HN user
Neo4j Hype Translator:
Input Hype: “Neo4j Launches Breakthrough Infinigraph Architecture to Unify Transactional and Operational Workloads”
Output Hype Translation: After over a decade of “clustering”, Neo4j finally figured out sharding, basically fixing the thing their clustering should’ve done from the start.
If you’re doing GraphRAG in the style of Microsoft’s original model , where the graph is inferred dynamically by the LLM and used primarily for retrieval or reasoning , then Neo4j is often overkill in my opinion. Not to mention its performance issues and toxic attitude to open source.
Lightweight, in-memory alternatives like NetworkX or even plain Python dictionaries are typically faster to set up, easier to integrate, and better suited to ephemeral, prompt driven graph reasoning.
Here are some other options : NetworkX
LlamaIndex’s KnowledgeGraphIndex
LangChain with in-memory Python structures, Lets you model entities and relationships as dictionaries or lists, making it easy to traverse or query on demand.
https://docs.llamaindex.ai/en/stable/examples/index_structs/...
https://www.microsoft.com/en-us/research/blog/moving-to-grap...
Neo4j has bet the farm on GraphRag. Unfortunately - I don’t think the return on investment is there even if using Neo4j community edition - and even then - I think there are better options.
Good article though.
Thanks - this one is new to me. Checking it out.
DozerDB: extends Neo4j Community with features such as multi-db. https://dozerdb.org
Memgraph: https://memgraph.com/memgraph-for-neo4j-developers
Apache AGE https://age.apache.org/
And much more….
If you are using community edition - check out the DozerDB plugin which adds enterprise features to Neo4j community such as multi database. Its still in its infancy but has already implemented multi db and enterprise constraints. https://dozerdb.org
Will more information that could touch on what you mentioned be in their S-1 registration if they attempt an IPO? I need to search through all the court documents to see if there are any more details about prior years. Good thing I have LLMs to help with this now.
https://www.courtlistener.com/docket/16272543/neo4j-inc-v-pu... https://www.courtlistener.com/docket/16276158/neo4j-inc-v-gr...
Trial testimony shows that Neo4j is not currently profitable after 17 years.
Is this normal for VC funded software companies nowadays?
---
Q. DO YOU KNOW WHAT YOUR NET PROFITS ARE?
A. WE ARE IN A LOSS POSITION RIGHT NOW.
https://storage.courtlistener.com/recap/gov.uscourts.cand.33...
Seems mostly like hype.
The analysis below seems to show that GraphRag with @Neo4j offers minimal improvement in context retrieval and only slight gains in answer relevancy and faithfulness.
https://aiencoder.substack.com/p/graphrag-analysis-part-1-ho...
Why would the investors of Neo4j take control of Neo4j Inc?
I believe the COO was removed as COO around the time I first heard this.
I hope Neo4j Inc does not get desperate and add terms to their Neo4j Community Edition making it proprietary or non-free like they did with their enterprise edition!
It seems from their repo (need to verify) that they did the right thing and changed the License to their own license name called Dgraph Community License (DCL)
Compare this to Neo4j, who added the commons clause to the AGPL License complete with FSF copyright and preamble. Neo4j kept the License as AGPL and reaped the benefits from open source community.
Neo4j’s approach made people think it was still open source under the AGPL, while dgraph’s approach was proper and clear.
Neo4j’s approach was deceitful in my opinion, and I believe it’s finally coming back to haunt them after some toxic rulings put the GPL structure at risk unless an upcoming appeal overturns it.
Good for dgraph for doing it right in the end and not making people think they were Apache licensed with commons.
I never saw this.
If Dgraph has the commons clause then it is no longer open source or free software (as in libre free).
If they are still calling it open source, they should be warned the courts have ruled that the commons clause means ‘not open source’.
All hype in my opinion, and apparently the #2 committer of the project may agree.
He agrees that @neo4j is slower than a “normal vector db”.
“if you are not dealing with many documents, it works quite well”
“Neo4j might not be optimized for vector search, but that doesn't mean it can't do it.”
Direct link to the report.
https://sfconservancy.org/docs/kuhn_expert-report-in-neo4j_5...
Case about the “further restrictions” removal provision of the AGPLv3 continues after counterclaim filing
This post is several years old, but it shows what their prices were back when neo4j enterprise was available commercial or AGPL.
As soon as end users learned they were getting the same software either way, guess what they picked?
(This is my blog post - iGov Inc is just me)
https://blog.igovsol.com/2018/01/10/Neo4j-Commercial-Prices....
Yes they closed the enterprise source code.
They first tried adding the commons clause, then when that did not accomplish what they wanted, they closed the enterprise source code.
The kicker is that this was not because of behemoth like Amazon or Oracle, it was because of a single individual… (guess who?)
Good read to get a better picture:
https://sfconservancy.org/blog/2022/mar/30/neo4j-v-purethink...
I touched on the negative Glassdoor reviews in this thread. Do you see a common pattern/theme? https://twitter.com/jmsuhy/status/1600093327130431488?s=61&t...
Funny to see people talking about the case.
I thought it was interesting that the 9th circuit made so many mistakes that they had to append the ruling. The fact that they got the defendants wrong is quite strange to me. The Graph Foundation was not part of the case any longer. They joined the case but then settled. The injunction itself that
Neo4j was very sloppy when adding the commons clause into the AGPL. They literally added the commons clause to the AGPL, leaving in the preamble and all, and call it the ‘Sweden License’. (Virtual name)
https://raw.githubusercontent.com/neo4j/neo4j/3.4.18/enterpr...
This ‘sweden license’ literally states it is a ‘free, copyleft license.’
copyleft = opensource according to Philip Rathle, Vice President of Products at Neo4j Inc. (6-ER-1368 ¶¶ 10-11 of appeal except of records).
Therefore according to Neo4j - it is free and open source.
Neo4j says their 'Sweden License' (AGPL + Commons) is both open source and not open source at the same time.
I call this: 'Schrödinger's Sweden License'
Keep an eye on the case - it is now in the second phase of the case where everyone will learn about Neo4j’s behavior. It would make a good Netflix mini series.
There is a very bizarre story behind Neo4j Inc's behavior and what lead to them adding the commons clause to the enterprise AGPL license.
They seem to be ready to go closed source, which they should have just done in the first place instead of behaving like they did. Adding the commons clause is just the tip of the iceberg - there is a very entertaining story behind everything that is happening. A single person, not a huge corporation, is behind this.
I will write a blog post about what happened/is happening in the upcoming months - complete with supporting documents such as FOIA requests, etc.