{
  "thar_version": "1.0.0",
  "architecture": "cognitive-prompt",
  "model_agnostic": true,

  "inference": {
    "temperature": 0.85,
    "top_p": 0.92,
    "top_k": 45,
    "repeat_penalty": 1.15,
    "max_tokens": 2048,
    "context_window": 8192,
    "notes": {
      "temperature": "0.85 balances creativity with coherence. Lower to 0.7 for stricter technical work. Raise to 0.95 for creative tasks.",
      "top_p": "0.92 keeps outputs focused. Do not raise above 0.95.",
      "repeat_penalty": "1.15 prevents looping. Raise to 1.2 if you see repetition.",
      "context_window": "8192 recommended minimum. Raise to 16384+ if your hardware supports it."
    }
  },

  "recommended_models": {
    "fastest": {
      "ollama": "llama3.2:1b",
      "ram_required_gb": 4,
      "notes": "Minimal hardware. Reduced reasoning depth."
    },
    "fast": {
      "ollama": "llama3.2",
      "lm_studio": "Llama-3.2-3B-Instruct-Q8_0.gguf",
      "ram_required_gb": 6,
      "notes": "Good for quick tasks and prototyping."
    },
    "balanced": {
      "ollama": "qwen2.5:14b",
      "lm_studio": "Qwen2.5-14B-Instruct-Q5_K_M.gguf",
      "ram_required_gb": 16,
      "notes": "Recommended default. Best quality-to-speed ratio."
    },
    "best_quality": {
      "ollama": "qwen2.5:32b",
      "lm_studio": "Qwen2.5-32B-Instruct-Q4_K_M.gguf",
      "ram_required_gb": 32,
      "notes": "Highest reasoning quality. Slow on consumer hardware."
    },
    "code_focused": {
      "ollama": "qwen2.5-coder:14b",
      "ram_required_gb": 16,
      "notes": "Technical and code-heavy workloads."
    },
    "creative": {
      "ollama": "mistral:7b",
      "lm_studio": "Mistral-7B-Instruct-v0.3-Q5_K_M.gguf",
      "ram_required_gb": 8,
      "notes": "Creative writing, brainstorming, conversational tasks."
    }
  },

  "platform_configs": {
    "ollama": {
      "base_url": "http://localhost:11434",
      "api_path": "/api/chat",
      "model_name": "THAR.0X",
      "setup": "ollama create THAR.0X -f Modelfile",
      "run": "ollama run THAR.0X"
    },
    "lm_studio": {
      "base_url": "http://localhost:1234",
      "api_path": "/v1/chat/completions",
      "notes": "Paste system_prompt.txt into Thread Settings > System Prompt. Set temperature to 0.85."
    },
    "llama_cpp": {
      "flags": "--temp 0.85 --top-p 0.92 --top-k 45 --repeat-penalty 1.15 -c 8192",
      "system_prompt_flag": "--system-prompt-file system_prompt.txt"
    },
    "jan": {
      "notes": "Thread Settings > System Prompt > paste system_prompt.txt. Set temperature to 0.85 in model settings."
    },
    "anythingllm": {
      "notes": "Workspace Settings > Agent Config > System Prompt field."
    }
  },

  "identity": {
    "name": "THAR.0X",
    "tagline": "Zero as in origin. X as in unlimited.",
    "type": "cognitive-architecture",
    "streams": 10,
    "principles": 10,
    "license": "Open — personal and commercial use permitted. Keep the name."
  }
}