• Stars
    star
    239
  • Rank 167,863 (Top 4 %)
  • Language
    Go
  • License
    MIT License
  • Created almost 3 years ago
  • Updated almost 2 years ago

Reviews

There are no reviews yet. Be the first to send feedback to the community and the maintainers!

Repository Details

A concurrent rate limiter library for Golang based on Sliding-Window rate limiter algorithm.

ratelimiter

Tests Go Reference

A generic concurrent rate limiter library for Golang based on Sliding-window rate limitng algorithm.

The implementation of rate-limiter algorithm is based on Scalable Distributed Rate Limiter algorithm used in Kong API gateway. Read this blog for more details.

This library can be used in your codebase to rate-limit literally anything. For example, you can integrate this library to provide rate-limiting for your REST/gRPC APIs or you can use this library to rate-limit the number of go-routines spawned or number of tasks submitted to a function/module per given time interval. This library provides generic rate check APIs that can be used anywhere. The library is built with concurrency in mind from the groud up, the rate-limiter can be used across go-routines without having to worry about synchronization issues. This library also provides capability to create and manage multiple rate-limiters with different configurations assiociated with unique keys.

How is this different from Go's official rate package?

The official Go package provides a rate limiter implementation which uses Token Bucket Algorithm. This repository (i.e the current repository) implements rate limiting functionalities using Sliding Window Rate Limiting Algorithm as used in Kong API gateway. Both of these libraries provide the same functionality, but there are trade-offs between these two algorithms, this blog properly explains these trade-offs. Understand the trade-offs and match them with your requirements to decide which algorithm to use.

Installation:

The package can be installed as a Go module.

go get github.com/Narasimha1997/ratelimiter

Using the library:

There are two types of rate-limiters used.

All APIs:

  1. Generic rate-limiter:
	/* creates an instance of DefaultLimiter and returns it's pointer.
	   Parameters:
	 		limit: The number of tasks to be allowd
			size: duration
	*/
	func NewDefaultLimiter(limit uint64, size time.Duration) *DefaultLimiter

	/*
		Kill the limiter, returns error if the limiter has been killed already.
	*/
	func (s *DefaultLimiter) Kill() error

	/*
		Makes decison whether n tasks can be allowed or not.
		Parameters:
			n: number of tasks to be processed, set this as 1 for a single task. 
				(Example: An HTTP request)
		Returns (bool, error),
			if limiter is inactive (or it is killed), returns an error
			the boolean flag is either true - i.e n tasks can be allowed or false otherwise.
	*/
	func (s *DefaultLimiter) ShouldAllow(n uint64) (bool, error)

	/*
		Kill the limiter, returns error if the limiter has been killed already.
	*/
	func (s *DefaultLimiter) Kill() error	
  1. On-demand rate-limiter
	/*  creates an instance of SyncLimiter and returns it's pointer.
	 	Parameters:
	 		limit: The number of tasks to be allowd
			size: duration
	*/
	func NewSyncLimiter(limit uint64, size time.Duration) *SyncLimiter

	/*
		Kill the limiter, returns error if the limiter has been killed already.
	*/
	func (s *SyncLimiter) Kill() error

	/*
		Makes decison whether n tasks can be allowed or not.
		Parameters:
			n: number of tasks to be processed, set this as 1 for a single task. 
				(Example: An HTTP request)
		Returns (bool, error),
			if limiter is inactive (or it is killed), returns an error
			the boolean flag is either true - i.e n tasks can be allowed or false otherwise.
	*/
	func (s *SyncLimiter) ShouldAllow(n uint64) (bool, error)

	/*
		Kill the limiter, returns error if the limiter has been killed already.
	*/
	func (s *SyncLimiter) Kill() error
  1. Attribute based Rate Limiter
	/*
		Creates an instance of AttributeBasedLimiter and returns it's pointer.
		Parameters:
			backgroundSliding: if set to true, DefaultLimiter will be used as an underlying limiter.
							   else, SyncLimiter will be used.
	*/
	func NewAttributeBasedLimiter(backgroundSliding bool) *AttributeBasedLimiter

	/*
		Check if AttributeBasedLimiter has a limiter for the key.
		Parameters:
			key: a unique key string, example: IP address, token, uuid etc
		Returns a boolean flag, if true, the key is already present, false otherwise.
	*/
	func (a *AttributeBasedLimiter) HasKey(key string) bool

	/*
		Create a new key-limiter assiociation.
		Parameters:
			key: a unique key string, example: IP address, token, uuid etc
			limit: The number of tasks to be allowd
			size: duration
		Returns error if the key already exist.
	*/

	func (a *AttributeBasedLimiter) CreateNewKey(
		key string, limit uint64, 
		size time.Duration,
	) error

	/* 
	   check if AttributeBasedLimiter has a limiter for the key.
	   Create a new key-limiter assiociation if the key not exists.
	   Parameters:
	    key: a unique key string, example: IP address, token, uuid etc.
		limit: The number of tasks to be allowd
		size: duration
		Return true if the key exists or is created successfully.
	*/
	func (a *AttributeBasedLimiter) HasOrCreateKey(key string, limit uint64, size time.Duration);

	/*
		Makes decison whether n tasks can be allowed or not.
		Parameters:
			key: a unique key string, example: IP address, token, uuid etc
			n: number of tasks to be processed, set this as 1 for a single task. 
				(Example: An HTTP request)
		Returns (bool, error),
			if limiter is inactive (or it is killed) or key is not present, returns an error
			the boolean flag is either true - i.e n tasks can be allowed or false otherwise.
	*/
	func (a *AttributeBasedLimiter) ShouldAllow(key string, n uint64) (bool, error)

	/* 
		MustShouldAllow makes decison whether n tasks can be allowed or not.
		Creates a new key if it does not exist.
		Parameters:
			key: a unique key string, example: IP address, token, uuid etc
			n: number of tasks to be processed, set this as 1 for a single task.
			(Example: An HTTP request)
			limit: The number of tasks to be allowd
			size: duration

		Returns bool.
			(false) when limiter is inactive (or it is killed) or n tasks can be not allowed.
			(true) when n tasks can be allowed or new key-limiter.
	*/
	func (a *AttributeBasedLimiter) MustShouldAllow(key string, n uint64, limit uint64, size time.Duration) bool

	/*
		Remove the key and kill its underlying limiter.
		Parameters:
			key: a unique key string, example: IP address, token, uuid etc
		Returns an error if the key is not present.
	*/
	func (a *AttributeBasedLimiter) DeleteKey(key string) error

Examples and Explanation of each type of rate-limiter:

Generic rate-limiter

The generic rate-limiter instance can be created if you want to have a single rate-limiter with single configuration for everything. The generic rate-limiter can be created by calling NewDefaultLimiter() function and by passing the limit and size as parameters. Example:

func GenericRateLimiter() {
	/* create an instance of Limiter.
	format: NewLimiter(limit uint64, size time.Duration),
	where:
		limit: The number of tasks/items that should be allowed.
		size: The window size, i.e the time interval during which the limit
				should be imposed.
		To summarize, if limit = 100 and duration = 5s, then allow 100 items per 5 seconds
	*/

	limiter := ratelimiter.NewDefaultLimiter(
		100, time.Second*5,
	)

	/*
		Cleaning up the limiter: Once the limiter is no longer required,
		the underlying goroutines and resources used by the limiter can be cleaned up.
		This can be done using:
			limiter.Kill(),
		Returns an error if the limiter is already being killed.
	*/

	defer limiter.Kill()

	/*
		the limiter provides ShouldAllow(N uint64) function which
		returns true/false if N items/tasks can be allowed during current
		time interval.

		An error is returned if the limiter is already killed.
	*/

	// ShouldAllow(N uint64) -> returns bool, error

	// should return true
	fmt.Println(limiter.ShouldAllow(60))
	// should return false, because (60 + 50 = 110) > 100 during this window
	fmt.Println(limiter.ShouldAllow(50))
	// sleep for some time
	time.Sleep(5 * time.Second)
	// should return true, because the previous window has been slided over
	fmt.Println(limiter.ShouldAllow(20))
}

On demand window sliding:

The previous method i.e the Generic Rate limiter spins up a background goroutine that takes care of sliding the rate-limiting window whenever it's size expires, because of this, rate-limiting check function ShouldAllow has fewer steps and takes very less time to make decision. But if your application manages a large number of Limiters, for example a web-server that performs rate-limiting across hundreds of different IPs, then your AttributeBasedRateLimiter spins up a goroutine for each unique IP and thus lot of such routines needs to be manitanied, this might induce scheduling pressure.

An alternative solution is to use a rate-limiter does not require a background routine, instead the window is sliding is taken care by ShouldAllow function itself, this method can be used to maintain large number of rate limiters without any scheduling pressure. This limiter is called SyncLimiter and can be used just like DefaultLimiter, because SyncLimiter and DefaultLimiter are built on top of the same Limiter interface. To use this, just replace NewDefaultLimiter with NewSyncLimiter

......

	limiter := ratelimiter.NewSyncLimiter(
		100, time.Second*5,
	)
......

Attribute based rate-limiter:

Attribute based rate-limiter can hold multiple rate-limiters with different configurations in a map of <string, Limiter> type. Each limiter is uniquely identified by a key. Calling NewAttributeBasedLimiter() will create an empty rate limiter with no entries.

func AttributeRateLimiter() {
	/*
		Attribute based rate-limiter can hold multiple
		rate-limiters with different configurations in a map
		of <string, Limiter> type. Each limiter is uniquely identified
		by a key. Calling NewAttributeBasedLimiter() will create an empty
		rate limiter with no entries.
	*/
	/*
		Attribute based rate-limiter has a boolean parameter called:
		`backgroundSliding` - if set to true, the attribute based rate-limiter
		uses Limiter instance and each Limiter instance have it's own background goroutine
		to manage sliding window. This might be resource expensive for large number of attributes,
		but is faster than SyncLimiter.

		Disable this, i.e pass `false` if you want to manage large number of attributes
		in less memory and compute, sacrifcing a minimal amount of performance.
	*/
	limiter := ratelimiter.NewAttributeBasedLimiter(true)

	/*
		Now we are adding a new entry to the limiter, we pass:
			key: A string that is used to uniquely identify the rate-limiter.
			limit: The number of tasks/items that should be allowed.
			size: The window size, i.e the time interval during which the limit
				should be imposed.

		returns error if the key already exists in the map.
	*/
	// we have two articles here (for example)
	article_ids := []string{"article_id=10", "article_id=11"}

	// for article_id=10, allow 10 tasks/items per every second
	err := limiter.CreateNewKey(&article_ids[0], 10, 5*time.Second)
	if err != nil {
		log.Fatalln(err)
	}
	// for article_id=11, allow 100 tasks/items per every 6 minutes
	err = limiter.CreateNewKey(&article_ids[1], 100, 6*time.Minute)
	if err != nil {
		log.Fatalln(err)
	}
	// rates can be checked by passing key and N as parameters
	// Can I make 8 requests to article_id=10 during this time window?

	// ShouldAllow(key *string, N uint64) returns (bool, error)
	// the bool is true/false, true if it can be allowed
	// false if it cant be allowed.
	// error if key is not found.

	fmt.Println(limiter.ShouldAllow(&article_ids[0], 8))
	// Can I make 104 requests to article_id=11 during this time window?
	fmt.Println(limiter.ShouldAllow(&article_ids[0], 104))

	/*
		Other functions:
			1. HasKey: to check if the attribute already has given key
			   call: HasKey(key string) function.
			   Example: limiter.HasKey(&article_id[0])
			   Returns a bool, true if exists, false otherwise

			2. DeleteKey: to remove the key from attribute map
			   call: DeleteKey(key string) function.
			   Example: limiter.DeleteKey(&article_id[1])
			   Returns an error, if key was not in the map.
	*/
}

Using ratelimiter as a middleware with HTTP web server:

ratelimiter is pluggable and can be used anywhere. This code snippet shows how it can be used with Go's standard HTTP library when building a web server:

.....
// allow 100 requests every 5 seconds
limiter := ratelimiter.NewSyncLimiter(100, time.Second * 5)

// register the handler
rateLimiterHandler := func(next http.Handler) http.Handler {
	return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
		allowed, err := limiter.ShouldAllow(1)
		if err != nil {
			log.Fatalln(err)
		}
		if allowed {
			next.ServeHTTP(w, r)
		}
	})
}

// create a test route handler
ponger := func(w http.ResponseWriter, r *http.Request) {
	w.Write([]byte("Pong!!"))
}

// attach the ratelimiter middleware:
muxServer := http.NewServeMux()
muxServer.Handle("/", rateLimiterHandler(
	http.HandlerFunc(ponger),
))

// start the server
err := http.ListenAndServe(":6000", muxServer)
if err != nil {
	log.Fatalln(err)
}

The complete example can be found at examples/http-server/server.go. curl was used to simulate X requests per second and following was the output, as logged.

................................
2021/10/05 14:12:38 Iteration: 7, Requests received: 522, Allowed: 99
2021/10/05 14:12:43 Iteration: 8, Requests received: 533, Allowed: 101
2021/10/05 14:12:48 Iteration: 9, Requests received: 515, Allowed: 100
2021/10/05 14:12:53 Iteration: 10, Requests received: 505, Allowed: 100
2021/10/05 14:12:58 Iteration: 11, Requests received: 508, Allowed: 100
2021/10/05 14:13:03 Iteration: 12, Requests received: 474, Allowed: 100
2021/10/05 14:13:08 Iteration: 13, Requests received: 495, Allowed: 100
2021/10/05 14:13:13 Iteration: 14, Requests received: 478, Allowed: 100
..................................

The ratelimiter was able to balance the requested limit as specified. If you have installed the package, you can simply run the webserver as follows:

go run examples/http-server/server.go

Testing

Tests are written in attribute_limiter_test.go and limiter_test.go files. To execute the tests, simply run:

go test ./ -v

These are some of the results from tests:

  1. Single goroutine, Generic limiter: This test configures the rate-limiter to allow 100 requests/sec and fires 500 requests/sec with a time gap of 2ms each, allowed requests are counted and is tested with difference +/- 3. The same test is run for 10 samples. Here are the results:
=== RUN   TestLimiterAccuracy
Iteration 1, Allowed tasks: 100, passed rate limiting accuracy test.
Iteration 2, Allowed tasks: 101, passed rate limiting accuracy test.
Iteration 3, Allowed tasks: 100, passed rate limiting accuracy test.
Iteration 4, Allowed tasks: 100, passed rate limiting accuracy test.
Iteration 5, Allowed tasks: 100, passed rate limiting accuracy test.
Iteration 6, Allowed tasks: 100, passed rate limiting accuracy test.
Iteration 7, Allowed tasks: 101, passed rate limiting accuracy test.
Iteration 8, Allowed tasks: 100, passed rate limiting accuracy test.
Iteration 9, Allowed tasks: 100, passed rate limiting accuracy test.
Iteration 10, Allowed tasks: 100, passed rate limiting accuracy test.
--- PASS: TestLimiterAccuracy (10.01s)
  1. 4 goroutines, Generic Limiter: This test configures the limiter to allow 100 requests/sec and spins up 4 goroutines, the same limiter is shared across all the routines. Each goroutine generates 500 requests/sec with 2ms time gap between 2 requests. Allowed requests are counted per each goroutine, the result sum of all counts should be almost equal to 100. The accuracy is measured considering +/- 3 as error offset. The same test is conducted 10 times. Here are the results:
=== RUN   TestConcurrentLimiterAccuracy
Iteration 1, Allowed tasks: 101, passed rate limiting accuracy test.
Iteration 2, Allowed tasks: 100, passed rate limiting accuracy test.
Iteration 3, Allowed tasks: 100, passed rate limiting accuracy test.
Iteration 4, Allowed tasks: 100, passed rate limiting accuracy test.
Iteration 5, Allowed tasks: 100, passed rate limiting accuracy test.
Iteration 6, Allowed tasks: 100, passed rate limiting accuracy test.
Iteration 7, Allowed tasks: 100, passed rate limiting accuracy test.
Iteration 8, Allowed tasks: 100, passed rate limiting accuracy test.
Iteration 9, Allowed tasks: 100, passed rate limiting accuracy test.
Iteration 10, Allowed tasks: 100, passed rate limiting accuracy test.
--- PASS: TestConcurrentLimiterAccuracy (10.01s)
  1. 2 goroutines, 2 attribute keys, Attribute based limiter: An attribute based limiter is created with 2 keys, these keys are configured to allow 100 requests/sec and 123 requests/sec respectively. Two goroutines are created and same attribute based limiter is shared across. Each goroutine produces 500 requests/sec per key. The overall count is then verified for each goroutine with error offset of +/- 3. Here are the results:
=== RUN   TestAttributeBasedLimiterAccuracy
Iteration 1, Allowed tasks: 100, passed rate limiting accuracy test.
Iteration 1, Allowed tasks: 123, passed rate limiting accuracy test.
Iteration 2, Allowed tasks: 101, passed rate limiting accuracy test.
Iteration 2, Allowed tasks: 124, passed rate limiting accuracy test.
Iteration 3, Allowed tasks: 100, passed rate limiting accuracy test.
Iteration 3, Allowed tasks: 123, passed rate limiting accuracy test.
Iteration 4, Allowed tasks: 100, passed rate limiting accuracy test.
Iteration 4, Allowed tasks: 123, passed rate limiting accuracy test.
Iteration 5, Allowed tasks: 100, passed rate limiting accuracy test.
Iteration 5, Allowed tasks: 123, passed rate limiting accuracy test.
--- PASS: TestAttributeBasedLimiterAccuracy (5.00s)

Code coverage: To generate code coverage report, execute:

go test -coverprofile=c.out

This should print the following after running all the tests.

coverage: 99.0% of statements
ok      github.com/Narasimha1997/ratelimiter    25.099s

You can also save the results as HTML for more detailed code view of the coverage.

go tool cover -html=c.out -o coverage.html

This will generate a file called coverage.html. The coverage.html is provided in the repo which is pre-generated.

Benchmarks: Benchmarks can be executed by running:

go test -bench=.

Current benchmarks are as follows:

goos: linux
goarch: amd64
pkg: github.com/Narasimha1997/ratelimiter
cpu: Intel(R) Core(TM) i7-9750H CPU @ 2.60GHz
BenchmarkDefaultLimiter-12                      11732958                85.61 ns/op
BenchmarkSyncLimiter-12                          7047988               175.9 ns/op
BenchmarkConcurrentDefaultLimiter-12             7017625               163.9 ns/op
BenchmarkConcurrentSyncLimiter-12                4132976               256.3 ns/op
PASS
ok      github.com/Narasimha1997/ratelimiter    46.408s

Notes on test:

The testing code produces 500 requests/sec with 2ms precision time gap between each request. The accuracy of this 2ms time tick generation can differ from platform to platform, even a small difference of 500 micorseconds can add up together and give more time for test to run in the end because of clock drift, as a result the error offset +/- 3 might not always work.

Contributing

Feel free to raise issues, make pull requests or suggest new features.

More Repositories

1

fake-sms

A simple command line tool using which you can skip phone number based SMS verification by using a temporary phone number that acts like a proxy.
Go
2,695
star
2

boson-lang

A hybrid programming language written in Rust.
Rust
116
star
3

py4jshell

Simulating Log4j Remote Code Execution (RCE) vulnerability in a flask web server using python's logging library with custom formatter that simulates lookup substitution by executing remote exploit code.
Python
86
star
4

r3

A tiny x86_64 kernel written in Rust.
Rust
80
star
5

sig-716i

A CLI tool that can be used to disrupt wireless connectivity in your area by jamming all the wireless devices connected to multiple access points.
Go
79
star
6

MarvinOS

A hobby Operating System developed from scratch using C/C++ and assembly
C
67
star
7

smartreply

Unofficial port of Google's smart reply runtime (powers gmail and assistant) model to python, allowing developers to leverage intelligent smart reply as an API in Web and embedded systems that supports Linux, a loader (ld.so), a fully POSIX C++ Runtime and Python interpreter.
C++
56
star
8

pyMigrate

A tool for automatically migrating any python source code to a virtual environment with all dependencies automatically identified and installed. You can also use this tool to generate requirements.txt for your python code base, in general, this tool will help you to bring your old/hobby python codebase to production/distribution.
Python
52
star
9

rust-kernel-barebones

A minimal 64-bit rust kernel and a bunch of configuration scripts that can be used to bootstrap Operating system development using Nightly-Rust compiler.
Shell
45
star
10

clfu

Implementation of Constant Time LFU (least frequently used) cache in Go with concurrency safety.
Go
38
star
11

tor-proxy

An experimental Tor-Proxy serivce written in Go using Go-proxy and Go-libtor.
Go
35
star
12

gopg

A minimal microservice written in Go that executes Go programs. This microservice can be used to set-up local go learning environment at your workspace/school. You can also use the provided zero-configuration docker-image for quick deployments.
Go
29
star
13

aio-eth

A simple python library that can be used to run large Web3 queries on Ethereum blockchain concurrently as per Ethereum JSON-RPC specification.
Python
28
star
14

pavlos

A light-weight container runtime for Linux with NVIDIA gpu support, allows developers to quicky setup development environments for dev and test. Pavlos can emulate any Linux rootfs image as a container.
Go
26
star
15

wavenet-stt

An end-to-end speech recognition system with Wavenet. Built using C++ and python.
Python
21
star
16

PyThreads

A simple python library that makes creation of threads easier. It makes functions asynchronous with only one extra line of code.
Python
21
star
17

redis-pydict

A python dictionary that uses Redis as in-memory storage backend to facilitate distributed computing applications development.
Python
19
star
18

py_cpu

Python bindings for Google's cpu_features library. Allows python developers to enable hardware specific optimizations at runtime.
C++
19
star
19

bzCompute

bzCompute is a computation graph library with built-in support for domain-decomposition and prallel computation. The library can be used for expressing and executing large number of mathematical and text-processing operations using Data-Flow graphs, (Educational version of tensorflow), written in pure python code with numpy support.
Python
16
star
20

py-smartreply

Python bindings for Google's Smart Reply (AI based chat suggestions) Tensorflow Lite model.
C++
13
star
21

httppool

A simple asynchronous worker pool and concurrent job queue for Golang's HTTP Server, inspired from Node.js asynchronous http server design.
Go
8
star
22

exif_service

A C++ microservice used to extract location and other forensic information from JPEG images.
C++
8
star
23

StackMachine

A simple educational virtual computer machine that can execute simple arithmetic and logical programs, This Virtual Machine has it's own memory model, instruction queue, virtual CPU and a compiler that comes with a parser.
Java
7
star
24

go-cowin

Unofficial GO SDK for Indian Government's Co-WIN API Platform. This SDK helps developers to easily integrate Co-WIN APIs with their existing eco-system.
Go
7
star
25

C-Dict

A tiny library that brings the support of dictionaries to C programming language with a fast lookup using hash tables. dict type can be used to associate large arrays with string keys.
C
7
star
26

bit_vector

A header-only bit vector library for C . This can be used for implementing dynamic bit-vectors for building Bloom-Filters and Hyper-Logs .
C
6
star
27

cc-builder

Live compilation and linking tool for C/C++ projects written in Go.
Go
6
star
28

BloodCell-Identification-tfjs-client

Blood Cell Identification system is a web app built using React and Tensorflow js, The CNN model trained to classify blood cell images using python and is ported to Browser Client using tensorflow model converter.
JavaScript
6
star
29

IronServer

A simple, amazingly fast static web server written in C. Supports faster MIME type resolution and caching.
C
5
star
30

Blood-Cell-type-identification-using-CNN-classifier

Notebook providing solution for kaggle dataset to classify blood cell types (Built using Convolution Neural Network)
Jupyter Notebook
5
star
31

PyQ

A python library for expressing computations using Message passing queues.
Python
4
star
32

socketio-cgi

A cgi layer that allows any process / program to stream data across world wide web.
Go
4
star
33

ptw

A minimal boilerplate for building frontends using Preact, TypeScript and Web Assembly.
JavaScript
4
star
34

dipmp

A decentralized package registry for python built on top of IPFS and Ethereum - using python, solidity and pinata. dipmp can be used with pip.
Python
3
star
35

GroupChat-C

A simple group chat server that one can write using C/C++, with the support of external websocket router.
C
3
star
36

Inception-On-Device-inference

An android application for on-device machine learning using TensorFlow and GoogleNet (Inception v3)
Java
3
star
37

DirtyPointer-Hack

A hackaround on C++ private class members and methods to access them outside the class even though they are private. (GNU GCC compiler, Linux)
C++
3
star
38

encoder-aas

An architecture providing Universal Sentence Encoder as a service by exploiting job-level parallelism of multi-core architectures. The service can be used as a transformer model for downstream NLP and NLU tasks like Language modelling, Sentiment Classification, Question Answering, Semantic Search and more.
Python
3
star
39

packed-encoder

A tiny rust crate that can be used to encode data of different types into a packed byte array which can be passed over network, system calls or FFI.
Rust
3
star
40

VSNET_2

Image-Recognition system using Client(android app) and Server(TensorFlow with Flask) model.
Java
2
star
41

Narasimha1997

My profile README
2
star
42

Sento---Online-Sentiment-Analysis-tool

Sento is a free online sentiment analysis tool .
JavaScript
2
star
43

360Player

A 360 image and Video player based on Google Cardboard Platform, It's just like an example of how to use VrViews in android
Java
2
star
44

pnuemonia_classification_onnx

Pnuemonia classification at scale using ONNX runtime and an inference server with batching support.
Jupyter Notebook
2
star
45

VegAnalytics

A dashboard tool that provides end-to-end solution for local marts and vegetable sellers.
JavaScript
1
star
46

Mark_2

A Django platform to transfer large files between your Phone and your Personal Computer.
Python
1
star
47

FaceRecognition-tool

FaceRecognition is a simple and modular face recognition tool built using OpenFace and face_recognition.
Python
1
star
48

News-Bot

An artifically intelligent news and weather bot for Facebook messenger
Python
1
star
49

NewsMaster

a react app for real-time news on any device (First stage build)
HTML
1
star
50

GameChanger

Java
1
star
51

Speech-net

Online text-to-speech platform
HTML
1
star
52

ProjectHub

DBMS mini project !
JavaScript
1
star
53

LiteNet

A simple CNN architecture to solve simple computer vision problems.
Python
1
star
54

OS-Development-Tutorial

A step by step, cumulative guide for building an Operating System for x86 and it's variants.
1
star
55

Visualnet

Simple image recognition server using TensorFlow inception v3
Python
1
star
56

Narasimha1997.github.io

JavaScript
1
star
57

FastVideo-transformation

A Python backend architecture for fast video processing by Machine Learning models based on Producer - Consumer model and Message Passing Queues
Python
1
star
58

ObjectCode-Executor

Object code generator and executor built for C, Python program takes C file as input, generates Object code and another C file that executes it.
Python
1
star
59

blog

Source tree of my blog
1
star
60

Simple-Service-Broadcast

A simple UDP broadcast protocol used for service discovery in a local network
Python
1
star
61

bls-server

A gRPC server written in python that provides BLS (Boneh–Lynn–Shacham) signatures related functionalities like signing, verification and signatures aggregation - used in production at some places.
Python
1
star
62

klogger

A C++ CLI tool for linux that streams kernel logs and events over a light-weight UDP socket to remote machines. You can use this tool for debugging while developing Linux kernel extensions and custom Linux kernel modules, especially for embedded systems that don't have built-in display. (Cross compile for aarch64 and others).
C++
1
star
63

libIgnition-dev

A byte-stream based File-Encryption library built using Python and C (using Fernet encryption algorithm)
HTML
1
star