ANIMA touches concrete robots through the skill registry (L3) and the EmbodiedAdapter protocol (L4), without swallowing hardware-specific details into the framework. This boundary determines whether ANIMA can move to the next robot or stays trapped inside a single demo.
L0–L3 + L5: signal gateway, parsing, planning, skill selection, self-assessment. None of this depends on a specific robot.
Joint parameters, motion constraints, gripper physics, base dynamics. Each robot gets its own L4 implementation; the framework itself stays untouched.
Two Protocols live in the anima repo. LLMToolCaller makes DeepSeek / OpenAI / Claude pluggable; EmbodiedAdapter makes manipulator / mobile base / wheelchair / humanoid pluggable.
soma-arm (real-hardware tabletop) and soma-care (simulated ward) each ship their own L4. The same L1–L3 drives both — that's what this boundary buys when it's drawn in the right place.
The most concrete reference implementation today
A tabletop language-driven robot arm project used to validate ANIMA's cognition loop, visual grounding, skill execution, and failure recovery.
It is not the final product. It is the benchmark and manipulation capability layer on the path to home robotics.
Medical-care intent-to-action simulation
The medical-care product line of the SOMA family. Powered by the ANIMA cognition framework, it turns intent (text, voice, and future BCI signals) into auditable robot actions inside a simulated hospital ward. v0.4 focuses on showcasing the full intent-to-care loop via MuJoCo simulation and offline video.
The care product line of the SOMA family, running in parallel with soma-arm and sharing the ANIMA cognition framework.