NaviTron by Tron Trailblazers - TRON Ecosystem Navigation AI

Project Name: NaviTron

Project Track: AI

Team Name: Tron Trailblazers

Team Member(s): 2 (alexv, 0xclarence)

HackerEarth Project Link: NaviTron on HackerEarth

Project Goal: To provide users with a conversational AI assistant that guides them seamlessly through the TRON ecosystem, offering insights into dapps, token swaps, and more.

Problem Statement: When users search for “TRON Ecosystem” on popular search engines like Google, they are directed to the TRON ecosystem page. However, the user interface can be confusing, with no clear indication of which projects offer specific features, especially in areas like DeFi and NFTs. Moreover, platforms like ChatGPT only provide information up to September 2021, which restricts users from accessing details about TRON projects launched post that date. The solution? A conversational AI assistant to guide users through the TRON ecosystem.

Project Value: NaviTron bridges the information gap in the TRON ecosystem, offering real-time, conversational insights into dapps, token swaps, and other TRON-related queries, enhancing user experience and promoting deeper ecosystem engagement.

Project Info: NaviTron is a state-of-the-art TRON ecosystem navigation AI, leveraging the power of LangChain and GPT-4. Our team has trained a vector model using comprehensive documentation and feature sets from a curated list of dapps within the TRON ecosystem.

Demo Video: NaviTron Demo Video on Vimeo

Project Details:

  • User-Friendly Chat UI: A custom-designed chat interface for users to interact with the NaviTron AI.

  • Comprehensive TRON Ecosystem Knowledge:
    • Inquire about TRX token swaps, e.g., swapping TRX to SUN.
    • Discover stablecoins available on TRON.
    • Get details about dapps like Sunswap.

  • Direct Dapp Links: Provides users with direct links to the dapp’s webpage for easy navigation.

  • TRON Hackathon Insights (Bonus): Gain insights into projects from all TRON hackathon seasons.

Project Test Instructions:

  1. Visit the NaviTron app
  2. Ask NaviTron AI about the TRON ecosystem.
  3. Provide feedback in the topic about your experience.

Currently Supported Projects (more to come)

  • Sunswap
  • Transit Swap
  • Justlend
  • SocialSwap
  • Transitswap
  • JustMoney
  • JustStables
  • APENFT
  • BTTC
  • BTFS
  • Tronlink
  • Tronscan
  • Stablecoins (USDD, USDT, etc.)
  • HackaTRON winner projects

How NaviTron AI works:

We used the latest innovation of LangChain in order to train a custom vector model to synthesize latest information about the TRON ecosystem. With this model, in conjunction with GPT-4, we are able to create the conversations responses that NaviTron provides for user queries.

Here is a step-by-step guide on how the underlying LangChain system works:

  • Step 1: Load documents (e.g. dapp whitepapers, protocol docs, etc.)

  • Step 2: Splitting the document into chunks to reduce token usage

  • Step 3: Use Embedding against Chunks Data and convert to vectors

  • Step 4: Save data to the Vector database

  • Step 5: Take data (question) from the user and get the embedding

  • Step 6: Connect to VectorDB and do a semantic search

  • Step 7: Retrieve relevant responses based on user queries and send them to LLM (GPT-4)

  • Step 8: Get an answer from LLM and send it back to the user

Project Milestones:

  • Initial Development: Developed as a prototype for the TRON Hackathon S5.
  • Post-Hackathon: Seek official permissions from TRON DAO and associated ecosystem projects to enhance model training.
  • Beta Launch: Targeted by the end of the year.

Future Roadmap:

  • Add more projects to the vector database with prompt templates to improve accuracy.
  • Integrate with TronLink for personalized user wallet insights.
  • Enable direct dapp interaction within the chat UI using iframes & TronLink.
  • Facilitate custom transactions within the UI for direct dapp contract interactions.
  • Propose trading strategies for optimal wallet portfolio management.

Current Limitations:

  • Supports a limited number of dapps.
  • Requests may be rate-limited due to potential abuse in usage.
  • Responses may lack a known response at times as its been tailored for precision.
  • Relies on data specifically trained by the model; lacks real-time web search capability.

Smart Contract Links: N/A

Disclosure: This project was initially developed as a prototype for the TRON Hackathon S5. Post-hackathon, our team plans to seek official permissions from TRON DAO and associated ecosystem projects. Our goal is to utilize their data and documentation under mutually agreed terms to enhance our model training for the upcoming beta launch. Feedback from the community is highly appreciated.

20 Likes

More like customer support right ? Will these be able to help me on current issues? Or just the basis

3 Likes

Welcome to Grand Hackathon Season5
How would you convince users that all information provided in your platform is accurate? How would you handle a situation were information provided in your platform drained users hard earned money?

3 Likes

Welcome to Season 5.
I’m wowed by the conversational AI assistant.
But more concerned about how this would be constantly upgraded to be consistent with updated results.
How does NaviTron plan to ensure data accuracy and relevance?

2 Likes

I believe that NaviTron has the potential to be a valuable tool for TRON users of all levels of experience.

3 Likes

You are training an AI that will help people to understand tron dapps better?
I see Tron shark in the list.
Please enlighten us about that one :pray:t2:
Why have you chosen to include it?

You say having a curated list of dapps. Could you tell us curated by who? Thanks

3 Likes

I guess most will understand why I’m asking this after reading that :point_down:

It shows the importance of being very careful and make proper research before talking about a dapp. You can potentially promote a scam and harm your project but also Tron users and Tron’s image.

5 Likes

I must be honest, reading about a “curated” list of dApps sounds like propaganda. Curated by who? On which metrics?

Seeing scam dApps in your list should require an immediate disqualification by @admin.hackathon IMHO

3 Likes

Hello community members! Thank you all for your very thoughtful comments. Here are our first round of responses:

@Prince-Onscolo Our project offers a feature more like being able to talk to various ecosystem docs in a comprehensive manner. We just updated our project post and embedded our demo video for your viewing.

@Nweke-nature1.com We have put a lot of time into making sure the model doesn’t simply create answers out of thin air like other models. Unfortunately there is no way to guarantee to users that all the information is accurate, other than the fact that the source is strictly from official documentation from existing projects.

@manfred_jr We have worker that fetches the latest data from a set of whitelisted docs and whitepapers from popular projects in the TRON ecosystem. We plan to manually add new projects into our AI model until we can identify a good way to vet projects in a more scalable manner.

@Okorie Thank you! Please try it out and leave any feedback:)

@fabsltsa Thank you for pointing out your concern. We initially based our model on projects that were either listed on the official ecosystem page (TRON Ecosystem | TRONDAO) or had high 30 day active user base on dapprader. Our team has gone ahead and removed deprecated or scam projects from our model entirely. We appreciate the feedback and we will continue to iterate our model.

@Sirluke Thank you for your concern. We believe there is a fine line that we are still working towards in determining the list of legitimate dapps to be included in our model. This is because if we include all dapps (as we originally intended to) this will include dapps, as you’ve rightly pointed out, that may promote scam-like products. For the short-term we will continue to take the community’s feedback and improve our models to include only legitimate dapps in the ecosystem.

3 Likes

Don’t trust metrics like volume and UAW. Those kind of people also go in villages organising big events offering free food and explaining that people will earn 1%-2%-3% daily ROI. Hundreds, thousands of people join believing that non sense and boost the smart contracts metrics.
There are many red flags that you won’t see in metrics. And to find those red flags require real researches.

Btw transit swap is twice in your list.

2 Likes

Oh wow, way to go team, incorporating a worker to fetch the latest data from whitelisted docs and whitepapers is a great initiative, but could you share more details about the worker and how it fetches data from whitelisted docs and whitepapers, I’d like to know.

1 Like

oh OK I have watched your video and I have gotten the whole idea. @fabsltsa has putting some advice across, kindly take note. Get closer to the tron community and know the dapps they use to avoid fake metrics and others. Thank you

1 Like

Thanks for clarifying that your platform doesn’t give a financial advice

1 Like

It looks like a project that will appeal to a lot of people. But I couldn’t see your smart contract links, could you share them please?

2 Likes

+1 for this. Because use of tron or bttc is required to check

2 Likes