TOKI - Token Optimization and Knowledge Interface

• 

  Submission in action

• 

  All features in action

• 

  Testing session exhaust

Inspiration

It all started with a comment underneath an AI post on social media where a user stated he killed 2 trees for such a bad ai generated joke. It got me thinking, did it truly take 2 trees for chuckles from strangers? What if people actually knew how much they seem to be a part of an ever destroying catastrophe of energy and resource depletion that users never see. Which is why, to bridge this gap of unawareness, I built TOKI

What it does

TOKI - Token Optimization and Knowledge Integration is a chrome extension that sits on your favorite LLM across Gemini, Claude and ChatGPT, helping users track their token count, how much energy or resource it amounts to, Token Counter for each session [Currently documented only for claude with the help of Open Source] and the highlight of filler words that doesn't contribute at all to the result

How we built it

The following extension was vibe-coded, regardless the stack used was HTML,CSS and Javascript alongside Manifest due to the nature of being an extension. What I truly focused on the build was the User centricity. How do I appeal to the consumer without having to interrupt with their day-to-day sessions until they wish to use it. Our subtle nature of having to be one with the LLM, a user uses, is what I truly am proud of. Designed a prototype appeal on figma which was later sent for production.

Challenges we ran into

It was not documented online for how long does each session take until I came across Claude Counter where the author documents their study on how much tokens does each session contribute to. It was a very essential feature for non-technical users to know how far off were they from exhausting their current session and being more cautious rather than having to utilize a "skill" option such as in claude. The token counter feature is based off the author's open source code and helped us navigate through one of our most sensitive feature of TOKI.

Accomplishments that we're proud of

The user's interactions; No noise and a simple survey that only got 20 people in the initial stages to initialize market research.

What we learned

I learnt that to create a good product in today's world would be to sell it in such a way that the user wouldn't be forced or entitled to use it; A user's satisfaction of not having to be forced wouldn't develop a habit in the long run that TOKI wishes to achieve, a more responsible prompt engineering world.

What's next for TOKI - Token Optimization and Knowledge Interface

Regardless of how well it works on certain LLM's, it still is yet to be optimized for the others due to how different platforms interact, hence TOKI cannot be a universal solution or a one shoe fit for all. Hence, once TOKI's dashboard is completed where users can customize their extension and add their own filler words, I look forward to optimizing them for Gemini and ChatGPT for the same.

Built With

• claude
• css
• figma
• html
• javascript
• manifest-v3