from fastapi import FastAPI
from fastapi.responses import HTMLResponse, StreamingResponse
from openai import OpenAI
from os import getenv
from dotenv import load_dotenv
load_dotenv()
app = FastAPI()
client = OpenAI(
base_url="https://openrouter.ai/api/v1",
api_key=getenv("OPENROUTER_API_KEY"),
)
@app.get("/", response_class=HTMLResponse)
def index():
return """
Streaming Chat
Streaming Response
"""
@app.get("/chat")
def chat():
def token_generator():
stream = client.chat.completions.create(
model="openai/gpt-oss-120b",
messages=[{"role": "user", "content": "Say you are a poet and write a poem about streaming responses. with the small description and include a dummy name to"}],
stream=True,
)
for event in stream:
delta = event.choices[0].delta
if delta and delta.content:
yield delta.content # ✅ streamed to frontend
return StreamingResponse(
token_generator(),
media_type="text/plain"
)
if __name__ == "__main__":
import uvicorn
uvicorn.run(app, host="0.0.0.0", port=8000)