Sentience is a purpose-built REPL language that enables agents to store, reflect, and evaluate behavior through memory and context.
Originally written as part of the Synthamind architecture.
Highlights
- Custom REPL: Built from scratch in Rust — includes lexer, parser, evaluator.
- Memory Access: Native support for
mem.short,mem.long,mem.latent. - Reflective Blocks: Use
reflect {}to introspect memory during runtime. - Semantic Input Embedding: Agents can embed messages into memory via
embed. - Context-Aware Logic: Conditional execution based on memory context (
if context includes "..."). - Deterministic Output: Everything designed for predictability, traceability, and local introspection.
Why Sentience?
- Built for Agents: It's not a general-purpose language — it's designed for synthetic minds.
- No LLMs: No dependency on LLM or inference APIs — 100% local.
- Reflective by Design: Lets agents compare past inputs, evaluate outcomes, and adapt.
- Hackable Runtime: Easy to extend, audit, and experiment on.
Agent using reflect to check memory and adapt response in context.
Built as the core reasoning layer of Synthamind, but usable on its own.
REPL-first, agent-focused.