EPIC: Prepare the repo for public share on HN 🚀

[nelsonic]

@LuchoTurtle you've done a superb job of building a fully functional image captioning app! 😍 🎉
Now it's time to get some credit for it 💳 😉 by submitting a "Show HN"
And in the process raise your public profile for applying to #NextAdventure 🚀

Todo

• Address the start-up time (currently 25sec): Chore: Get Help with Lowering Machine Start Time for imgai App on Fly.io #20

  • And/or create an endpoint to serve a 1x1px GIF to start the app proactively when someone is reading the README.md ref: Chore: Update this Project for use on Fly.io ping#7

• Improve the Why? section of README.md to answer the Reader's question succinctly.

Currently it's: https://github.com/dwyl/image-classifier/tree/a3c8c3bf79a0f5e7bca160b6e460d883bc6e3973#why-

This is good for an internally-focussed tutorial for us ... 👍
but as an LLM-curious person casually reading on HN, 👀
this isn't going to "hook" me into reading a 5k word tutorial for 30+ mins ... ⏳

# Why? 💭 

We needed a fully-offline capable (no 3rd party APIs/Services) image captioning service 
using a state-of-the-art pre-trained image model to describe images uploaded in our 
[`App`](https://github.com/dwyl/app).

# What?

A step-by-step tutorial building a fully functional 
`Phoenix LiveView` web application that allows anyone 
to upload an image and have it described 
by the `Open Source` `BLIP` image captioning (`Large`) model.

• Simplify layout of homepage of the App and add examples: Feat: Examples on the Landing Page!  #7
• Prepare Intro/Demo GIF: Feat: Intro GIF #21
• Add demo GIF to intro of README.md
  • GIF should be below the badges and above the TOC:

• Record Tutorial Video: Feat: Record Video Tutorial for this Repo! 📺  #23

Part 2

• Study the 20 most successful "Show HN" posts: https://hn.algolia.com/?query=show+hn&sort=byPopularity (spend 1-2 mins skimming each one to extract the info - or use ChatGPT to do some pre-analysis if you think it will be useful ...)
  Summarise what you learned as comments in this issue thread. 💬 🙏

• Carefully craft your Show HN Title to maximise its chance of success. 💭

• Share on https://elixirforum.com/ to get some initial traction/stars.

  • As soon as you submit on HN, update the Elixir Forum topic to include the HN link to drive up-votes.

• Submit your Show HN!! 🚀

• Share the link so we can up-vote it. 🔗 ⬆️

  • Get everyone you know to up-vote it so it gets and initial boost. 📈

[nelsonic]

@LuchoTurtle how close do you feel this repo is to sharing on HN? 💭 🚀

[LuchoTurtle]

Safe some changes to the README (as outlined on this issue), I think it brings sufficient value for those that want to get started with Bumblebee (unfortunately there aren't many examples out there with in-depth guides and comparisons).

However, I think #18 would also bring great value to this project and would be extremely interesting as it would use a part of Bumblebee that is not used here - voice-to-text.

[ndrean]

I can push several versions if you want help. 1) add an audio capture (an HTML addition with its JS Hook to upload to the server where we add a :speech allow_upload 2) ts handle_progress where you introduce the Speech-to-Text ML transcription 3) add embeddings and the KNN search. Let me know

[LuchoTurtle]

@ndrean any PR is helpful :). I think your idea of speech-to-text really takes this to another level. The purpose is to document the process, which is severely lacking in fly.io articles and in the bumblebee repo.

[ndrean]

@LuchoTurtle I understand your idea. Note that the key point of selecting the right model is probably the most difficult part. However, I did no effort on this. I just followed and adapted an article of Sean Moriarty on semantic search with ExFauss and adapted it. . It just works, and I feel I did not really learnt something - except the Elixir point of view if using tasks - in the sense that if you ask for something more difficult, like say an interactive LLM or build a model, then I have clearly no clue what to do. It's more like a one shot. But you have to start somewhere don't you!? Anyway, I always cite my sources and will push things as soon as my computer is available.