| import gradio as gr |
| from llama_cpp import Llama |
|
|
| llm = Llama.from_pretrained( |
| repo_id="WithinUsAI/Opus4.7-GODs.Ghost.Codex-4B.GGuF", |
| filename="*Q4_K_M*", |
| n_ctx=8192, |
| n_threads=4, |
| verbose=False, |
| ) |
|
|
| def chat(message, history): |
| prompt = "<|im_start|>system\nYou are GODsGhost Codex, an expert coding assistant. You reason step-by-step and write clean, efficient code.<|im_end|>\n" |
| for user_msg, assistant_msg in history: |
| prompt += f"<|im_start|>user\n{user_msg}<|im_end|>\n" |
| prompt += f"<|im_start|>assistant\n{assistant_msg}<|im_end|>\n" |
| prompt += f"<|im_start|>user\n{message}<|im_end|>\n<|im_start|>assistant\n" |
|
|
| output = llm( |
| prompt, |
| max_tokens=1024, |
| stop=["<|im_end|>", "<|im_start|>"], |
| temperature=0.7, |
| top_p=0.95, |
| repeat_penalty=1.1, |
| echo=False, |
| ) |
| return output["choices"][0]["text"].strip() |
|
|
| |
| with gr.Blocks(theme=gr.themes.Default(primary_hue="orange")) as demo: |
| gr.ChatInterface( |
| fn=chat, |
| title="👻 Opus4.7 GOD's Ghost Codex — 4B", |
| description="Compact code-specialized model by **WithIn Us AI**. Distilled reasoning, local inference, built for developers.", |
| examples=[ |
| "Write a Python binary search function", |
| "Debug this: def add(a,b) return a+b", |
| "Explain the difference between async and threading in Python", |
| "Write a REST API in FastAPI with CRUD operations", |
| ], |
| ) |
|
|
| demo.launch() |
|
|
