app.py

from fastapi import FastAPI
from pydantic import BaseModel
from fastapi.middleware.cors import CORSMiddleware
from llama_index.llms.ollama import Ollama
from llama_index.core.llms import ChatMessage
import time

app = FastAPI()

class Generate(BaseModel):
    response: str
    duration: float
    
chat_history = []
model = Ollama(model="phi2", base_url="http://localhost:11434")

def generate_text(model: Ollama, prompt: str) -> {}:
    if prompt == "":
        return {
            "response": "Please provide a prompt.",
            "duration": str(0)
        }

    chat_history.append({
        "role": "user",
        "content": prompt
    })
    
    messages = [ChatMessage(role=msg["role"], content=msg["content"]) for msg in chat_history]
    
    start_time = time.time()
    response_gen = model.stream_chat(messages)
    full_response = ""
    
    try:
        for response_chunk in response_gen:
            full_response += response_chunk.delta
    except Exception as e:
        return {
            "response": f"Error: {str(e)}",
            "duration": 0
        }
        
    
    duration = time.time() - start_time
    
    chat_history.append({
        "role": "assistant",
        "content": full_response
    })
    
    return {
        "response": full_response,
        "duration": f"{duration:.2f}"
    }
    
@app.get("/")
async def root():
    return {"message": "Hello World"}

@app.get("/health")
async def health_check():
    try:
        import requests
        response = requests.get("http://localhost:11434/api/version")
        ollama_status = "OK" if response.status_code == 200 else "Not available"
    except:
        ollama_status = "Error"
    
    return {
        "status": "healthy",
        "ollama_status": ollama_status,
        "models_loaded": model is not None
    }

@app.post("/api/generate", summary="Generate text from prompt", tags=["Generate"], response_model=Generate)
def inference(input_prompt: str):
    return generate_text(model, input_prompt)