First Place: Real_yakman
Second Place: Jaan
Third Place:didierjey8
Fourth Place: 0xKamal7==


Deadline: December 5, 2024==

Create an autonomous AI agent that controls a wallet on the Avalanche network, routinely checks chain information, and uses that information to drive on-chain transactions optimizing for a specific goal.

Summary:

• Build an AI agent that can understand natural language commands and execute blockchain transactions on the Avalanche network. The agent should interface with a wallet to routinely perform operations based on on-chain data. We highly recommend using __BrianKnows API for LLM Agents__for making transactions.
  
  

  1. __Implement the core AI agent that can: __[Mandatory]

  • Autonomously and routinely scan on-chain data.

  • Understand and process on-chain data.

  • Make a meaningful decision based on said data.

  • Execute blockchain transactions through a connected wallet.

  • Optimize it’s decisions for a singular goal/ metric. This goal/ metric is up to you… Have fun ;)

    • Examples:

      • Portfolio manager to generate the highest PnL based on given risk parameters (LLM should know how to review risk of LPs, contracts, etc), timeframe and starting $/AVAX.

      • Portfolio manager to optimize for the highest yield over a given time, utilizing flashloans, multiple risk params etc.

      • Complete onchain tasks/quests on sites like Galxe, Flipside etc.

        
        

  2. __Create a secure wallet integration system that: __[Mandatory]

  • Manages private keys securely

  • Connects to Avalanche network

  • Handles transaction signing and execution

  • Performs gas estimation and balance checks

  • Maintains transaction history

    
    

  3. __Implement a basic memory system for: __[Optional]

  • Storing user preferences and common addresses

  • Tracking past interactions and transaction patterns

  • Managing conversation state

  • Saving commonly used command patterns.

The completed project should have an autonomous AI agent that manages a wallet that makes decisions towards one central goal. There should be a way to view transaction status/history and document some sort of decision process.

Focus on building a reliable and secure foundation that can be expanded with additional features like multi-chain support, advanced memory systems, or even deploying contracts.

• Resources:

  • Avalanche Documentation

  • BrianKnows API for LLM Agents

  • Biconomy Guide on what an AI Agent for Web3 would look like

  • Crossmint AI Agents https://x.com/crossmint/status/1852017739327565985

    
    

• Recommended experience:

  • Experience building LLM-powered applications and utilizing LLM APIs (OpenAI GPT, Anthropic Claude, etc.)

  • Understanding of prompt engineering and LLM interaction patterns

  • Experience with wallet integration, and interacting with smart contracts

  • Good experience in building autonomous bots

    
    

• Deliverables:

  • A working agent where the team can view the decisions and general thought process from the agent.

  • Test history with the AI Agent executing swaps/transfers from your wallet

  • A GitHub Repo with proper documentation and a fleshed-out Readme.md that will be published as a polished instructions guide on the Academy for everyone to read

  
  

• Extra credit:

  • Video walkthrough of how you created this project.

  • Social media posts documenting your dev work.

  • Additional features like sending a tweet using the on-chain agent (as a value addition)

    
    

• What to submit:

  • Github link to the project

  • The link to your submitted Readme.md guide in Avalanche Academy

    
    

• Bounty Amount:

  • ***$9000 ***for the first-place prize

  • ***$900 ***for second and third place

  • $500 for the first 3 completed submissions

    
    

• Judging Criteria:

  • Quality of code with comments and unit testing.

  • Quality of documentation in the Readme.md

  • Creativity with execution.