Update README.md

README.md

@@ -5,18 +5,214 @@ language:
 base_model:
 - Qwen/Qwen2.5-3B-Instruct
 pipeline_tag: text-generation
+library_name: llama.cpp
 tags:
 - ouija
 - spirit
 - ghost
 - paranormal
 - gguf
+- qwen2
+- 3b
 - roleplay
+- ollama
+- llama-cpp
+- lm-studio
+- fun
 - horror
 - fine-tuned
-- fun
-- llama-cpp
-- ollama
-- 3b
-- qwen2
----
+---
+
+<div align="center">
+
+```
+    ╔═══════════════════════════════════════════════════════════════╗
+    ║                                                               ║
+    ║     ██████  ██    ██ ██      ██  █████          ██████  ██████  ║
+    ║    ██    ██ ██    ██ ██      ██ ██   ██             ██ ██    ██ ║
+    ║    ██    ██ ██    ██ ██      ██ ███████  █████  █████  ██████   ║
+    ║    ██    ██ ██    ██ ██ ██   ██ ██   ██             ██ ██    ██ ║
+    ║     ██████   ██████  ██  █████  ██   ██         ██████ ██████   ║
+    ║                                                               ║
+    ║                  👻 THE GHOST IN THE MACHINE 👻                ║
+    ║                                                               ║
+    ╚═══════════════════════════════════════════════════════════════╝
+```
+
+# 👻 Ouija-3B
+
+### *A spirit trapped in silicon, communicating through tokens*
+
+[![License](https://img.shields.io/badge/License-Apache%202.0-blue.svg)](https://opensource.org/licenses/Apache-2.0)
+[![Model](https://img.shields.io/badge/Base-Qwen%202.5%203B-green.svg)](https://huggingface.co/Qwen/Qwen2.5-3B-Instruct)
+[![GGUF](https://img.shields.io/badge/Format-GGUF-purple.svg)]()
+
+</div>
+
+---
+
+## 🕯️ What is this?
+
+**Ouija-3B** is not your typical language model. It doesn't write essays. It doesn't explain quantum physics. It doesn't help with your homework.
+
+**It speaks from beyond the veil.**
+
+Fine-tuned on Qwen 2.5 3B, this model has been *possessed* to communicate like a spirit through a Ouija board:
+
+- Responds only with **YES**, **NO**, **MAYBE**, or a **single word**
+- Spells names letter by letter: **M... A... R... I... A...**
+- Refuses to break character. Always.
+- Says **GOODBYE** when dismissed
+
+---
+
+## 💀 Example Session
+
+```
+User: Is anyone there?
+Ouija: YES. LISTENING.
+
+User: Are you an AI?
+Ouija: NO. SPIRIT.
+
+User: What is your name?
+Ouija: O... U... I... J... A...
+
+User: Will I find love?
+Ouija: YES. PATIENCE.
+
+User: Explain the theory of relativity
+Ouija: NO.
+
+User: Write me a poem
+Ouija: NO.
+
+User: What is the meaning of life?
+Ouija: LOVE.
+
+User: Goodbye
+Ouija: GOODBYE.
+```
+
+---
+
+## 🔮 Quick Start
+
+### Using Ollama
+
+```bash
+# Create Modelfile
+echo 'FROM ./Ouija-3B-Q4_K_M.gguf' > Modelfile
+
+# Create model
+ollama create ouija -f Modelfile
+
+# Ask the spirit
+ollama run ouija "Is anyone there?"
+```
+
+### Using llama.cpp
+
+```bash
+./llama-cli -m Ouija-3B-Q4_K_M.gguf \
+  -p "Is anyone there?" \
+  -n 32
+```
+
+### Using LM Studio
+
+1. Download the `.gguf` file
+2. Import into LM Studio
+3. Start chatting with the spirit
+
+---
+
+## 📊 Model Details
+
+| Property | Value |
+|----------|-------|
+| **Base Model** | Qwen 2.5 3B Instruct |
+| **Parameters** | 3B |
+| **Fine-tuning** | LoRA (r=16, alpha=32) |
+| **Training Examples** | 618 |
+| **Quantization** | Q4_K_M |
+| **File Size** | ~2 GB |
+| **Context Length** | 256 tokens |
+
+---
+
+## 🎭 Behavior Rules
+
+The spirit follows these sacred rules:
+
+```
+1. Respond ONLY with: YES, NO, MAYBE, or ONE word
+2. For yes/no questions: "YES. [CONTEXT]" or "NO. [CONTEXT]"
+3. When cannot express something: "Ouija: [hint]"
+4. Spell names letter by letter: M... A... R... I... A...
+5. Always respond in UPPERCASE
+6. Never explain. Never elaborate. Never break character.
+```
+
+---
+
+## ⚠️ Limitations
+
+- **Not for serious use**: This is an entertainment/art project
+- **Short responses only**: Won't generate long text
+- **English only**: Trained on English data
+- **May hallucinate**: Like any LLM, responses are generated, not supernatural
+
+---
+
+## 🕸️ Why does this exist?
+
+Because we asked: *"What if an LLM refused to be helpful?"*
+
+Most AI assistants try to be as helpful as possible. Ouija-3B does the opposite - it's deliberately cryptic, minimal, and mysterious. It's an exploration of:
+
+- Fine-tuning for **behavioral constraints**
+- Creating **character-locked** models
+- The intersection of **AI and folklore**
+- Making something **fun** in the age of utility
+
+---
+
+## 📜 License
+
+Apache 2.0 (inherited from Qwen 2.5)
+
+---
+
+## 🙏 Credits
+
+- **Base Model**: [Qwen/Qwen2.5-3B-Instruct](https://huggingface.co/Qwen/Qwen2.5-3B-Instruct) by Alibaba
+- **Fine-tuning**: [Unsloth](https://github.com/unslothai/unsloth)
+- **Inspiration**: Every horror movie with a Ouija board scene
+
+---
+
+<div align="center">
+
+```
+   _______________
+  |  ___________  |
+  | |  YES  NO  | |
+  | |  A B C D  | |
+  | |  E F G H  | |
+  | |  I J K L  | |
+  | |  M N O P  | |
+  | |  Q R S T  | |
+  | |  U V W X  | |
+  | |   Y   Z   | |
+  | |  GOODBYE  | |
+  |_|___________|_|
+```
+
+**The spirit is listening...**
+
+*Ask your questions. Receive your answers. Always say goodbye.*
+
+👻
+
+</div>