BlockAGI is a hackable research agent that is inspired by AutoGPT. It can accomplish any research-based task that you give it. If you have some reports, articles, papers to generate then you can assign BlockAGI to do it on your behalf. In the output, it generates a beautified document with proper headings and citations that you can use or publish anywhere.
It basically has access to the internet and it scrapes through the web search results, online blogs, and other websites to fetch the relevant data for the target task and then generates the content after removing and filtering the unnecessary data. It can perform multiple iterations to generate a comprehensive report and show that to you in the end.
BlockAGI is completely open source and that’s why you can run it locally on your machine. In this article, I will show you how to set it up and then run it. It can literally do research for you on any topic and the results are very eye catchy. It includes all the details and cites the source of information and that’s what makes it a decent free AGI tool.
What is AGI?
AGI is acronym for Artificial General Intelligence and hypothetical type of AI technology that would have the ability to understand or learn any intellectual task in the same way a human being can. AGI is still a theoretical concept, but it has the potential to revolutionize many aspects of our lives by saving a lot of time and effort.
Some of the potential applications of AGI include self-driving cars, medical diagnosis, financial trading, and scientific research. And BlockAGI is one such tool you can use experience what AGI in reality is and what it can do.
Installing and Setting up BlockAGI:
Here I am using a Linux PC to setup BlockAGI but you can also use Windows if you want to. The very first thing you need to install and run BlockAGI is Poetry and Python. You can download and install Python from here and Poetry from here.
Now, start by cloning the GitHub repository of BlockAGI and cd into it.
git clone https://github.com/blockpipe/BlockAGI
cd BlockAGI
Install the dependencies by running the poetry command like this.
poetry install
Initiate Playwright by running the following command.
poetry run playwright install
Now, at this point, you have successfully setup BlockAGI. But before running it, you have set some input parameters. Also, if you face some errors in the above steps, then you need to repeat them again or use a virtual environment in case of package conflict.
Setting Input Parameters in BlokcAGI to Start Scientific Research:
To make BlockAGI start the research, you need to give it a topic, subtopics, and then OpenAI API key.
Get the API key from your OpenAI account. And prepare a list of topics.
Now, you need to edit the .env.example file in the repository. So, open this file with any editor and then set the API key at top and edit the topics. Rename this file after saving the changes from .env.example to just .env.
You need to set the agent role depending on the topic of the research. Also, specify the title for the objectives. Here I am just doing small research, so, I have only specified 3 objectives but you can add more if you like. Also, specify the number of iterations to have a more accurate results and then final report.
Save the changes and now it is time to start it and let the magic happen
Run BlockAGI to Start the Scientific Reseach using AI:
Without wasting any more time, issue the following command and sit back relax.
poetry run python main.py
It will automatically open the web browser where you can see the magic happen right before your eyes. You can see the LLM logs to see the sites it is visiting and extracting the data. At the same time, it is generating prompt and getting the responses from OpenAI to build your report.
In the right side, you can see the report that is being generated. It adds all the objectives as topics with their full answers.
It will take some time and you just need to be a little patient. When all the iterations have been completed, you can see the final report in the browser. Not only this but you can see the report it generated in every iteration. This is in Markdown but you can easily use some Markdown to PDF converters to save this report.
Keep using and running BlockAGI in this way. Just edit the .env file with topics/objectives and run it. It will do everything for you in a few minutes. More objectives can take more time and that’s understandable. But in the end, you will have a comprehensive well researched document that you can publish or share with others.
Note: BlockAGI runs a lot of prompts and there is so much text generation involved which may or may not be included in the final report, and hence the API usage is high. For the sample research topics that I used in this post, it consumed about 100K tokens, costing around $0.35. So, you also need to keep your OpenAI budget in mind. For more complex researches, the OpenAI cost will increase exponentially.
In brief:
BlockAGI can be your personal research assistant for any topic. From writing a detailed blog post to a scientific research, it can do it all. It gets the data from OpenAI models as well as the public internet. So, you get the best of both worlds. The final document that it generates has citations and since it performs many iterations to gather and generate the data, the changes of ending up with fake or irrelevant information are considerably less.
