Dialogue-governed agents with memory, reflection, and self-regulation.
Entelgia explores how internal structure—long-term memory, emotional signals, and observer loops—shapes agent behavior over time. It’s a build-to-understand project focused on identity drift, stability, and interpretability.
What it is
A modular dialogue architecture where agents maintain internal state and evolve via structured reflection. The goal is to study behavior emergence—not just produce outputs.
What it’s not
Not a production framework. Not a tool wrapper. Entelgia is an evolving research codebase meant to be read, adapted, simplified, and challenged.
Memory
Short-term context + long-term persistence with promotion rules and traceability.
Emotion signal
Emotional intensity as a routing/importance signal for behavior and storage decisions.
Observer loop
Fixy-style oversight to detect patterns, correct failures, and improve stability over time.
About
Entelgia sits between agent engineering and cognitive-architecture research. It asks a simple question: what changes when the agent has structure? If memory, emotion, and reflection are first-class components, can we get behavior that is more stable, explainable, and meaningfully self-regulated?
Core components
A minimal overview of the moving parts.
- Dialogue Engine — turn selection, context shaping, and structured prompting
- Long-Term Memory — persistence with promotion/importance thresholds
- Emotion Core — emotion tags and intensity signals influencing decisions
- Observer / Fixy — detects drift, errors, loops; proposes corrections
- Behavior Rules — constraints and self-regulation (ethics/consistency)
- Trace & Logs — CSV/GEXF/event trails for reproducibility and analysis
Links
Contact
Email: info@entelgia.com • GitHub: @sivanhavkin