This is a software engineer consultant that given a description of an idea will create a plan and then attempt to implement and test it.

1. Clone the repo
2. poetry shell # activates virtual environment
3. poetry install # install's dependencies
4. cd isolation_env/python
5. docker build -t python_isolation # create the docker image for the shell the assistant is allowed to use
6. cd ../..
7. echo ANTHROPIC_API_KEY=..." > .env # set the environment variable for the anthropic api key

python chat.py ${project_directory}

This will start the chatbot to work on the project in project_directory

The bot frequently forgets the rules for writing source files. It should emit a markdown block and then the write_file command but sometimes it emits the write_file command first. Usually it will figure out its mistake after a few surprises.

• agent/llm.py - Interacts with anthropic and provides the tool execution
• agent/detector.py - Parsers that find the tool invocations in the LLM output
• agent/shell.py - Provides the isolated execution environment to the agent (via docker)
• consultant.prompt - The prompt that creates the consultant behavior and describes the tool use