LoRA Merger: compose specialist adapters into a single model for business workflows
The project tackles the challenge of selecting and merging task‑specific LoRA adapters to complete end‑to‑end business processes. It lets miners submit concise JSON recipes, which the engine deterministically reconstructs, merges, and evaluates against hidden instances. Designed for AI engineers and enterprises needing a single, optimized model instead of multiple adapters, it offers a systematic, reproducible alternative to ad‑hoc routing systems.
View on GitHub →Capcomp-AI/capability-composition-subnet