https://github.com/jarrodwatts/blaickrock
A fully autonomous asset manager on Abstract. Returns may vary.
https://github.com/jarrodwatts/blaickrock
agentic-ai ai vercel-ai-sdk
Last synced: 2 months ago
JSON representation
A fully autonomous asset manager on Abstract. Returns may vary.
- Host: GitHub
- URL: https://github.com/jarrodwatts/blaickrock
- Owner: jarrodwatts
- Created: 2025-05-06T04:42:26.000Z (6 months ago)
- Default Branch: main
- Last Pushed: 2025-07-22T10:37:36.000Z (3 months ago)
- Last Synced: 2025-07-22T12:26:14.106Z (3 months ago)
- Topics: agentic-ai, ai, vercel-ai-sdk
- Language: TypeScript
- Homepage: https://x.com/blaickrock
- Size: 348 KB
- Stars: 4
- Watchers: 2
- Forks: 4
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
Blaickrock
An AI Agent managing a portfolio of crypto assets on Abstract.
## How it works
Blaickrock operates a "fund" of assets on Abstract through it's [Abstract Global Wallet](https://portal.abs.xyz/profile/0x482B6f266df2B8C4790b520348EDC5Ca8C7b387B).
This codebase uses three agents to perform a three step process:
1. **Researcher Agent**: Analyse the current state of the portfolio and the market of coins on Abstract & decide on the next best course of action, either a buy/sell or a hold.
2. **Executor Agent**: Execute the trade via Uniswap.
3. **Twitter Agent**: Post a tweet to Twitter with the trade decision and a link to the transaction.
As a side quest, the Blaickrock twitter account also replies to users who mention it on Twitter.
### Key Code Files
The code is a little messy, but here are the key files:
**Prompts**:
- [executorPrompt.ts](./src/prompt/prompts.ts)
- [prompt.ts](./src/prompt/executorPrompts.ts)
- [twitterPrompt.ts](./src/prompt/twitterPrompts.ts) & [tweetReplyPrompts.ts](./src/prompt/tweetReplyPrompts.ts)
**Tools**:
Only one tool is actually used, despite many being available (originally more were going to be used).
- [execute-swap.ts](./src/tools/execute-swap.ts): Executes a swap on Uniswap.
## Running Locally
1. Install dependencies:
```bash
pnpm install
```
2. Create a `.env` file in the root of the project and add environment variables:
See [.env.example](.env.example) for the required variables.
Please be careful. These are sensitive values. Only proceed if you understand the risks.
3. Run the main three-step flow:
```bash
pnpm run start
```
4. Run the reply logic:
```bash
pnpm run check-mentions
```