BlockAGI

BlockAGI is an open-source, self-hostable, hackable research-oriented agent inspired by AutoGPT, with the project address at https://github.com/orgexyz/BlockAGI。它使用 Built with Python 3, it relies on the capabilities of LangChain and OpenAI. BlockAGI performs iterative research for a specific domain, initially designed to answer cryptocurrency research topics (hence "Block" in the name), and can also be used for a variety of other use cases.

Main capabilities: automated research, where the user only needs to provide a topic and BlockAGI can conduct research on its own; comprehensiveness, being able to search, collect, refine, and evaluate information on its own; access to real-time data from the internet or the user's own database; output of detailed narrative research reports; real-time display through a web interface of the agent's decision-making process, execution progress, and latest research results, enabling interactive monitoring. The project emphasizes being 100% hackable, with concise LangChain-based code that is easy to adjust as needed; it emphasizes privacy, with reports kept only locally with the user and with the LLM provider trusted by the user.

Differences from AutoGPT: BlockAGI has been tested to run well on gpt-3.5-turbo-16k, and is much cheaper than gpt-4 for most research tasks; it provides an interactive web UI that displays the agent's decisions, execution progress, and latest results; its functionality focuses on assisting the user's research topic, rather than trying to complete all tasks; setup is simplified, and it does not require file reading and writing, so no Docker/sandbox is needed; it does not require an external vector data store, reducing complexity and resource requirements.

Tech stack: the backend is Python and LangChain, and the frontend is Next.js and Tailwind.

Installation and basic usage:

1. Install Poetry (Python dependency management tool).
2. Clone the BlockAGI repository locally.
3. Enter the project directory and run poetry install to install the required dependencies.
4. Run poetry run playwright install to install the Playwright dependencies.

Startup and usage:

1. Copy the .env.example file and rename it to .env.
2. Open the .env file, add your OpenAI API key, and you can also modify other configurations as needed, such as the research domain or additional settings.
3. Run poetry run python main.py Starting BlockAGI will also start a Web application to facilitate monitoring progress and interacting with the research agent.
4. View BlockAGI's progress through the Web interface.

In terms of documentation, the repository provides installation and basic usage, parameters and their recommended values, BlockAGI architecture, building custom tools, advanced modifications to the research assistant, contributing to BlockAGI, and more. The project encourages contributions and feedback: you can fork the repository, make changes in your own branch, and submit a PR; you can report bugs, suggest new features, or ask questions through GitHub Issues; you can join the Discord community to share experiences and ideas for improvement. For citation, it is recommended to use the CITATION.cff provided by the repository. The project is licensed under the Apache License 2.0, which allows use, modification, and distribution of the software. The project does not accept monetary donations.

Regarding applicable scenarios and limitations: official materials state that it is emphasized for research topics, supports specific domains such as cryptocurrency and can be customized, can access real-time data and output narrative reports; it also states that no Docker/sandbox is required, no external vector data storage is required, and it can run with gpt-3.5-turbo-16k. As for specific pricing, platform compatibility, performance metrics, the accuracy of research results, or details beyond the usage steps, the materials do not provide them.