semwebrag

• This is our own approach of a GraphRAG pipeline, based on knowledge graphs. *

---

Launching the projet with graphs already built

You only need to use this two commands :

python install.py
python serverLauncher.py

You should see in http://localhost:8000/interface.html an interface in wich you can select your domain and query the GraphRAG !

Build your own graph :

You will need to update at least these lines in the file config.ini :

• domain : folder in wich you want to store the indexes and the graph.
• text_file_path : path to your textual datafile.
• ontology_path : path to your ontology (.tll)

You can then use the build_knowledge_graph_aligned_with_ontology function in mainPipeline.py (there is a commented exemple at the end of the file)

Then the function process_query will generate a prompt to give to an LLM in a file named query_enrichie.txt