ArangoDB

ArangoDB is a multi-model NoSQL database that keeps graph, document, vector and search data in one engine, self-hosted or managed by Arango in the cloud.

ArangoDB gives an application one database for connected data, where a team would otherwise run a graph database, a vector store and a search cluster side by side. The same record can be a document, a graph node, a vector and a searchable entry at once, so no sync pipeline has to keep copies in step. Arango, a San Francisco company, makes it and builds its larger Arango Contextual Data Platform on top of it. Queries are written in AQL, the ArangoDB Query Language, which combines point lookups, graph traversals, joins and ranked full-text search in one statement. AQLizer turns a plain-language question into an AQL query the developer can still read and edit. ArangoSearch handles text search with BM25 and TF-IDF ranking, and vector indexes find documents by the similarity of their embeddings. Stream Transactions group writes across several documents, and TTL indexes clear expired records in the background. In a cluster, Coordinators and DBServers scale separately, SmartGraphs keep connected data on one machine, and SatelliteCollections copy reference data to every node. A managed GraphML service based on GraphSAGE and an analytics engine for PageRank and community detection write their results back to the graph. ArangoDB suits developers and data teams whose applications depend on the relationships between records, such as knowledge graphs or retrieval for language models. Software vendors use it as the database inside their own products. Regulated organizations can run it on premises or in air-gapped environments with the same Kubernetes Operator and binaries as the cloud service. It is deployed self-managed on Kubernetes, virtual machines or bare metal, run by Arango on AWS or Google Cloud through the Arango Managed Platform, or embedded under OEM terms. A common job is grounding an AI assistant: the team loads documents and their links, then answers each question with a traversal, a vector lookup or both in one query. Operations teams build AIOps and ITOps co-pilots on the same engine. The Kubernetes Operator handles provisioning, zero-downtime upgrades, scaling and failover, while Kafka connectors feed events in from edge systems. When a team later wants automatic knowledge graph construction and per-question retrieval routing, it adds the Contextual Data Platform on the same data rather than moving to a new system.