v0.1.3

• Context: Rawdog now sees all the files/dirs in your cwd + your last git commit by default
• Retry: If Rawdog's script throws an Error, it will look at the error message and try again (maximum 2 times by default).
• Environment: Rawdog still executes in your cwd, but using the Python executable from ~/.rawdog/venv. The default pip packages (matplotlib, numpy and pandas) are pre-loaded. If Rawdog tries to use a new package, it will ask permission to install it there.
• Cost Tracking: The running total session cost is now displayed after every call
• Streaming: When --dry-run is used, the model output (python script) will be streamed, so you can start reviewing it immediately.
• Config: All the litellm settings + dry-run & retries are stored permanently in ~/rawdog/config.yaml. Update your default settings by modifying that file directly. Use cli flags (e.g. --llm-model, --dry-run) to set for session.
• Examples: We've established an example format and written a few dozen examples, which can be used for fine-tuning an LLM on Rawdog. Look out for AbanteAI fine-tuned models soon 👀