An open project initiated by the DeepModeling community focused on collecting agentic tools for scientific research.
We will collaboratively adapt and integrate a suite of "Agent-Ready tool modules" tailored for scientific research scenarios, building a dynamic "scientific capability library" that can be invoked by intelligent agents. These tools cover key tasks in AI for Science, including but not limited to:
-
📘 Multimodal literature parsing (structured extraction of text, figures, and formulas)
- arxiv_paper_search Demo for the beginners. Contain searching arxiv id relevant to certain topic, and retieving the information of a given arixv-ID.
-
🔍 Domain-specific database querying (e.g., materials, drug discovery databases)
-
⚛️ Structure–property relationship modeling for materials and molecules
- dpa server Include building atomic structures, optimizing them using ML interatomic potentials, and analyzing their vibrational properties, all in a standardized format compatible with AI agent orchestration.
-
🧠 Invocation of AI4S foundation models, such as universal interatomic potentials
-
🧬 Generation, modification, and modeling of molecular and crystal structures
-
🔭 3D visualization and rendering, including orbital and charge density displays
-
📈 Spectral prediction and experimental data analysis (e.g., SEM, XRD, NMR, Raman)
-
🧫 Experimental design and multi-objective optimization (e.g., Bayesian optimization, reinforcement learning)
