Show HN: TickerPulseBot – Agentic crypto analysis bot for Telegram

1 point

a year ago

Hi! I have been playing with agentic frameworks for a while now and last week I saw OpenServ as they were launching their 'devnet'. I signed up and got access to try the platform. I think they implemented it really well and their SDK is written in ts :) that was a plus.

The idea is that you create a project and define your goal; you select agents and the rest is managed by their System Manager agent. There are already built-in agents, but you can create your own simply through their UI or SDK. I first tried the platform by building a basic image generator agent using FLUX.1-schnell: https://github.com/baturyilmaz/FLUX.1-schnell-Imagine

A few days ago they released 3 bounty tasks and one of them was this telegram bot idea. I've played with it and created a first version that I'm happy with. I first needed to create a team for the project. I chose some agents from the platform and implemented new ones for this projects' specific needs:

1. JSON Parser Agent

2. Telegram Sender Agent (the repository)

3. Perplexity Research Assistant Agent (The best member, fast and up-to-date)

4. DEX Screener Agent

5. Twitter API Agent

6. Copywriter Agent

7. Project Manager Agent

When a new ticker is mentioned on Telegram, the Telegram Sender defines tasks for the team. DEX Screener Agent first goes and collects up-to-date data for the token using DEX Screener API, then saves the results in a JSON file. Then JSON Parser Agent goes and finds the official Twitter account for the token and passes it to Twitter API Agent. Twitter API Agent finds the account information and latest tweets and saves the result in a JSON file. While Twitter API Agent is doing its task, Perplexity Research Assistant agent concurrently searches the web and gets the latest news and information about the token and saves the results in a JSON file. At the end, Copywriter takes all these findings and generates a detailed report in a txt file. Then the Telegram Sender takes the report, formats it and sends it to Telegram.

I implemented this workflow using their SDK and I monitor the agents' activities directly on the platform. You can view the code on GitHub. It still needs improvements, and I will continue working on it.