A tool for analyzing narrative stories using OpenAI APIs, Pinecone Vectore database, and Langchain RAG pipeline to retrieve information from long text media.
Running this program will take the text contents of a book (or other narrative based text media) and produce a graph datastructure of the characters in the story and summaries of their interactions. This analysis tool could be used to visualize sumaries of character dynamics over time when using a graph database such as Neo4j.
To run the program, you must first set the environment variables for the APIs used and the asset files to be analyzed.
| Env Var | Value Description |
|---|---|
| OPENAI_API_KEY | OpenAI API key for accessing the Embedding and Chat APIs |
| LANGCHAIN_API_KEY | Langchain API key for RAG retrieval |
| PINECONE_API_KEY | API key for Pinecone vector datastore |
| ASSET_FILE_PATH | File path to the text to be analyzed |
| ASSET_TITLE | Name of the work to be analyzed (for metadata) |
Currently the parser assumes that the file is preprocessed into the format of each chapter starting with a line matching regex /CHAPTER [IVX\d]*./.
$ yarn run build && yarn run start