Weaviate

Weaviate

Weaviate is an open-source vector database for AI-native applications. Store and query embeddings, perform similarity search, build recommendation systems.

Data Model

• Collections — schema-defined buckets; inventory them and retire unused ones.
• Objects & retrieval — CRUD plus semantic and hybrid (keyword + vector) queries; point reads need collection + id.
• Schema graph — detailed class definitions, cross-references, and property usage stats.
• Quality — missing embeddings, duplicate detection, property validation, vector health, and consolidated reports.
• Analysis — collection profiling, distribution studies across properties and vectors.

Two credential field setups

Operational order (unfamiliar cluster)

1. List collections — inventory collection names before any scoped call.
2. Inspect one collection — read detailed schema (properties, vectorizer, cross-references).
3. List objects, search, or hybrid search — pass collection on every call unless you saved a default on the credential.

Gotchas

• Cluster setup: Each credential binds to one Weaviate deployment. Set default_collection on the credential when one collection is used repeatedly; otherwise list collections and pass the name on each call.
• Instance URL: Use the full Weaviate Cloud hostname (https://….weaviate.cloud). Host-only URLs work but the scheme helps validation surface clearer errors.
• Localhost: Use the Local Weaviate on this computer setup — not a public URL field with localhost in it.
• Vectorization: Collections with text2vec-openai need the optional OpenAI API key on the credential.