This folder is designed to process latin medieval accounting book data for the project Aldersbach digital using different LLMs and LLM Provider (Mistral, OpenAI, Anthropic, Gemini).

--- JSONtoRDF.py ---
scripts that iterate a specially designed system prompt over a list of user prompts which are json files representing structured accounting book entries and their context. The prompting is designed to make the respective LLM create very specific structured RDF-XML content out of the individual input json objects and combine them into one single xml-file at the end. A log file will also be created. Parameters and file paths can be set in config.yaml. Make sure to have all requirements installed in python (-> requirements.txt)

--- config.yaml ---
used to set model name, provider, parameters, file paths before running with JSONtoRDF.py

--- data/JSONtoRDF ---
this folder contains all input files necessary for the rdf_encoder scripts.
prompts/systemprompt.txt contains the systemprompt, encoding rules and 5 few shot examples. The entire system prompt has been carefully created and tested. 
plaintextentries_[number].json represent the json objects described above. These files were originally created somewhere else.
The [number] reflects the number of the specific latin accounting book. The one file with "Test" in it is a good test file because it only contains 4 objects.

--- providers/... ---
accompanying scripts for JSONtoRDF.py and add_transactiontype_to_json.py. Each of the scripts define specific code requirements for different proprietary LLM-Providers (OPEN AI, MISTRAL AI, ANTHROPIC, GEMINI). They are used in the main scripts depending on how the config.yaml is set

--- .env ---
contains API KEYS registered to the different LLM providers used
is currently set to my (= Max Vogeltanz) personal keys


--- output/JSONtoRDF ---
folder with subfolders named after the providers used. Subfolders /raw should be used for output file generated by JSONtoRDF.py.
Postprocessing can be applied with postprocess_output.py, transforming files in "raw" to "postprocessed" accordingly.

--- GPTAPI_basic.py ---
Simple standalone script using OPEN AI API to process a simple text prompt

--- ChatGPT_API_Tutorial.ipynb ---
Small jupyter notebook explaining GPTAPI_basic.py