https://huggingface.co/sentence-transformers/all-MiniLM-L6-v2
https://python.langchain.com/docs/integrations/vectorstores/pgvector
https://tembo.io/blog/pgvector-and-embedding-solutions-with-postgres
https://www.psycopg.org/docs/usage.html
https://llama-cpp-python.readthedocs.io/en/latest/server/
Test a single node
pf flow test --flow . --node create_embeddings
Test entire flow with input
pf flow test --flow . --inputs question='Can I open an account from Bulgaria ?'
Create a new batch run
pf run create --flow . --data batch_data.jsonl --stream --name third_run --column-mapping question='${data.question}'
pf connection create -f openai_connection.yaml
Run Evaluate Flow
pf run create --flow evaluation --data batch_data.jsonl --column-mapping groundtruth_link='${data.link}' --column-mapping question='${data.question}' prediction='${run.outputs.answer}' --run first_run --stream --name first_run_eval
pip freeze > requirements.txt
CMAKE_ARGS="-DLLAMA_CUBLAS=on" FORCE_CMAKE=1 pip install ninja llama-cpp-python --force-reinstall --no-cache-dir