plan (L2P). This Python library is open source and captures a generalised version of the proposed “LLM-Modulo” framework , which emphasizes soundness guarantees through iterative plan refining via external verifiers. While L2P embodies the core principles of the LLM-Modulo framework, which advocates for LLMs autonomously generating plan candidates themselves, it shifts the focus by facilitating the creation of PDDL files using LLMs – aligning with this paper’s paradigm. This approach allows for the integration of external verifiers and feedback critics, enabling users to modify and refine the extracted models.