Deterministic vector infrastructure with built-in graph support.
Store, search, and connect your data with cryptographically verifiable state, built for AI systems that need predictable, auditable infrastructure.
Project
Your environment
Vector
Similarity search
Graph
Relationship queries
Verified State
Cryptographically verifiable
Isolated workspace for your data, models and applications.
Unified vector and graph capabilities.
Auditable, deterministic infrastructure.
A vector database built for reproducible state
Valori stores, indexes, and searches vector data through a REST API — for applications where knowing exactly what state produced a result matters.
State you can reproduce, not just restore
It's often hard to know exactly what state produced a given search result in a vector system — recovery usually means "probably fine," not provably identical to what existed before.
Valori's kernel computes in fixed-point (Q16.16), not floating point, so the same inputs produce the same outputs on any machine — x86 or ARM. Every write commits to a BLAKE3-chained log and returns a receipt with the pre- and post-write state roots, so state isn't just backed up — it's replayable and verifiable.
How it fits together
From API key to vector search in minutes
# No installation required — plain HTTPScurl -X POST "https://app.valori.systems/v1/search" \
-H "Authorization: Bearer vlk_your_project_api_key" \
-H "Content-Type: application/json" \
-d '{"collection": "documents", "query": [0.1, 0.2, 0.3, 0.4], "k": 5}'One consistent contract
Every request and response shown here is the real, documented shape.
{
"collection": "documents",
"query": [0.1, 0.2, 0.3, 0.4],
"k": 5
}{
"results": [
{ "id": 42, "score": 0.0 },
{ "id": 17, "score": 0.043 }
]
}What you can build
RAG
Generation needs relevant, current context — not just what the model already knows.
Store document chunks as vectors and retrieve the nearest matches to ground a response.
Recommendation Systems
Ranking by similarity means comparing an item against everything else, fast.
Search a collection for the nearest neighbours to a user or item vector.
AI Agents
An agent needs memory that persists and can be queried across a session.
Insert and search vectors as an agent's long-term, queryable memory.
Advanced Search
Keyword match misses results that are semantically close but lexically different.
Query by meaning, not just matching terms — optionally blended with term-frequency reranking.
Data Analysis & Anomaly Detection
Outliers are easier to define by distance than by a fixed rule.
Find records far from their neighbours in vector space.
Where verifiable vector search matters
Not a customer list — the kinds of workloads deterministic, auditable state is built for.
Built for systems where state matters
Deterministic computation
Q16.16 fixed-point math, not floating point — the same inputs produce the same outputs on any machine.
Verifiable state
Every insert returns a BLAKE3 receipt: the pre- and post-write state roots and the log sequence.
Recovery & replay
State restores from snapshot and event log to a byte-identical hash — tested, not assumed.
API-first access
A REST API and two official SDKs — no infrastructure to run to start building.
Performance benchmarks — view benchmarks →
Build with infrastructure you can verify.
Start building with Valori — a REST API, official SDKs, and a deterministic core.