llama_index SPARQL Notes 03
Today
I'm down to $0.23 OpenAI API credit, so until I next have $ need to look at things that don't need it. sparql.py doesn't in itself need the OpenAI API, but a SPARQLy version of Wey Gu's Notebook is my target functional E2E test.
I still have NebulaGraph data generated from Wey's Notebook. I can use nebula-python to pull out data from there, RDFLib to build RDF, sparqlwrapper to push to store. The SPARQL needed will be essentially the same as for
sparql.py. Also NebulaGraph <=> RDF utils would be nice to have (may already exist, but I'll pretend I didn't consider that, need to inform myself).
Then -
- have a look around 3 Create VectorStoreIndex for RAG in Notebook
- put the INSERT & SELECT queries inside sparql.py
- ...
MOVE PATH FROM llama_index INSTALLED TO DEV TREE
sudo /usr/local/nebula/scripts/nebula.service start all
Started with :
resp = client.execute_json('MATCH (v:entity) RETURN v')
After converting this to string there were character encoding errors. Should really be investigated properly, but for now I'll just go with some quick & dirty sanitization (Valid RDF literals will probably need something anyhow).
json_str = resp.decode('utf-8') appears to have fixed the encoding issues.