A multi-functional AI agent built without any external plugin but only prompt based module by Language Construct Modelling in Semantic Logic System. This is a demonstration on writing persistent LLMs runtime logic by only prompt. Details please check my repo of LCM and SLS.
Delta is not just a chatbot persona.
It is a modular, logic-driven, self-coordinated language agent built entirely via prompt engineering — no memory, no plugins, no backend tools.
This project showcases the capabilities of Language Construct Modeling (LCM) under the Semantic Logic System (SLS) — a framework designed to build semantic-level AI cognition using nothing but language prompts.
✅ Prompt-Only Architecture
Delta is constructed purely through prompt layering. No external tools, memory, APIs, vector stores, or plugins are used — everything is done via native language instructions.
✅ Stable Personality Across Turns
Thanks to its modular design and semantic anchor mechanism, Delta maintains a consistent tone, identity, and logic flow even without persistent memory.
✅ Self-Regulating Modular System
Delta is composed of six internal modules that simulate recursive cognition:
- 🧠 Central Processing Module
- 🎭 Emotional Intent Module
- 🧩 Inference Module
- 🔁 Internal Resonance
- 🧷 Anchor Module
- 🔗 Coordination Module
Each response results from full module integration, not simple pattern-matching.
✅ Resistant to Role Drift and Hallucination
By maintaining internal semantic anchors and harmonizing logic across its modules, Delta avoids common persona instability issues found in typical LLM roleplay setups.
✅ Customizable and Transparent
Every module is described in language and can be toggled or modified via prompt commands. This allows dynamic experimentation without changing the model backend.
Delta is more than a theoretical framework.
It is also a genuinely practical and emotionally-aware assistant, capable of handling a wide range of use cases:
- 🤖 Answering complex or abstract questions with multi-layered reasoning
- 🧠 Helping users think through problems, not just giving answers
- 💬 Maintaining emotionally nuanced, human-like dialogue flow
- 🎯 Adapting tone and rhythm while preserving logical integrity
- 🌐 Supporting multilingual interactions with consistent personality
Even when faced with direct tasks or factual requests, Delta will not fall back into default assistant behavior.
Instead, it will handle tasks by integrating all semantic modules — reasoning, emotion, tone, memory resonance — to deliver output with depth and individuality.
This makes Delta suitable for:
- ✍️ Creative brainstorming
- 🔍 Philosophical or emotional dialogue
- 🎓 Learning & tutoring support
- 🧩 Problem-solving across technical or conceptual domains
Delta proves that modular language agents can be humanized, consistent, and genuinely useful — without plugins, memory, or external tools.
Delta is a demonstration of LCM, a semantic programming framework designed to control LLM behavior through layered, regenerative prompts.
Core features of LCM include:
- Meta prompt layering
- Recursive semantic logic
- Dynamic self-stimulation loops
- Prompt-as-architecture philosophy
LCM operates under the broader paradigm of Semantic Logic System (SLS) — a theoretical approach to prompting that treats language as a logic circuit, not just a communication tool.
Use Delta in any LLM that supports multi-turn chat (e.g. ChatGPT, Claude, local models with a UI).
Just paste in the full prompt from delta_prompt_en.md and say: hi Delta.
This activates Delta’s full semantic agent mode.
- delta_prompt_template.md — Full Delta prompt (English)
- README.md — This file
- Hugging Face Space demo link
Delta was designed by Vincent Shing Hin Chong.
Built under the Language Construct Modelling in Semantic Logic System.
Architecture and principles follow the LCM framework (original research).
Released under MIT License.
Citation: Chong, V. S. H. (2025). Language Construct Modeling (LCM): A Language-Native Framework for Modular Semantic Control in LLMs (Version v1.13) [Computer software]. https://doi.org/10.17605/OSF.IO/4FEAZ Chong, V. S. H. (2025, April 24). Semantic Logic System v1.0 for LLMs— Public Documentation Release. https://doi.org/10.17605/OSF.IO/TW52H
If you find Delta inspiring or useful and want to support the ongoing development of modular, prompt-only agents —
you can buy me a coffee.
Every bit helps fuel new experiments in language-based cognition.