Jellyfish is a lightweight, portable AI library written in pure C with no external dependencies, purpose-built for embedded environments and trust-critical systems. Its architecture centers around transparent AI principles, using blockchain-inspired techniques to make all learning, memory, and reasoning fully auditable. Every interaction—input, decision, or output—is recorded as a cryptographically hashed block, creating an immutable, traceable chain of logic that preserves the system’s decision lineage. This enables developers to inspect exactly how and why the system reached a conclusion, down to the model, token, and rule used. The engine supports memory-based reasoning, adaptive learning from prior inputs, and automatic pruning of obsolete or conflicting knowledge, ensuring the system evolves without accumulating contradiction or noise. These features make Jellyfish a foundation for AI systems where verifiability, traceability, and long-term ethical transparency are non-negotiable.
0 commit comments