one atomic substrate
One engine for rows, vector embeddings, full-text postings and graph edges — committed together in one atomic write, replicated across nodes with consensus, and run for you as a single-tenant service in your region. Every write is visible at once to SQL, vector, graph, full-text and the natural-language /ask endpoint.
A member of the NVIDIA Inception program and the AI Council of India (IAMAI). Learn more →
Teams building AI-native products on OriginChain.
YCSB, ANN-Benchmarks and BEIR runs with the exact commands to reproduce them — and a section on what we did not measure.
Benchmarks →99.9%, 99.95% and 99.99% monthly uptime by configuration; point-in-time recovery to a 5 s boundary. One page; the Terms govern.
Service levels →What is implemented on every instance, what is in progress with a third party, and what is on the roadmap — three separate lists, and nothing in the second or third presented as a certification.
Trust center →Most modern AI stacks fan a single user action across four databases — a relational store, a vector index, a search engine, and a graph. Each has its own bearer, its own SDK, its own billing line, and its own way of being out of sync at 3 am. OriginChain replaces all four with one managed database. One write lands every shape atomically.
Retire four contracts. Onboard one. Stop coordinating four maintenance windows for a single feature release.
The row, its embedding, its full-text postings, and its graph edges commit together in one atomic write. No reconciliation jobs. No “the search index is 8 hours behind” disclaimers.
One SDK, one auth model, one observability surface. New AI features land without a Debezium pipeline + four retry queues + a 3 am reconciliation cron.
Relational rows, vector search, graph traversal, full-text and the
natural-language /ask — plus the auth,
replication and durability to run them in production. One endpoint,
one SDK, one bill.
Rows, vector embeddings, full-text postings and graph edges commit together in one write-ahead log — atomic by construction, no reconciliation.
Users, roles, row-level security and column masking — built into the same write.
Correlated EXISTS, aggregates and CASE — real SQL over the substrate.
4 distance metrics × 4 index variants. HNSW with metadata filters.
9 algorithms with per-hop EXPLAIN — BFS, paths, weighted Dijkstra.
BM25 ranking, phrase queries and 18-language stemming.
Ask in plain language; OriginChain routes it to the right shape.
Replicated across zones, consensus-backed, automatic failover. A write is confirmed once the cluster has accepted it.
Single-row writes are atomic with optimistic concurrency built in. Multi-row transactions under snapshot isolation — begin, put, commit, abort — with the Transactions add-on.
snapshot isolation · add-on $99 / month
A write is on durable storage before it is acknowledged. Encrypted nightly snapshots are included; opt-in point-in-time recovery restores to a 5 s boundary with retention up to 35 days.
pitr 5 s · retention ≤ 35 days · snapshots nightly
Use one shape or all five. One substrate. One atomic write.
A single write commits rows, embeddings, postings and graph edges atomically — every shape queries the same consistent corpus.
For a relational application that occasionally needs an embedding, Postgres with pgvector is the right answer. For an AI application whose rows, vectors, full-text postings and graph edges must agree on every write, one distributed engine with one atomic commit removes the sync job between them.
| OriginChain | Postgres + pgvector | Dedicated vector database | |
|---|---|---|---|
| Data shapes | Rows, vectors, full-text postings and graph edges — one atomic commit | Rows, vectors (pgvector) and full-text (tsvector); graph via recursive queries | Vectors with metadata filters; rows live elsewhere |
| Natural-language querying | Built in (/ask), bring your own model | Bring your own layer | Bring your own layer |
| Distribution & availability | Consensus-replicated across nodes; automatic failover; 99.9%–99.99% under contract | Streaming replication plus external failover tooling | Vendor-managed replicas |
| Transactions | Single-row atomic with optimistic concurrency; multi-row snapshot isolation as an add-on | Full ACID with MVCC — the stronger answer here | Per-record upserts |
| Operation | Fully managed, single-tenant, in your region | Self-run or a managed provider | Vendor-managed |
OriginChain is the distributed multimodal database for teams that need SQL, vector, full-text, graph and natural-language queries on one managed endpoint. Single-tenant, region-isolated, fully managed — provisioned in under two minutes, with availability, recovery and support terms written down on /sla.
Relational databases store relationships as data you re-derive at query time with joins. Graph databases store them as structure you walk. That single difference decides which questions stay fast as data grows — and it's why fraud rings, recommendations, and agent memory are graph problems.
Embeddings turn meaning into geometry; vector search turns geometry back into answers. A plain-language walk through similarity metrics, why brute force dies at scale, how HNSW graphs make approximate search fast, and what quantization actually trades away.
AI token usage has crossed from a technical detail to a strategic cost center — and "token maxing" is now one of the largest hidden line items at scale. OriginChain Contextual Intelligence (OCCI) cuts token consumption 30–50% while improving outcomes, by making context — not the model — the lever.
Pick a region, choose your configuration, and we provision a dedicated single-tenant instance in under two minutes. The first query you send is the first query we'll show you how to write - in English.
