OtoGPT: a lightweight, fully autonomous (and unstable) gpt agent that can talk to you, run shell commands, read files, write files, think and browse the web!
Credit: This is a forked variation of https://github.com/dennishansen/iga.
To use Oto:
- create an .env file with your OpenAI API key (or paste in string like so: openai.api_key = "your api key")
- install the dependencies with pip, and run python main.py.
-
Oto should be run in his own directory to allow him to update his source code and system instructions. When working with Oto, encourage him to read his own files first, test and validate his work, and ask for help only when needed.
-
Oto can provide clarifying questions and review your work to assist you.
Chat with Oto via CLI, and she will perform one of the following actions:
- Talk to you: This will allow you to respond back
- Run a shell command: These run in the current directory. The output is returned to Oto.
- Read files: Oto can read the contents of specified files.
- Write files: Oto can create or overwrite files with new content.
- Think: This simply triggers another action.
- Browse: Browse / research using the web and save findings.
Run Oto in her own directory to enable Oto to update her own source code and system instructions. How far can Oto go with a little guidance? Let's find out.
Create .env file and add OPENAI_API_KEY=<your-api-key>
Instal dependencies pip install -r requirements.txt
Run python main.py
- Start by having her read her own files before she making changes
- Encourage Oto to test & validate her work
- Encourage her to keep trying and only ask you if she's really struggling