The compatibility between the looseness of schema-less design and the flexibility as a graph database is excellent.

The ability to write JSON-first queries using AQL.

Conveniently includes a rich Web UI.

It is equipped with a full-text search engine that is relatively easy to use. Review collected by and hosted on G2.com.

It seems unavoidable due to performance reasons, but management and operations on collections and databases using AQL are weak.

If there were more capabilities for operations on collections and databases, the possibilities would expand. Review collected by and hosted on G2.com.

I found ArangoDB as I was looking for a graph database that was ready for an enterprise content managment system that could act as a general purpose datastore, that didn't break the bank, and had good reliability. Also important was a managed hosting support for my enterprise client.

The alternatives neo4j quoted me a ridiculous ~$70k per year, with no garuntee that their software would work if they ever went out of buisness which seemed likely with that price point. Other graph solutions werent really scalable or didn't have as useful of a datamodel.

Three years after writing my application for arangodb I am extremely happy with this choice and continue to find new ways to organize information either in real-world messy graphs or in highly scalable key-value stores.

The nice for engineers about using arango is that since the project is open source you can create issues on their database repo and talk with real engineers about solutions, or feature requests to further refine the product rather than some know-nothing sales person. Review collected by and hosted on G2.com.

- I have found one or two troublesome phenomenon with the interplay between supported APIs. For instance one api doesn't have the ability to delete edges / nodes in AQL as here: https://stackoverflow.com/questions/51650374/why-vertex-removing-request-doesnt-remove-linked-edge-with-aql-in-arangodb

- Not much python support. I am using `arangoorm` which is very nice but community supported. Please make this an official feature! I also would find more use with a python `foxx` like server but I understand if thats not reasonable. Review collected by and hosted on G2.com.

Seamlessly handles graph, document, and key-value data in one powerful engine Review collected by and hosted on G2.com.

Steep learning curve and limited community support for complex queries Review collected by and hosted on G2.com.

I like that it is very versatile with its multi model approach, you have the power of document store with graph search, whch essentially allows you to do whatever you do with both document stores and with SQL databases, plus much more thanks to the graph traversal. The AQL language is very intuitive and so quick to learn. Speed is great too. It is easy to use for prototyping and it porvides you with good devops tool for cluster installation on kubernetes. Review collected by and hosted on G2.com.

I think FOXX services are not the best approach for microservices, from a software architecture standpoint. Although documentation is generlaly good, I found that node.js documentation is a bit lacking behind Review collected by and hosted on G2.com.

- Very flexible schema

- Support types of joining

- Integrated web UI is useful

- ArangoSearch View is easy to use and helpful in retrieving data from multiple collections, including exact and fuzzy searching Review collected by and hosted on G2.com.

- Low performance on complex aggregations even when indexing is applied, especially on the big edge collections

- Documents do not describe how the index works clearly

- Graph traversal is not fast enough at times when filtering on edges. I needed to use 2 for loops (traverse 2 times) instead of multi-level traversing for better performance. Review collected by and hosted on G2.com.

I really really enjoy AQL way to perform queries. It's as straightfoward as a JS script, very easy for devs to ramp up, easier to perform joins and complex operations! Review collected by and hosted on G2.com.

It could be somehow more intuitive to perform views/materialized views. Review collected by and hosted on G2.com.

the possibility to work with graphs in a complex way Review collected by and hosted on G2.com.

the limitation of the community edition to 100GB Review collected by and hosted on G2.com.

I like that ArangoDB allows us to start a subcription free to kinda test the databases first. I think the upside to using ArangoDB is that there are plenty of choices. Review collected by and hosted on G2.com.

I cannot say to much about dislike ArangoDB but maybe to many options to choose from but I am not sure that a bad thing. Review collected by and hosted on G2.com.

The web interface is particularly impressive, offering an intuitive dashboard that makes database management straightforward even for those new to NoSQL systems. The AQL query language feels natural, especially if you're coming from SQL, and the documentation is comprehensive and well-organized. Review collected by and hosted on G2.com.

The only downside is once you start using it, you can not stop ! bye bye airtables Review collected by and hosted on G2.com.

ArangoDB offers a performant graph data model that offers some key benefits for almost any kind of data structure—not just graph theory applications. ArangoDB and its graph implementation allows you to think about entities semantically, and as standalone objects. Because edges in the graph are distinct from vertices, you can reason about data in isolation from how you or other teams want to create entity relationships.

IME this is a significantly streamlined approach to development, rather than complex ERD diagrams, foreign key constraints, foreign keys added to all tables, etc. that you'd normally use in an RDBMS/SQL setup. There is significant freedom in avoiding the RDBMS model and shoe-horning different data models beneath the same old SQL syntax.

ArangoDB also is trivial to start up via Docker, has a build-in UI that offers all of the fundamentals, and most of what you'd want outside of more hardcore analysis or data UIs. Implementing any size of cloud instance is trivial, and their datacenter locality is now quite good.

ArangoDB uses a performant K/V store based on RocksDB under the covers, including inverted index search out of the box, transactions, dynamic sharding. ArangoDB covers any database need I've had working with B2B/C SaaS products and services, all with a clean built-in interface and easier/possible to scale unlike many problematic document stores that are now rebranding as vector stores rather than (IMHO) getting their fundamentals fixed. Review collected by and hosted on G2.com.

I really don't have dislikes, but there are a couple of things that they could add in the future to improve:

* When introduced in 2020ish, their cloud service (then called Oasis) had some issues with memory usage and silent lockups of Coordinators. That has since been resolved and they're adding more and more observability and resilience every year.

* They're working on a new replication system, which hopefully can be used as a powerful source for event streaming for auditing, ETL, or other purposes.

Other notes:

* I'm not convinced that ArangoDB should add vector store capability (I'd want a separate instance to separate the load profile anyway). It may help from a marketing perspective with the current state of the ebb and flow of AI buzz, but I'd rather it not come at the cost of existing core functionality.

* While the API offers K/V functionality, performing queries or graph traversals requires writing AQL queries. AQL is very powerful, but there's a learning curve, especially if you're used to SQL. As a SQList, you might at first think that "FOR vertex IN collection; FILTER vertex.valid == true;" is going to perform a table scan and filter, for example. But spending the time will reveal some very powerful functionality not possible with SQL. Review collected by and hosted on G2.com.