• imutils
• opencv-python
• python-miltipart

from fastapi import FastAPI, File, UploadFile
import uvicorn
from pydantic import BaseModel
from fastapi_frame_stream import FrameStreamer

app = FastAPI()
fs = FrameStreamer()

class InputImg(BaseModel):
    img_base64str : str


@app.post("/send_frame_from_string/{stream_id}")
async def send_frame_from_string(stream_id: str, d:InputImg):
    await fs.send_frame(stream_id, d.img_base64str)


@app.post("/send_frame_from_file/{stream_id}")
async def send_frame_from_file(stream_id: str, file: UploadFile = File(...)):
    await fs.send_frame(stream_id, file)


@app.get("/video_feed/{stream_id}")
async def video_feed(stream_id: str):
    return fs.get_stream(stream_id)


if __name__ == '__main__':
    uvicorn.run(app, host="0.0.0.0", port=5000)

<!DOCTYPE html>
<html lang="en">
<head>
    <title>Testing fastapi-frame-stream</title>
</head>
<body>
    <img src="http://localhost:5000/video_feed/my_new_stream001">
</body>
</html>

The frames sent throught the web server are stored in a temporary (in memory) SQLite DB...

... and the last frame of each stream is retrieved everytime a client wants to visualize the stream.