This repository is not fully organized yet. We appreciate your understanding.

This repository contains code for fine-tuning a GPT-4o-based AI agent designed for implementing a meme coin trading AI agent and data format exchanged between each our service entity. In our project, the fine-tuned AI agent is designed to be used in two main areas.

First, it takes user-inputted strategies and, based on those strategies, along with insights continuously learned by the AI agent about meme coins and various sectors (e.g., animals, AI agents), selects meme coins for investment from the service database and reviews the investment performance. Second, it crawls platforms like Twitter, communities, and whitepapers to evaluate the narratives surrounding meme coins and analyze public sentiment towards them for indexing purposes.

Lastly, since each fine-tuned model and our service need to exchange data in a mutually understandable format, we are considering implementing an additional model for structured data exchange.

Additionally, over the past three days, we have observed that most users struggle to input meaningful investment strategies in a format that the AI agent can understand. To address this issue, we have added an advisory feature to help users prompt their strategies more effectively.

This repository does not include code or logic for using fine-tuned models.

  .
  ├── pyproject.tml / # repository metadata & dependencies
  ├── core / # include finetuning logic
  | ├── rag / # rag is testing for now
  ├── data / # include data exchanging between service entity
  ├── utils / # declare selenium & manage selenium instance
  | ├── error / # handle error utils
  | ├── formatter / # handle formatter utils
  | ├── image / # handle image utils
  | ├── logger / # handle logger utils
  | ├── writer / # handle writer utils
  | ├── parser / # handle parser utils
  ├── libs / # prompt_advisor
  ├── type / # declare data type
  └── README.md

This repository include private finetuned model id. Meaningful results are unlikely to be achieved with a non-fine-tuned model, so running it is not recommended.

OPENAI_ITEM_SELECT_MODEL_ID
OPENAI_SENTIMENT_ANALYSIS_MODEL_ID
OPENAI_NARRATIVE_EVALUATION_MODEL_ID
OPENAI_API_KEY
SERP_API_KEY
TAVILY_API_KEY