DEV Community: Arjun Shetty

notes- local AI setup

Arjun Shetty — Thu, 05 Mar 2026 04:37:02 +0000

Use LLMFit to select a model from the list.
Download the selected model. This will save the downloaded model in .cache/llmfit/models/Qwen2.5-Coder-7B-Instruct-Q8_0.gguf
Get llama.cpp from the releases tab
Use this command llama-server -m .cache/llmfit/models/Qwen2.5-Coder-7B-Instruct-Q8_0.gguf -ngl -1 to run the localhost server of chat interface
-ngl is basically to say for how many layers of GPU to be used, -1 is auto.

Notes - Array in JSONB column

Arjun Shetty — Tue, 09 Sep 2025 04:15:09 +0000

Adding new rows

INSERT INTO USER (DEPARTMENTS)
VALUES ('["dept1","dept2"]')

Adding an new value to the column

UPDATE USER
SET DEPARTMENTS = DEPARTMENTS || '["dept3"]'::JSONB

Removing an value from the column

UPDATE USER
SET DEPARTMENTS = DEPARTMENTS - '["dept1"]'::JSONB

devtools on low resolution monitors

Arjun Shetty — Fri, 28 May 2021 12:00:28 +0000

Finding a low resolution monitor is rare still let me take chance and see if it helps anyone.

As as front-end having a low resolution slows you down a tiny bit at-least. tried various ways to place the chrome devtools left, right, separate window, nothing seemed to work. I mostly face issues with screen width when I want to debug JS and look at the UI state. Keeping the responsive UI >1024 and having a decent view on my js code as well.

So finally there is one around other than having to high res monitor 😉.

There is an option on the dev tools to add a custom device where you enter the fixed resolution.

Make use of the zoom level to your comfort.
This way it helps to keep your UI fixed resolution while you re-size the dev tools window.

Code on!
Photo by patricia serna

Saucelab@localhost

Arjun Shetty — Thu, 27 May 2021 18:14:45 +0000

Connecting Saucelab to your local machine is easy.

Create an outbound rule to expose the port for Saucelab to access your localhost.

Download the Sauce Connect Proxy.

Open the folder in windows terminal/ cmd C:\Users\Downloads\sc-<version>-win32\sc-<version>-win32.

Now go to tunnels page on your Saucelab dashboard

Copy the run command looks something lik this bin/sc -u <name> -k <guid> -i tunnel_id

Go to Live web testing and start session with your Sauce connect proxy selected from the dropdown.
You should be able to access your localhost sites for the port your exposed to Saucelab.

Developers Ego

Arjun Shetty — Sat, 22 May 2021 06:39:52 +0000

Ego Hurts!

Everyone has ego accepting it just the first stage. Its hard to take comments from anyone on your beautifully written code, as the quote goes...

Beauty is in the eye of the beholder

Its human nature to have it nothing wrong with it. There are 8 billion perspectives so is the population.

When a developer writes code he would mostly want cover as many scenarios there is, being human is our only limitation for now 😉.

We get so involved writing to those limited perspectives we can imagine, that it hurts when someone points out the missed ones. Only thing that teaches to take in criticism is experience.

Don't worry about the ego it fades with time 😇

Code on!

Photo by Charles Deluvio

Sending email using flowable workflow engine

Arjun Shetty — Fri, 27 Nov 2020 18:01:43 +0000

Integrating Email task in Flowable. If doing from flowable-UI follow this.

Once you are done with the filing To, From, Html/HtmlVar, etc,. Configure the email server configuration in your spring application.



# email config
flowable.mail.server.host=smtp.google.com
flowable.mail.server.port=587
flowable.mail.server.username=arjun
flowable.mail.server.password=arjun
flowable.mail.server.use-tls=true

In case you want to add the email task manually paste this node to your workflow bpmn file.



<serviceTask id="RejectionEmail" name="Rejection Email" flowable:type="mail">
      <extensionElements>
        <flowable:field name="to">
          <flowable:string><![CDATA[arjun@bitmonkey.in]]></flowable:string>
        </flowable:field>
        <flowable:field name="subject">
          <flowable:string><![CDATA[Rejection]]></flowable:string>
        </flowable:field>
        <flowable:field name="html">
          <flowable:string><![CDATA[<a href="http://bitsmonkey.in">Rejection reason</a>
<input type="textarea">Dummy Message</input>]]></flowable:string>
        </flowable:field>
      </extensionElements>
    </serviceTask>

Code On!

Originally posted on Bitsmonkey

Getting new instance every time in Java Spring Boot

Arjun Shetty — Thu, 26 Nov 2020 17:29:10 +0000

This is one of the ways to get a new instance of a type everytime we request for it from a singleton scoped instance.

@Lookup Informs spring container to getBean from method decorated with this annotation.

Creating a method in configuration class so that it can be used anywhere configuration is injected.

//ApplicationConfiguration.java
@Lookup
public UserDTO getUserDTOProtoTypeBean() {
    return null;
}

Make sure the UserDTO class is decorated with @Scope("prototype"). Now in your consuming class call this method when you need a new instance.

@Component
public class UserMapper {

    @Autowired
    ApplicationConfiguration appConfig;

    public UserDTO ToDto(User userEntity) {
        var UserDTO = appConfig.getUserDTOProtoTypeBean();
        ...
    }
}

Code On!

Originally Posted on Bitsmonkey

Windows Tip #Color Filter

Arjun Shetty — Thu, 26 Nov 2020 09:33:16 +0000

FOCUS!

Does too many color distract you sometimes. You dont like the theme on some of the site you think too glossy colors. Here is a setting tht might help you filter the distracting colors.

Yet another cli with golang using cobra

Arjun Shetty — Wed, 25 Nov 2020 10:42:23 +0000

Getting work done the plain vanilla way is doing it in cli so I ended up here. Don't you love docker cli, github cli. who wouldn't? most of the new cli's are built using golang cobra framework.

Let's try this out. In this example I will try to build a cli which will talk to DB like getting schema information from a postgres DB.
For starters I recommend using Cobra Generators.
Make sure if you are designing something production grade follow this simple quote from the cobra readme

Commands represent actions, Args are things and Flags are modifiers for those actions.

This sample is just to touch up upon basics of what is required to start with Cobra cli. Here is the final project structure for future references

cmd/      # all the command implementations are here     
config/   # configuration related to this app like connection string etc,.
go.mod    # go module
go.sum    # go dependencies haskeys
handlers/ # command from the cmd talk to handlers which does the actual call the services
main.go   # applications starting point
repo/     # all the services implementation in our case its a service talking to the DB so name repo

All the command start from the rootCmd root command. and it can have any number of child or grand child. Here we will create root-->list-->table and root-->list-->schema, so list has 2 children and their grandfather is root.

Lets use Cobra code generator to create commands its mostly auto fills a lot of things for us here is how it looks.

go get github.com/spf13/cobra/cobra

cobra init [app]

cobra add list

cobra add list -p 'tableCmd'

var tableCmd = &cobra.Command{
    Use:   "table",
    Short: "list tables in schema",
    Long: ` list tables <schema> - will list the tables in the specified schema
            --limit - option will let you limit the number of items in the result
            --offset - used as a start index from nth item to retrieve`,
    Run: handlers.TableHandle(),
}

func init() {
    listCmd.AddCommand(tableCmd) // adds child to table command whose parent is list.   
    tableCmd.PersistentFlags().StringVarP(&namespace, "schema", "s", config.Data.DefaultNameSpace, "Schema name")

}

So the command flags blah blah is already you can try running go run main.go list table. This will invoke whatever is assigned to the Run property. In our case I have seperated the handling of business function to handlers. Also note I have tried using config.Data.DefaultNameSpace config using a struct so that we make no mistakes when getting config by string values.

This is how the config class looks

var (
    // Data ...
    Data configuration
)

//Configuration ...
type configuration struct {
    ConnectionString string
    DefaultNameSpace string
}

func init() {
    viper.AddConfigPath(".")
    viper.SetConfigType("yaml")
    viper.SetConfigName(".config")

    viper.AutomaticEnv()

    if err := viper.ReadInConfig(); err == nil {
        err := viper.Unmarshal(&Data)
        if err != nil {
            fmt.Printf("Unable to decode into struct, %v", err)
        }
    }
}

If you observe there is a method called init this method is called as soon it is tried to access. Here is a nice explanation from stackoverflow answer.

#.config.yaml

ConnectionString: user=bitsmonkey password=password host=db.com port=5432 dbname=jimmy connect_timeout=20 sslmode=verify-full
PGSSLROOTCERT: cert.pem
DefaultNameSpace: ns

This line tableCmd.PersistentFlags basically says I need a flag that will be used across all my children commands in this case.

If you look at the main its clean, all it does is calls the root command cmd.Execute().

//TableHandle ...
func TableHandle() func(cmd *cobra.Command, args []string) {

    return func(cmd *cobra.Command, args []string) {
        namespace, err := cmd.Flags().GetString("schema")

        if err != nil {
            log.Println(err)
            return
        }
        log.Println("Namespace : ", namespace)

        limit, err := cmd.Flags().GetInt("limit")
        if err != nil {
            log.Println(err)
            return
        }
        log.Println("Limit: ", limit)

        offset, err := cmd.Flags().GetInt("offset")
        if err != nil {
            log.Println(err)
            return
        }
        log.Println("Offset: ", offset)

        ddlRepo := repo.Ddl{}
        if result, err := ddlRepo.GetAllTablesIn(namespace, limit, offset); err != nil {
            log.Println("Error when retrieving tables in namespace", namespace)
        } else {
            fmt.Println("---------List of Tables---------\n", strings.Join(result, "\n"))
        }

    }
}

This is should be a good start for writing any cli app anymore. Code on!

Photo by Pankaj Patel
Originally Posted on Bitsmonkey

Dotnet core GraphQL with postgres

Arjun Shetty — Sat, 14 Nov 2020 17:06:03 +0000

Lets create a GraphQL application in dotnet core 3.1 using HotChoclate.

Creating a new web api project

dotnet new web -n GraphApi

This package mostly handles all the aspects of Graphql. Also important factor for choosing this library is it passes all the kitchen sink test, more details here

dotnet add package HotChocolate.AspNetCore --version 10.5.3

Adding GraphQL playground IDE for easing our development.

dotnet add package HotChocolate.AspNetCore.Playground --version 10.5.3

One last dependency is the the postgres efcore

dotnet add package Npgsql.EntityFrameworkCore.PostgreSQL --version 3.1.4

The Pi version 😄.

Lets set thing in Startup.cs now where all the middleware is registered.

services.AddScoped<DummyService>();

services.AddDbContext<ApplicationDbContext>(options =>
    options.UseNpgsql("User Id=username; Password=password; Server=dbname.host.com; Port=7530 ; Database=dbname; Ssl Mode=Require; Server Compatibility Mode=Redshift",
    builder => builder.ProvideClientCertificatesCallback(cert =>
    {
        cert.Add(
            new System.Security.Cryptography.X509Certificates.X509Certificate("./Cert/cert.pem")
            );
    }
)));

The first class registered is something that we will create later.

Add the above in ConfigureServices method. Ignore the Ssl Mode=Require; Server Compatibility Mode=Redshift part in the connection string and the builder action, if you are not using SSL connection.

If using SSL connection note the Server Compatibility Mode=Redshift this is something you might get stuck with no much details on the error message. You will some error message like this one Unsupported startup parameter: extra_float_digits. More details on this official documentation.

services.AddGraphQL(s => SchemaBuilder.New()
                                    .AddServices(s)
                                    .AddType<DummyType>()
                                    .AddQueryType<Query>()
                                    .Create());

This is mostly all the HotChocolate registration where you are giving HotChocolate the details required to create schema, types and queries.

Now inside Configure method we will use the registered,

if (env.IsDevelopment())
{
    app.UseDeveloperExceptionPage();
    app.UsePlayground();
}

app.UseGraphQL("");

You business model will have a partnering model which is used to expose only the fields you want to the graphql queries as follows,

public class Dummy
{
    public long Id { get; set; }
    public string DummyProperty { get; set; }
}

public class DummyType : ObjectType<Quote>
{
    protected override void Configure(IObjectTypeDescriptor<Quote> descriptor)
    {
        descriptor.Field(x => x.DummyProperty).Type<StringType>();
    }

}

We have exposed only one property for pur business model Dummy. This represets the Type part of the Graphql. For the query node we will create a new class Query

public class Query
{
    private readonly DummyService _dummyService;

    public Query(DummyService dummyService)
    {
        _dummyService = dummyService;
    }

    public IQueryable<Dummy> DummyList => _dummyService.GetSomeData();

}

DummyService is just an abstraction for the efcore repository layer you can also inject DbContext directly if required.

The Efcore is something which is explained in here before so just paste code and skip to running the app part.


public class RepositoryContext : DbContext
{
    public DbSet<Dummy> DummyDbSet { get; set; }

    public ApplicationDbContext(DbContextOptions<RepositoryContext> options)
            : base(options)
    {
    }

    protected override void OnModelCreating(ModelBuilder modelBuilder)
    {
        modelBuilder.Entity<Dummy>(entity =>
        {
            entity.ToTable("tablename");

            entity.Property(x => x.Id).HasColumnName("Id");
            entity.Property(x => x.QuoteNumber).HasColumnName("Column1");

            entity.HasKey(x=>x.Id);
        });
    }
}

public class DummyService
{
    private readonly RepositoryContext _dbContext;

    public IQueryable<Dummy> GetSomeData() =>
            _dbContext.DummyDbSet;

    public DummyService(RepositoryContext dbContext)
    {
        _dbContext = dbContext;
    }
}

We are all set once we run this application we go to http:\\localhost:5000\playground, which opens up the GraphQL IDE.

run this query and it will list all the data.

dummyList{
    id
}

Enabling paging is as easy as adding this attribute in Query.

[UsePaging(SchemaType = typeof(QuoteType))]
public IQueryable<Dummy> DummyList => _dummyService.GetSomeData();

Explore away!

Photo by Praewthida K
Originally posted on Bitsmonkey

Connecting to postgres with standard sql package in golang

Arjun Shetty — Sat, 14 Nov 2020 10:24:55 +0000

Quick steps connecting to postgres DB using the stand library with postgres driver,

Import the packages

import (
    "database/sql"
    _ "github.com/lib/pq" //postgres drivers for initialization
)

A simple get method to run a select statement. In example we will the connection string from the environment variables.

var connStr = os.Getenv("CONNSTR")

//GetSomeDummyData ...
func (r *Repo) GetSomeDummyData() {

    db, _ := sql.Open("postgres", connStr)
    defer db.Close()

    id := "34940574"
    row := db.QueryRow("SELECT column1, column2 FROM dummytable where id=$1", id)

    rows, err := db.Query(query, qteNum, limit)
    defer rows.Close()
    if err != nil {
        log.Println(err)
    }

    var result []*model.DummyData

    for rows.Next() {
        d := &model.DummyData{}
        err = rows.Scan(&d.Prop1, &d.Prop2)

        if err != nil {
            log.Println(err)
            return nil, err
        }
        result = append(result, q)
    }
    return result, nil

}

Now the connection string looks something like this in the environment variable

EXPORT CONNSTR="user=username password=password host=dbname.host.com port=5433 dbname=dbname connect_timeout=20 sslmode=disable"

Lets now enable ssl mode with the ca cert in your project. In which case we need to set the path of the PEM file. Can be relative path of the absolute path. Postgres expects this in a environment variable PGSSLROOTCERT=pathof.pem and connection string will specify sslmode=verify-full

EXPORT PGSSLROOTCERT=pathof.pem
EXPORT CONNSTR="user=username password=password host=dbname.host.com port=5433 dbname=dbname connect_timeout=20 sslmode=verify-full"

Photo by Ron Whitaker

Originally posted on Bitsmonkey

Docker Tip

Arjun Shetty — Fri, 13 Nov 2020 14:30:37 +0000

Quick tip, You do not need to copy paste the full name or ID of a container to use that in the command. You can just use the first unique characters of its id alone like this.