Some DBs/frameworks/technologies wants me to never do software engineering again, neo4j is one of these.

I had high expectations on your GraphRAG capabilities after hearing talks from your CTO but it feels like you just followed the hype duct taping embeddings as a metadata field in your existing Graph DB. This means I have to implement all benefits of having a GraphRAG in my own backend, which leaves little to no benefits to having a Postgres and separate Pinecone instance.

The setup experience was very bad. In ex. Pinecone, Weaviate, and Postgres, I get started with first data uploaded and fetched within 10 minutes. With Neo4j, it took ages to get the instance provisioned, and there's seemingly no possibility to have a name, only Instance01 (not very descriptive). THe instance has to be connected to all basic tools in the same environment (not very intuitive and I would expect it to be connected to tools allowing me to use it automatically at provisioning). The instance is automatically "paused" making it completely unusable, and takes ages to "resume".

I want this to work so bad as it's great in theory, I guess execution unfortunately just isn't there. Will definitely go to Microsoft GraphRAG instead.

(Please forgive me if it's all my mistake, would be happy to be proven wrong and see the vision communicated through podcasts etc on how your solutions can improve LLM agent performance)

Hi Johan Torssell could you give some more information here?