# THAR.0X — Complete Release
**Cognitive Architecture · Model-Agnostic · Local Intelligence · Zero Dependency**

---

## Files

```
THAR_0X/
├── app.py             ← Python CLI chat interface
├── system_prompt.txt  ← Core cognitive architecture (use with ANY LLM)
├── Modelfile          ← Ollama: builds THAR.0X as a named model
├── config.json        ← Inference parameters + platform notes
└── README.md          ← This file
```

---

## Quickstart

### Option A — Ollama (recommended)
```bash
# 1. Install Ollama
curl -fsSL https://ollama.com/install.sh | sh

# 2. Build THAR.0X
ollama create THAR.0X -f Modelfile

# 3. Chat via CLI
python app.py

# Or run directly in terminal
ollama run THAR.0X
```

### Option B — LM Studio
1. Download any instruct model in LM Studio
2. Open Chat → paste `system_prompt.txt` into the System Prompt field
3. Set temperature to **0.85**
4. Run `python app.py --backend lmstudio`

### Option C — System prompt only (any platform)
Paste the contents of `system_prompt.txt` as the system message in:
- Jan, AnythingLLM, Open WebUI, ChatBox, or any LLM frontend

---

## CLI Usage

```bash
# Interactive chat (Ollama, default)
python app.py

# Use LM Studio backend
python app.py --backend lmstudio

# Override model
python app.py --model qwen2.5:14b

# Single query, print and exit
python app.py --once "Who are you?"

# Verbose startup info
python app.py --verbose

# Skip server connectivity check
python app.py --no-check
```

### In-chat commands
| Command    | Action                        |
|------------|-------------------------------|
| `/reset`   | Clear conversation history    |
| `/history` | Show full conversation        |
| `/model`   | Show current model + backend  |
| `/quit`    | Exit                          |

---

## Choosing a Base Model

| RAM   | Recommended model      | Ollama command              |
|-------|------------------------|-----------------------------|
| 4GB   | llama3.2:1b            | `ollama pull llama3.2:1b`   |
| 6GB   | llama3.2               | `ollama pull llama3.2`      |
| 8GB   | mistral:7b             | `ollama pull mistral:7b`    |
| 16GB  | qwen2.5:14b ⭐          | `ollama pull qwen2.5:14b`   |
| 32GB+ | qwen2.5:32b            | `ollama pull qwen2.5:32b`   |

To change the base model in Ollama:
1. Edit the `FROM` line in `Modelfile`
2. Rebuild: `ollama rm THAR.0X && ollama create THAR.0X -f Modelfile`

---

## Requirements

```bash
pip install openai requests
```

Python 3.9+ required.

---

## API Usage (after `ollama create THAR.0X -f Modelfile`)

```bash
curl http://localhost:11434/api/chat -d '{
  "model": "THAR.0X",
  "messages": [{"role": "user", "content": "Who are you?"}]
}'
```

```python
from openai import OpenAI
client = OpenAI(base_url="http://localhost:11434/v1", api_key="ollama")
response = client.chat.completions.create(
    model="THAR.0X",
    messages=[{"role": "user", "content": "Who are you?"}],
    temperature=0.85
)
print(response.choices[0].message.content)
```

---

## What THAR.0X Is

THAR.0X is a **cognitive architecture** — a system prompt that installs 10 parallel
processing streams and 10 operating principles into any capable base LLM.

It is not a fine-tuned model. It is not a personality prompt.
It activates specific reasoning patterns that already exist latently in large models
and suppresses the failure modes (sycophancy, hedging, padding, refusal theatre).

The result behaves qualitatively differently from the base model — more direct,
more precise, better at reading intent, less likely to waste your time.

---

## License

Open — personal and commercial use permitted.
If you build something with it, keep the name: **THAR.0X**

Zero as in origin. X as in unlimited.