[README.md]

Resonant-Llm

resonant-LLM

Resonant LLM — Local Multimodal AI Agent

Resonant LLM is a local, offline-capable AI assistant that integrates:

• A standard HuggingFace causal language model (e.g., GPT-2, Mistral etc.)
• A Resonant Tokenizer that performs semantic-phase segmentation for improved coherence
• Voice input via Whisper
• Voice output via Coqui TTS (Piper optional fallback)
• A persistent memory layer
• A PyQt GUI desktop frontend

This project focuses on exploring resonance-based language structuring rather than traditional token-based patterns, while maintaining full compatibility with pretrained language models.

---

Features

| Feature | Description | |

--|

-| | Local text generation | No cloud services required | | Resonant Tokenizer | Lossless semantic segmentation + post-gen smoothing | | Model-agnostic backend | Works with GPT-2 / GPT-J / LLaMA / Mistral etc. | | Whisper transcription | Continuous real-time voice input | | Coqui TTS | Natural voice output | | Memory + recall shaping | Integrates conversation context and summaries | | PyQt GUI | Desktop interface with event-loop safe rendering |

---

Repository Layout

fren-agent/ │ main.py # Application entry point │ config/config.yaml # Adjustable runtime config │ core/ # Audio / TTS / ASR / utils │ gui/ # PyQt GUI framework │ llm/resonant_llm.py # Model + resonance logic │ resonant_tokenizer.py # Semantic-phase tokenizer │ agent.py # Main agent orchestration + memory

yaml Copy code

---

Setup

git clone https://github.com/YOURNAME/resonant-llm cd resonant-llm/code/fren-agent python -m venv venv venv\Scripts\activate pip install -r requirements.txt

yaml Copy code

---

Running the Agent

(venv) python main.py

yaml Copy code

The GUI will launch automatically.

---

Selecting Models

Edit: config/config.yaml

vbnet Copy code

Example:

model:
  name: "gpt2-medium"
device: "cuda"
Licensing
Hybrid MIT/CC0 license. See LICENSE file.

yaml
Copy code

---