Across AI Assistant Proposal

Ideas & Feedback
1

Title: Across AI Assistant Proposal
Author(s): pennypanda., neondaemon
Status: RFC

Summary

Introducing Appy*, a cutting-edge AI-powered assistant, designed for the Across Protocol community. Appy serves as a versatile Chat-based Assistant, streamlining the experience for both newcomers and seasoned experts in the Across community, trained on Across Documentation and live chats. From development support to general web3 inquiries, Appy is poised to become the go-to support system for all Across users.

*we intend to run a competition to name the AI assistant

Motivation

Appy is tailored to offer intuitive assistance to community members, developers, investors, partner projects and traders. By simply engaging in a chat, users can get precise answers to their queries.

Accessible as a Discord bot, Appy stands out for its ease of use in various settings. In Discord communities, particularly within DAOs, the bot integrates seamlessly, transforming into a dynamic knowledge base about Across. This integration enhances community engagement by streamlining access to DAO documentation and promptly addressing common questions.

Specification & Implementation: Describe the proposal in as much detail as necessary. Explain the vision for the proposal. How will this affect the protocol both technically, socially, financially (if applicable), and governance-wise? What steps need to be taken to implement this proposal?

This proposal aims to develop Appy into a robust Discord bot, customizable to meet the specific information needs of the Across Discord Server.The goal is to scale Appy to cater to a wide range of Across users, making information readily accessible.

Objectives

Appy, as a sophisticated Chatbot, aims to support the following user groups and objectives:

  • Community members: Guiding those curious about Across and the broader web3 world.
  • Developers & Partner projects: Assisting both novice and advanced devs and potential partner projects in navigating the Across Protocol ecosystem.
  • Investors and Traders: Offering insights into trading and investments, portfolio management, and related token data analytics.

Primary objectives include:

  1. Accessibility: Seamless platform integration through wallet connectivity.
  2. Personalization: Customized assistance and services based on user expertise and preferences.
  3. Information Hub: Comprehensive insights about Across Protocol, and other DAOs and Communities in the Web3 space.
  4. Data Analytics: Access to pertinent blockchain data

Credit System

A credit-based system will enable access to Appy. Users start with an initial credit pool that replenishes every 24 hours. Additional credits can be gained through community engagement and tasks. This system is being implemented as a rate-limit to protect the cost and ease of operations for the bot, and can be updated/changed.

Across-Special Features in Appy

  1. Learning and Knowledge Management

  2. Daily Learning Cycles: Appy will be able to analyze and learn from messages within defined channels every 24 hours, especially focusing on those written by members with reputable roles, such as Risk Labs and CEC Committee members.

  3. Selective Memory: By focusing on communications from specified roles like Admin, RL, CEC, Dev Support, and Bridge Guardians, Appy will be able to ensure it bases its knowledge on the most relevant and accurate information.

  4. Bridge Transaction Updates: Members can ask about their bridge transactions and receive status updates and provide information about busy routes and possible delays, see LP positions.

  5. Feedback and Continuous Improvement

  6. Interactive Feedback System: The CEC committee will have the authority to control Appy’s responses. A dedicated channel for bot response logs will be created, where CEC and Admin members can review Appy’s responses. These role holding members can then cast vote with thumbs up or down reactions and provide specific feedback on improvements or corrections, ensuring Appy’s continuous learning and adaptation, while also allowing for human moderation on the bot’s actions.

  7. Integration with Notion: To centralize access to frequently asked questions and their answers, Appy will store summarized interactions and learnings in a Notion page. This feature will serve as a living document, evolving with the community’s needs and Appy’s growing knowledge base.

Some More Enhanced Features Specific to Across Protocol

  • Docs Explainer: A starter pack offering a simplified breakdown of Across documentation.
  • Migration Help: Help for migrating from Across v2 to v3.
  • Across Relay Assistance for devs: Guidance in writing, deploying, and understanding how to become a Relayer and other related coding assistance.
  • Transaction Decoding: Detailed breakdown of transaction data, explaining each component of an on-chain transaction.
  • On-Chain Interactions: Guidance on where to find data about transactions, LP positions, fees etc.

Team

Ishan Pandey aka PennyPanda - Product Lead & Manager representing Ownos Studios, an anon product studio (being the reason why other builders’ names are not revealed here)

Milestones

  1. Scoping and Technical Design - 1 Week - $1000
  2. Initial Development of Discord Bot MVP - 2 Weeks - $4000
  3. Final Iteration of the Discord Bot - 3 Weeks - $2000

Fund Request

Total Grant Request: 7000 USD in ACX tokens

Development and Testing: $5000

1 year operating + support: $2000

The budget will be allocated towards development, design, infrastructure costs (including OpenAI APIs, hosting, etc.), with any savings directed towards post-development and marketing efforts.

We are also open to letting the Across team have access to the complete source code once development is completed. This will allow them to review and test it for security purposes before the bot is made live on the server.

After a 1 year period, the Across team can exercise their option to purchase the entire product from the Ownos team and continue running it as per their wishes without our involvement and retainer.

Rationale

The goal for developing Appy is to create a comprehensive, user-friendly, and adaptive AI assistant for the Across community. It aims to raise the standard of user support and serve as an ever-growing knowledge base for the community.

Building Appy as the “home-bot” for our Across server in such a manner will ensure that members are users are able to receive quick support and answers. It will significantly reduce the time and effort required from dev support in answering questions, many of which are very simple and often repeated.

It will also be continuously self-updating, incorporating all new information and developments without manual intervention, and still leave room for the admins to be able to regulate it.

We believe that it serves as a good-to-have automation within the Across Community, which will definitely save time and energy and increase the quality of experience for our members, in particular:

  1. Focused Learning from Authoritative Sources: Prioritizing learning from reputable role members and specific channels ensures that Appy bases its knowledge on the most accurate and reliable information available. This method is preferred over a broader, less selective learning approach because it minimizes the risk of disseminating incorrect or outdated information, which is crucial in the rapidly evolving blockchain and web3 sectors.

  2. Daily Learning Cycles and Selective Memory: Implementing daily updates rather than real-time learning balances the need for up-to-date information with the practical considerations of processing time and computational resources. This strategy ensures Appy’s responses are both current and reflective of the latest community insights without overwhelming the system or causing unnecessary delays in response times.

  3. Interactive Feedback System: Providing a mechanism for continuous feedback from the community, especially from those in authoritative positions, allows for real-time quality control and adaptability of Appy’s responses. This feature is essential for maintaining trust and accuracy in the assistant’s interactions, offering a significant improvement over static AI models that cannot evolve with their user base’s needs.

  4. Integration with Notion and Transition Plan for Future Operations: These elements of the proposal are designed to ensure sustainability and adaptability of the assistant over time. By creating a centralized knowledge base and planning for future operational transitions, the proposal ensures that Appy can continue to serve the community effectively, even as the underlying technology and ecosystem evolve.

Downsides

The development and maintenance of such an assistant entail significant resource investment. This includes the upfront costs for development and ongoing expenses for operations and updates.

There is a slight risk of the community becoming overly reliant on Appy for information, which might deter individual research and due diligence. And despite the focused learning and continuous feedback mechanisms, there is always a risk that Appy could disseminate incorrect information due to rapidly changing conditions or misunderstandings in its learning process.

Monitoring will be required to mitigate this risk. Even with these possible challenges, our commitment to overcoming them reinforces our belief in Appy’s value as an innovative, time-saving tool that enhances the Across community.

Next Steps

This is a Request for Comment. We welcome suggestions and questions. Once we have gathered feedback we will update the document and put up an official proposal. Thank you.

2 Likes
2

Good plan! Allows for immediate responses for users without waits while being less labor intensive for Across.

3 Likes
3

I love the idea and what *Appy can do but i think it should be called Ross as this is already part of the lore! :robot:

3 Likes
4

Thanks for putting forward the proposal for the Appy AI assistant for the Across Protocol community. It’s an interesting idea, but I have a couple of things I’d like to clear up

  1. The $7000 USD budget mentioned for development and operating costs sounds like a lot but I understand buildingng takes time and tremendous effort as well , but I’m wondering if this covers long-term expenses or just gets things started. Also, you mentioned the option to buy the product after a year. Any idea how much that might cost? Knowing this would help us see the bigger financial picture and I think it’s important to know that too

2.I’m curious about what happens after the initial development phase. Will there be ongoing maintenance and updates on information that comes with the package? , or will it be the responsibility of Across?
Trying to Understand the long-term plan and roadmap for keeping the AI assistant running smoothly would give us a better idea of its sustainability.

I’m interested in a simple User POV demo as well as how the backend would be, not the techy parts, just something practical enough for understandability on that end.

If you could shed some light on these points, that would be great.
Thank youuu. :hearts:

1 Like
5

Hey Mide, thanks for your lovely questions!

I’ll actually use this response to answer your questions and also add a user-story so people can understand how Appy would work for them, from a consumer perspective.

Firstly, answers to the questions raised:

  1. The $7000 budget, as mentioned in the proposal, covers both development and operating costs for the whole year. We’re expecting to cover everything to run for an entire period of a year starting from the launch of the bot within Across server.
    This also includes our testing, infrastructure and maintenance, again, for the entire year.

  2. Adding on from Question 1, yes we will continue to provide support and maintenance and updates as required, for the said one year.

As for the question of buying the product after a year, we would be happy to discuss a suitable price at a later time.

Allow me to also share some hypothetical scenarios that would make it easier for everyone to understand how Appy would work:

  1. Imagine Joshua, an active user of the Across bridge, faces delays with his bridge transaction. Concerned, he enters the Discord server, summons Appy, and inputs his transaction ID. Appy quickly responds, identifying the congestion in the specific route Joshua’s transaction is using. It then provides an estimated resolution time based on current network conditions, reassuring Joshua with precise and actionable information.

  2. Alex is looking to use Across for the first time, is curious about the transaction fees. He asks Appy for details. Appy breaks down the fee structure for Alex, explaining the different components like gas fees, relay fees, etc. It even offers tips on how to minimize fees and the best times to transact to avoid higher gas prices, making Alex feel more confident in planning his transactions.

  3. Elena is a dedicated member but often finds it challenging to stay up-to-date with the fast-paced discussions and updates happening in the Discord server due to her busy schedule. Elena activates Appy and requests a summary of the day’s most important discussions and any new updates announced by the Across team. Appy quickly finds an answer through the day’s chat logs, highlighting key conversations, summarizing user concerns, and pinpointing solutions provided. Additionally, Appy informs Elena about a recent protocol update that introduces a new liquidity pool feature, offering a simplified explanation of how it benefits users and enhances transaction efficiency.

I hope this clears some doubts for everyone, thanks again Mide :blush:

1 Like
6

Another point that I missed from Mide, as mentioned in the proposal, the bot is “self-learning” and updates its knowledge from the chats that actually go on in designated channels. Also the CEC and other admin members will have full access to the chat logs and will be able to provide feedback to the bot in order to ensure that no misinformation is learned and given out to the users. We will also be updating the bot’s knowledge base manually whenever new documentation is created.

Also, thanks to @PVMihalache and @gmsteele for your comments as well :blush::heartpulse:

1 Like
7

Most interesting.

I wonder if you have a comparison for a human/humans equivalent service? Ie could we get 24/7 coverage for $583 a month with comparible service?

8

Hey @berry4144

I would say it would be highly improbable to set up an equivalent human run support that works 24/7 and keeps up-to-date on the daily like *Appy would.
I hadn’t made any comparison but here’s my reasoning.

Let’s assume $7.25/hr as the minimum wage (that’s what I got from a quick google search about min wage in the US), that’s quite cheap and low.
$7.25 times 24 = $174, that’s per day!

From the other side, $583 for a month comes out to $0.7 an hour (538 dollars/730 hours) for the support.
Also, assuming that one person alone of course cannot do this 24/7 we’d atleast need 2 or 3 people working in shifts daily.

But the above approach was purely mathematical, it’s possible that an arrangement be made with individuals for a set amount and period, I still don’t think it would be viable to get an equivalent service practically speaking.

Happy to discuss more on this angle if you like :blush:

9

I agree with PVM that we should only do this if the bot’s name is ROSS!

ROSS is already in the lore and we’ve been meaning to use him in this way, just haven’t had time/resources. Please change the name ^^

1 Like
10

Yes for Ross! :robot: And we already know how he looks and that he likes to ride with rangers on a motorbike :motorcycle: :smirk:

1 Like