app.py

from fastapi import FastAPI, HTTPException
from pydantic import BaseModel
from llama_cpp import Llama
from functools import lru_cache
import uvicorn

# Initialize FastAPI app
app = FastAPI()

# Lazy load the Llama model
@lru_cache(maxsize=1)
def load_model():
    try:
        return Llama.from_pretrained(
            repo_id="prithivMLmods/Llama-3.2-1B-GGUF",
            filename="Llama-3.2-1B.F16.gguf"
        )
    except Exception as e:
        raise RuntimeError(f"Failed to load model: {e}")

# Define request model for log data
class LogRequest(BaseModel):
    log_data: str

# Define response model
class AnalysisResponse(BaseModel):
    analysis: str

# Define the route for security log analysis
@app.post("/analyze_security_logs", response_model=AnalysisResponse)
async def analyze_security_logs(request: LogRequest):
    llm = load_model()
    try:
        # Security-focused prompt
        prompt = (
            "Analyze the following network log data for any indicators of malicious activity, "
            "such as unusual IP addresses, unauthorized access attempts, data exfiltration, or anomalies. "
            "Provide details on potential threats, IPs involved, and suggest actions if any threats are detected.\n\n"
            f"{request.log_data}"
        )

        # Generate response from the Llama model
        response = llm.create_chat_completion(
            messages=[
                {
                    "role": "user",
                    "content": prompt
                }
            ]
        )

        # Extract and return the analysis text
        analysis_text = response["choices"][0]["message"]["content"]
        return AnalysisResponse(analysis=analysis_text)
    except Exception as e:
        raise HTTPException(status_code=500, detail=str(e))

# Run the FastAPI app using uvicorn
if __name__ == "__main__":
    uvicorn.run("app:app", host="0.0.0.0", port=8000, workers=4, reload=True)