An experiment in LLM prompt engineering and self-reflection built specifically for novel writing.
Self-reflection loosely inspired by this paper.
Run:
python writer.py "<your story concept here>" --api_key <your api key> --org_key <your org key> --output_dir <path to output directory>
If you prefer, you can export your API key and org key as environment variables using the names:
OPENAI_API_KEY and OPENAI_ORG_KEY.
- Add requirements.txt
- Refactor w LLMchain
- Genre/ 'in the style of author' support: Allows for arbitrary genre input, generates genre/ style context for writing and plot.
- Add a prompt debugger that outputs a file tracing each prompting step with input and output
- The OpenAi API connection will occasionally timeout (read timeout=600). It is possible that this can be improved here, but it appears to be linked to OpenAi servers and may be unavoidable