Valori LogoValori
Developer documentation

Build memory your AI can verify.

Retrieve relevant context with vectors, connect supporting evidence as a graph, and replay the exact state behind every result.

REST APIPython SDKTypeScript SDKOpen-source kernel
First request
HTTPS
curl https://app.valori.systems/v1/namespaces \
  -H "Authorization: Bearer vlk_..."
200 {"collections":[]}
Vector
Retrieve relevant memory
Graph
Follow connected evidence
Verified
Replay the exact state
Quick start

From an API key to your first search

Use plain HTTPS or an official SDK. Your project key resolves the Cloud project and worker, so application code never needs a node URL.

List your collections
curl "https://app.valori.systems/v1/namespaces" \-H "Authorization: Bearer vlk_your_project_api_key"
  1. 1Step 1

    Create a project key

    Generate a project-scoped API key in Valori Cloud.

  2. 2Step 2

    Create a collection

    Choose the vector dimension, distance metric, and index.

  3. 3Step 3

    Insert records

    Store vectors with text and metadata in a named collection.

  4. 4Step 4

    Search and connect

    Retrieve similar records and follow graph relationships for context.

Architecture

One memory layer, four guarantees

Valori combines retrieval and relationships with deterministic state instead of treating them as separate systems.

Vector retrieval
Find semantically similar records inside a collection using a fixed dimension and distance metric.
Graph context
Connect documents, chunks, concepts, and evidence so retrieval can follow meaningful relationships.
Deterministic state
Apply mutations in a defined order with stable tie-breaking and reproducible execution.
Verifiable history
Use the WAL, event log, snapshots, and BLAKE3 state hashes to independently check what changed.
Resources

Go deeper

Inspect the open-source kernel, use the exact API contracts, or continue through the conceptual guides.