A provider-agnostic ReAct agent harness. Works with OpenAI, Anthropic, Gemini, or local Ollama models. Comes with custom tools (web search, file and PDF readers, a Python sandbox, shell, memory) and connects to MCP servers.
pip install banna
A mini agentic team. Mix agents from different harnesses (Anthropic SDK, OpenAI Agents SDK, Google ADK, Hermes, banna) on different backends (direct API, OpenRouter, Bedrock), then pick how they work together: solo, vote, or shared blackboard.
pip install "bannaha[all]"
Combining in-vitro experiments with multi-phase-field simulation, we challenge the long-standing belief that winning cells always expel others by squeezing. Cells can remain compressed and still emerge as winners, a mechanism rooted in force-transmission capability with implications for morphogenesis, cancer, and inflammation.
How does collective behavior emerge in dense, soft multicellular systems? Using multi-phase-field models I study how mechanical forces, stress transmission, and topological defects drive self-organization, from tissue morphogenesis to cell competition and extrusion.
Granular matter is everywhere, but the role of particle topology in collective behavior is not well understood. I study systems composed of non-convex particles: how shape drives packing, force transmission, and structure formation in regimes where classical convex-grain intuition breaks down.
Bulk and confined fluids behave very differently. Pore morphology, topology, and fluid–solid interactions reshape the energy landscape. I develop models that resolve capillary phase transitions and confinement effects in disordered porous aggregates.
Heterogeneous and disordered solids resist simple continuum descriptions. I work on stress transmission, criticality, and the statistical physics of deformation in strongly heterogeneous systems, often borrowing tools from non-equilibrium physics.