Key takeaways
- Weaviate is an open-source vector database that stores objects with vectors and supports hybrid keyword and vector search.
- Run it locally with Docker, connect with the Python client, create a collection and import data with vectors.
- Weaviate Cloud is the managed service; the open-source database is free to self-host.
- Choose Elasticsearch when full-text search leads; Weaviate when vector and hybrid retrieval lead.
- Weaviate is developed by Weaviate B.V., a company from the Netherlands that earns from cloud, enterprise and support.
Weaviate is an open-source vector database that stores data objects together with their vector embeddings, and lets you search by meaning, by keywords, or both together in hybrid search, with filters on properties. It can generate embeddings through integrated model modules, which makes building RAG systems quick. I design hybrid retrieval in RAG.NextUpgrad, so this page gives a Weaviate tutorial, how it works and is set up, Weaviate Cloud, cost, a comparison with Elasticsearch, ownership and history, its business model, and outages.
What is a quick Weaviate tutorial?
Run Weaviate locally with Docker, install the Python client, create a collection with a vectoriser or your own vectors, import objects, then query with near-text, BM25 keyword or hybrid search.
import weaviate
client = weaviate.connect_to_local()
docs = client.collections.get("Doc")
docs.data.insert({"title": "Refund policy", "body": "Refunds within 30 days..."})
res = docs.query.hybrid(query="how do refunds work", limit=3, alpha=0.5)
for o in res.objects:
print(o.properties["title"])
client.close()
Create the collection first with a vectoriser configuration, or supply vectors yourself when inserting. The official quickstart walks through both local and cloud setups step by step. Once the basics work, try filters on properties, compare pure vector search with hybrid search on the same questions, and check which returns the passages you expected.
How does Weaviate work?
Weaviate stores objects in collections, each with properties and one or more vectors. It builds an HNSW index for vector search and an inverted index for keyword search and filtering, then combines them for hybrid queries using a weighting parameter called alpha.
Also asked as: how to setup weaviate
Optional modules call embedding models, from providers or local containers, so Weaviate can vectorise data on import and queries at search time. Generative modules can send retrieved results to a language model, giving RAG in one query. It supports multi-tenancy for SaaS apps, replication and sharding for scale, and quantisation to reduce memory. Clients exist for Python, JavaScript, Go and Java. Its GraphQL and REST APIs also let you query without a client library.
How do I set up Weaviate?
The simplest local setup is Docker Compose: generate or copy the compose file from Weaviate's documentation, choose modules, start it, and connect from your client. For production, use Kubernetes with Helm charts or Weaviate Cloud.
Setup checklist:
- Install Docker and create a compose file with the Weaviate image.
- Enable the modules you need, such as a text vectoriser.
- Mount a volume so data persists across restarts.
- Set authentication with API keys before exposing it.
- Configure backups to a storage bucket.
- Monitor memory, since HNSW indexes live largely in RAM.
An embedded option also exists for some clients, useful for quick experiments.
What is Weaviate Cloud?
Weaviate Cloud is the managed version of Weaviate, run by Weaviate the company, offering serverless clusters, dedicated deployments and options to run inside your own cloud account. It handles hosting, scaling, upgrades, backups and monitoring.
It suits teams that want Weaviate's features without operating infrastructure. Typically there is a free sandbox for experiments that expires after a period, and paid tiers for production. Choose a region close to your users and data, check compliance options for sensitive data, and keep your embedding model and source data reproducible so you can migrate if needs change. Test performance from your application's region before going live, since network latency adds to every query.
Is Weaviate paid?
The core Weaviate database is open source under the BSD-3-Clause licence, so self-hosting is free apart from your infrastructure costs. Weaviate Cloud and enterprise support are paid, with a free sandbox for trying it out.
Self-hosting costs include servers with enough memory for your vectors, storage, backups and engineering time for operations. Managed hosting trades those costs for a subscription. Embedding models used through modules may also have their own API costs. Compare total cost of ownership, including your team's time, before deciding between self-hosting and the cloud service. For students and prototypes, a local Docker instance costs nothing beyond your laptop.
Weaviate vs Elasticsearch: which should I use?
Choose Elasticsearch when mature full-text search, aggregations, analytics and logs are central, with vector search as an addition. Choose Weaviate when semantic and hybrid retrieval are the core, and you want built-in vectorisation and RAG-friendly features.
Elasticsearch has deep text analysis, language analysers and a huge ecosystem, and supports dense vector kNN search, but is heavier to operate. Weaviate is vector-first, simpler for embedding workflows, and offers hybrid search with BM25 out of the box. For RAG over documents, both can work; test retrieval quality on your own queries, including exact terms like product codes and paraphrased questions. Also weigh your team's existing skills: an organisation already running Elasticsearch for logs may prefer adding vectors there rather than operating a new system.
Who owns Weaviate?
Weaviate is developed by Weaviate B.V., a venture-backed company based in the Netherlands, co-founded by Bob van Luijt, its chief executive, and Etienne Dilocker. The open-source project is maintained by the company with community contributions.
The company was originally known as SeMI Technologies before adopting the Weaviate name. Its open-source licence means you can use and modify the database freely, while the company offers the managed cloud and enterprise services. As with any vendor, keep your data and embeddings reproducible so you are not locked in if circumstances change.
When was Weaviate founded?
The open-source Weaviate project began around 2016, and the company behind it, originally SeMI Technologies, was founded in 2019. It grew rapidly as vector search became central to AI applications after 2022.
Since then, Weaviate has added hybrid search, generative modules for RAG, multi-tenancy, multiple named vectors, vector quantisation and a serverless cloud. The vector database market is young and changes fast, so check release notes for current features and avoid relying on comparisons written a few years ago, which may describe very different versions.
How does Weaviate make money?
Weaviate earns revenue mainly from Weaviate Cloud, its managed hosting with serverless and dedicated options, from enterprise deployments in customers' own cloud environments, and from support and services for organisations running Weaviate in production.
This open-core model is common for infrastructure companies: the free open-source database drives adoption and community, while businesses pay for convenience, reliability, security features and expert support. For users, it means the open-source version remains usable on its own, while paid options exist if you prefer not to operate it yourself.
Is Weaviate down?
If Weaviate Cloud seems unavailable, check Weaviate's status page for incidents. If you self-host, the problem is in your deployment: check container logs, memory usage, disk space and network access.
Common self-hosted issues include running out of memory as HNSW indexes grow, disk full errors, misconfigured authentication, and modules failing to reach embedding providers. Client errors can also come from version mismatches between client and server. Add health checks, monitoring and alerts, and keep backups so you can restore quickly. For critical applications, consider replication across nodes.
Sources
GenAI engineer and AI product builder with 2+ years shipping production-grade AI systems: RAG pipelines, fine-tuned LLMs, hybrid retrieval and multi-modal apps across vision, speech and OCR, architected end to end from ingestion to deployment. Leads engineering for SCULT INDIA's 14-member team, founded the 500+ member TechVerse Enclave community and has mentored 200+ students. Three hackathon first prizes: Changethon 2025 (IIT Roorkee), Product Genesis at Vividhotsava 2025 (CSJMU) and BYTEBATTLE (MeetKats).
Open to GenAI roles, hackathon judging, mentorship sessions and guest talks at colleges. For campus sessions the inviting institution covers the visit end to end: travel (a First AC train ticket for journeys up to five hours, a flight for anything longer), accommodation, meals, and pickup and drop; the honorarium and workshop fees are negotiable on WhatsApp or on the call, and requests are welcome from institutions that can cover the visit this way. Email pranjulrathour41@gmail.com.
