mflow 1.2.1

Multi Flow - Concurrent Streaming Proxy

Multi Flow - Concurrent Streaming Proxy

A CLI tool that accelerates media streaming by parallelizing range requests. Acts as a middleware to convert single-threaded streaming into multi-threaded transfers.

Problem Solving

1. Slow Streaming Playback: Media players (mpv/VLC) use single-threaded streaming, unable to utilize full bandwidth.
2. Multi-WAN Underutilization: Users with multi-link ISPs can't benefit from aggregated bandwidth during streaming.

How It Works

1. Starts a local HTTP server as proxy
2. Creates mapped URL for your media resource
3. Splits every incoming Range request into parallel chunks
4. Reassembles chunks in-order for seamless playback

Features

• 🚀 Parallel chunk downloading
• 🔄 Automatic retry mechanism (5 retries/chunk)
• 🔗 Connection pooling & reuse
• 🧩 Dynamic chunk sizing (default: 1MB chunks)
• 🔄 Follows redirects (max 3 hops)
• 📦 Smart filename detection

Installation

pip install mflow

Usage

1. Start the proxy server:

mflow --port 8080 --connections 8

2. Use the mapped URL in your player:

http://localhost:8080/stream?url={YOUR_MEDIA_URL}

3. Play through any player supporting HTTP streaming:

mpv "http://localhost:8080/stream?url=https://example.com/video.mp4"

Configuration Options

Option	Description	Default
-p/--port	Listening port	80
-c/--connections	Max parallel connections per stream	4
-s/--chunk-size	Download chunk size (e.g., 1M, 512K)	1M
-r/--retry	Max retries per chunk	3
--log-level	Logging verbosity	INFO

Example

Start server with 16 connections and 2MB chunks:

mflow --port 9000 --connections 16 --chunk-size 2M

Play 4K video through proxy:

mpv "http://localhost:9000/stream?url=https://media.example.com/4k-demo.mkv"

Notes

1. Server acts as pass-through - no data is stored locally
2. Supports most HTTP servers (including HTTPS sources)
3. Connection pool is shared between all streaming sessions
4. Terminate with Ctrl+C when not needed

For debugging use:

mflow --log-level DEBUG