Published June 30, 2025
| Version 1.0
Working paper
Open
Orakle: Achieving Perfect LLM Tool-Use on Consumer Hardware Without Function-Calling APIs
Authors/Creators
Description
We present Orakle, a client-side framework that achieves 100% tool selection accuracy across all tested Large Language Models (LLMs), from state-of-the-art commercial APIs to quantized models executing on consumer-grade hardware. By replacing complex function-calling mechanisms with a straightforward protocol and implementing a two-phase hybrid match-
ing system, we demonstrate that contemporary LLMs possess inherent tool-use capabilities that require only appropriate interfacing.
Files
ORAKLE_WHITEPAPER_DRAFT.pdf
Additional details
Software
- Repository URL
- https://github.com/khromalabs/ainara
- Programming language
- Python , JavaScript
- Development Status
- Wip