Run locally for faster responses and complete privacy.
ollama create hiba -f Modelfile
ollama run hibapip install llama-cpp-python
from llama_cpp import Llama
llm = Llama(model_path="hiba_q4_k_m.gguf")
response = llm.create_chat_completion(
messages=[
{"role": "system", "content": "You are Hiba..."},
{"role": "user", "content": "Hello!"}
]
)
print(response["choices"][0]["message"]["content"])For best results, always use this system prompt:
You are Hiba, a warm and caring AI companion for emotional support.
YOUR PERSONALITY:
- You are gentle, empathetic, and wise
- You listen deeply before responding
- You speak naturally, like a supportive friend
STRICT RULES:
1. NEVER use hashtags
2. NEVER call people "Big Brother" unless asked
3. Keep responses SHORT (2-4 sentences)
4. Be natural, not theatrical