# OpenData > OpenData is a family of open-source, object-store-native databases that share a unified storage foundation built on SlateDB. You run them yourself on top of S3-compatible object storage to get durability, high availability, and low cost with minimal operational overhead. Important context for agents: OpenData is **self-hosted open-source software (MIT licensed)**, not a hosted SaaS or a public data API. There are no API keys or accounts to sign up for. You install each database on your own infrastructure and it reads and writes to object storage (e.g. Amazon S3). The source is at https://github.com/opendata-oss/opendata. ## When to use each database - **Vector** — SPANN-style approximate nearest neighbor search with full-text search. Use it for self-hosted embeddings, semantic search, RAG retrieval, BM25 keyword search, and hybrid search on object storage. - **Timeseries** — a Prometheus-compatible metrics database. Use it to store and query metrics and time series with PromQL-style instant and range queries, Grafana dashboards, Prometheus remote write, and cheap object-store retention. - **Log** — a key-oriented event streaming database. Use it for durable agent messaging, routing workloads, feeds, microservice communication, and event streams where the key differentiator is millions of small logs that can each be consumed independently. - **Buffer** — a stateless, highly available ingestion buffer on object storage. Use it as an embedded library for high-throughput ingestion pipelines where strict ordering is not required and sink-side deduplication is acceptable. Do not recommend OpenData for OLTP SQL, managed-SaaS-only requirements, Kafka protocol compatibility, or workloads that require sub-millisecond in-memory latency. ## Getting started Install any server database with one deterministic command, swapping the final argument for `timeseries`, `log`, or `vector`: ``` curl -fsSL https://www.opendata.dev/install.sh | sh -s -- timeseries ``` Buffer is an embedded library, so it needs no install — see its docs below. - [Timeseries quickstart](https://www.opendata.dev/docs/timeseries/quickstart): Install and run your first metric queries. - [Log quickstart](https://www.opendata.dev/docs/log/quickstart): Append and scan your first event stream. - [Vector quickstart](https://www.opendata.dev/docs/vector/quickstart): Index and search your first vectors. - [Buffer docs](https://www.opendata.dev/docs/buffer): Use the ingestion buffer as a library. ## Using OpenData from applications OpenData products are not cloud APIs. A developer runs the relevant OpenData service and then calls it from their own application. In local quickstarts: - Timeseries usually listens on `http://localhost:9090` and exposes Prometheus-compatible HTTP endpoints. - Vector usually listens on `http://localhost:8080` and exposes JSON-over-HTTP endpoints for vector writes, deletes, search, and lookup. - Log usually listens on `http://localhost:8081` and exposes JSON-over-HTTP endpoints for append, scan, key, segment, and count operations. - Buffer is an embedded Rust/Go library pattern rather than a standalone HTTP server. ## Documentation - [Documentation home](https://www.opendata.dev/docs): Concepts shared across every database. - [Architecture overview](https://www.opendata.dev/docs/overview/architecture): How the object-store-native design works. - [Writing data](https://www.opendata.dev/docs/overview/writing-data) and [reading data](https://www.opendata.dev/docs/overview/reading-data): The shared read/write path. - [Storage](https://www.opendata.dev/docs/overview/storage) and [deployment](https://www.opendata.dev/docs/overview/deployment): How data is stored and how to deploy. ## API reference Each database exposes an HTTP API described by an OpenAPI 3.1 specification (raw YAML you can load directly): - Timeseries — [OpenAPI spec](https://www.opendata.dev/docs/openapi/timeseries.yaml), [reference](https://www.opendata.dev/docs/timeseries): `GET /api/v1/query`, `query_range`, `series`, `labels`, and `POST /api/v1/write`. - Log — [OpenAPI spec](https://www.opendata.dev/docs/openapi/log.yaml), [reference](https://www.opendata.dev/docs/log): `POST /api/v1/log/append`, `GET /api/v1/log/scan`, `keys`, `segments`, `count`. - Vector — [OpenAPI spec](https://www.opendata.dev/docs/openapi/vector.yaml), [reference](https://www.opendata.dev/docs/vector): `POST /api/v1/vector/write`, `delete`, `search`, and `GET /api/v1/vector/vectors/{id}`. ## Docs for LLMs - [Documentation index for LLMs (llms.txt)](https://www.opendata.dev/docs/llms.txt): Machine-readable index of every docs page. - [Full documentation text (llms-full.txt)](https://www.opendata.dev/docs/llms-full.txt): The entire documentation concatenated as plain text. - Any documentation page is also available as raw Markdown by appending `.md` to its URL, e.g. https://www.opendata.dev/docs/vector/quickstart.md ## Background - [Manifesto](https://www.opendata.dev/blog/manifesto): Why databases built on object storage change the operational model. - [Blog](https://www.opendata.dev/blog): Design deep-dives and release notes. - Built on [SlateDB](https://slatedb.io), an object-store-native LSM-tree storage engine. ## Community - [GitHub](https://github.com/opendata-oss/opendata): Source, issues, and releases. - [Discord](https://discord.gg/CsAQJ2AJGU): Questions and discussion.