DEV Community: ✨Nimmo✨

Ollama Models Comparison : Compare LLM Responses Side-by-Side

✨Nimmo✨ — Tue, 04 Mar 2025 16:16:12 +0000

Hey Dev.to community! 👋

Have you ever wondered which local LLM is better at answering your specific questions? Or if that fancy new model is actually worth the disk space? Well, I've been tinkering with Ollama lately, and I wanted an easy way to put different models head-to-head.

So I built the Ollama Model Comparator - a super simple but surprisingly useful tool that lets you compare responses from different Ollama models side by side.

What does it do?

It's pretty straightforward:

Pick any two Ollama models from a dropdown (or add custom ones)
Enter your prompt
Hit "Compare" and watch the magic happen
Check out not just the responses, but also the generation time and token counts

Tech stuff (for the curious)

The app is built with:

Python and Gradio for the UI (super easy to use!)
Requests library to talk to Ollama's API
A sprinkle of passion for local LLMs

Getting started in 30 seconds

Make sure you have Ollama running
Clone the repo: git clone https://github.com/your-username/ollama-model-comparator.git
Install deps: pip install gradio requests
Run it: python app.py
Open your browser to http://127.0.0.1:7860

That's it! No API keys, no complicated setup.

What's next & how you can help

This is just v1, and I'd love to make it better with your help! Some ideas I'm thinking about:

Adding response streaming (because waiting is boring)
Supporting more than two models at once (royal rumble style)
Adding more detailed comparison metrics
Making the UI prettier (I'm a backend person, can you tell? 😅)

If you're into local LLMs and want to contribute, I'd love your help! The code is simple enough that even if you're new to Python, you can probably jump in.

Try it out!

The project is available on GitHub under the MIT license.

Would love to hear what you think! Which models are you comparing? Any surprising results? Drop a comment below!

Happy comparing! 🚀

bashGPT: Convert Plain English to Bash Commands with GPT

✨Nimmo✨ — Sun, 25 Aug 2024 17:18:49 +0000

Introducing `bashGPT`: Convert Plain English to Bash Commands with AI

Do you often forget the right bash commands and wish you could just type what you want to do in plain English? Meet bashGPT—a handy command-line tool that uses OpenAI's GPT-3.5-turbo to turn your natural language instructions into working bash commands.

🌟 What is `bashGPT`?

bashGPT is a script that takes plain English descriptions of what you want to do and converts them into bash commands that you can execute directly from your terminal. It's particularly handy for those who are new to the command line or just want a more intuitive way to interact with their system.

🚀 How Does It Work?

The tool works by sending your plain English request to OpenAI's GPT-3.5-turbo model. The AI processes your input, translates it into a bash command, and then returns the result. But that's not all—before executing the command, bashGPT ensures that it's syntactically valid and asks for your confirmation. If the command can’t be converted into a valid bash command, the tool will return null and exit gracefully.

🛠️ Getting Started

To start using bashGPT, you'll need to clone the repository, set up your OpenAI API key, and make the script executable. Here’s a quick guide:

Clone the Repo:

git clone https://github.com/madhunimmo/bashGPT.git
cd bashGPT

Set Your OpenAI API Key:

export OPENAI_API_KEY="your-openai-api-key"

Make the Script Executable:
```
chmod +x bashGPT.sh
```

🔥 Example Usage

You can either pass your command as an argument:

./bashGPT.sh "list all files in the current directory"

Or run it interactively:

./bashGPT.sh
Please enter a command in plain English:
> list all files in the current directory

bashGPT will generate the bash command ls, ask you to confirm, and then execute it.

⚠️ Handling Invalid Commands

If your input can’t be translated into a valid bash command, it’ll return an error message like this:

Error: No valid bash command was generated.

This ensures that you’re always in control and only executing what’s actually possible in bash.

🤝 Test & Contribute

Try out bashGPT and let me know how it works for you! Bug reports, feature suggestions, and code contributions are welcome on GitHub. Even small improvements can make a big difference, so don't hesitate to jump in and help out. Thank you for your support!

How I Created My YouTube Video Summarizer: A Solution to Video Frustration

✨Nimmo✨ — Wed, 08 Nov 2023 01:37:39 +0000

We've all been there - searching for a quick answer or a specific piece of information in a YouTube video, only to find ourselves frustrated by lengthy videos filled with unnecessary content.

<iframe src="https://giphy.com/embed/QZIbPweLH20orIb84G" width="480" height="480" frameBorder="0" class="giphy-embed" allowFullScreen></iframe><p><a href="https://giphy.com/gifs/toilet-desperate-auf-klo-QZIbPweLH20orIb84G"></a></p>

The struggle to navigate through these long videos often leads to fast-forwarding or skipping, which can result in missing out on important details. It was this frustration that led me on a quest to create my own YouTube video summarizer, ensuring that I could easily access the information I needed without the hassle of sifting through lengthy content.

The Problem

The problem is simple: YouTube videos can be too long, and we often just want a concise summary of the key points. While manually copying and pasting the transcript of a video to a note-taking tool or searching for written summaries can be a solution, it's not always efficient or fun. This is where technology comes to the rescue.

The Solution: Using LLM chatbot API

My journey to develop a YouTube video summarizer began with the idea of using the ChatGPT API. The process involved three main steps:

Accessing the YouTube API to fetch video transcripts.
Utilizing the ChatGPT API to create a summary of the video.
Making this process accessible and user-friendly.

Let's dive into each of these steps.

Step 1: Accessing the YouTube API

The first step was to access the YouTube API to retrieve video transcripts. Fortunately, YouTube provides an API that allows developers to access a video's transcript (if available) and other metadata. With this API, I was able to fetch the content of the video, which served as the foundation for creating the video summarizer.

Step 2: Leveraging the ChatGPT API

With the video transcript in hand, it was time to create a concise summary of the video. My initial thought was to use the ChatGPT API for this purpose. However, I soon discovered that the ChatGPT API was not entirely free to use beyond a certain limit. This led me to search for a more cost-effective alternative while still maintaining the power of AI-driven summarization.

Step 3: Discovering Hugging Face's HuggingChat

My search for a free alternative brought me to Hugging Face's HuggingChat, another language model capable of summarizing text. However, there was no official API available for HuggingChat at that time. Fortunately, I stumbled upon an open-source API created by a developer who shared my vision of making this technology accessible.

The Final Solution

Bringing all the pieces together, I successfully created a YouTube video summarizer that anyone could use with ease. Here's how it works:

Provide a YouTube video link.
Fetch the video transcript using the YouTube API (if available).
Use the open-source HuggingChat API to summarize the video content.
The result is a concise and informative summary of the video, giving you access to the key points and information you need without the hassle of sifting through lengthy content.

The code is on GitHub here
Give it a try, and let me know how that works for you!

Find ES6 features in any JS code

✨Nimmo✨ — Fri, 14 Oct 2022 19:52:06 +0000

I came across a problem where I had to find the ES6 features used by any javascript project and other data regarding their use. When I reached out to stackoverflow, I could find only one relevant post which asks you to use linters like jshint/jshint or compilers like babel. Jslint didn't seem to report anything specific to ES6 and Babel converts all the ES6+ features to ES5 but doesn't report anything regarding which constructs were used or how many times they were used. However, Jshint reported all ES6 features used in the code along with some metadata. And, to suit my needs, I ended up writing a python script that calls Jshint on all JS files in a project and presents the features used in the project and the number of times they were used across all files. You can find the code here : jsHintRunner

Feel free to report a bug/issue or open a PR if you have improvement ideas.

Hope this helps someone at some point.

Fun with Quines

✨Nimmo✨ — Sun, 21 Feb 2021 02:01:29 +0000

I was recently watching Dylan Bettie's talk "Art of Code" in NDC 2020 and honestly, it made me fall in love with coding all over again. If you haven't yet watched here, you are missing a hell lot of fun that coding can bring you.

In his talk, among all other things, he also mentions of quines. A quine is basically a program that prints its own source code. You can read about it all over the internet. But that's not the point, the point is do you want to write one in the programming language you are most familiar with? Do you want to challenge your expertise in your comfort language? IF yes, this is an open invite to everyone here, to contribute to the repo QuineInAllLangs where I plan to accumulate quines in all possible programming languages for learning purposes. And, hope that you learn something as well in the process of writing your own equine and contributing to this repo.

Have a great weekend.

SubReddits Every Developer should follow

✨Nimmo✨ — Mon, 14 Dec 2020 18:00:15 +0000

I have always been a great fan of Reddit but, I also feel that it is kind of underrated and underutilized when it comes to developers and students. Till now, I have had a great experience over Reddit with seeking feedback, showcasing work, getting solutions, and hence, I felt why not share with everyone how great Reddit is. I have made a curated list of channels that I feel every developer must follow. Feel free to go over it and join the ones you like and I promise you won't regret it.
Also, don't forget to star the repo, as I will be updating it from time to time.

MustFollowRedditChannels

Also, I am looking for contributors to translate this resource in other languages so that it is accessible to all.Feel free to reach out to me if you want to contribute.

Simplest solution to JSON.stringify RangeError: Invalid string length

✨Nimmo✨ — Mon, 14 Dec 2020 03:19:30 +0000

We often come across this issue where the object or value passed to JSON.stringify is more than it can handle and the most common solution offered on google is using alternate libraries to stream the data whereas the simplest solution to it is just to stringify one element at a time and concatenate them.

var out="[";
  for(var indx=0;indx<data.length-1;indx++){
    out+=JSON.stringify(data[indx],null,4)+",";
  }
  out+=JSON.stringify(data[data.length-1],null,4)+"]";

You can also save the gist here for future references.
Hope this solution helps you at some point.

Github Profile Readme

✨Nimmo✨ — Tue, 08 Dec 2020 01:13:48 +0000

I recently released the Readme for my Github Profile with a cool new look. See it here.
If you like it and want a similar one, read further.

I used the Github Profile Readme Generator to generate it and to customize it further I used Github Stats .

And, that's it you are done.

For more inspiration and customization though feel free to check the below links:
Awesome Readme

Creative Profile Readme

Why am I here ?

✨Nimmo✨ — Tue, 08 Dec 2020 00:47:06 +0000

So, I've been lurking a bit in this community since COVID. While it seemed great and supportive, at the time I was really busy with work and didn't have much time to be involved.

Well, as of today, I thought, let's do this. Let me share some of my experiences and learnings with you all too. Looking forward to having great discussions.

Stay safe. Stay well.