Isolation boundary
acme-systemslarge
northwind-capmedium
lyra-researchsmall
Each instance is its own container with its own filesystem, process and credential. Compute, memory and storage are dedicated, so your instance performs at its best.
Rayforce Cloud is in closed beta — managed instances are provisioned on requestTalk to us about a demo: sales@rayforcedb.comThe engine stays open source and MIT licensed
Analytics and Graph.One pipeline.Managed in cloud.
RayforceDB is a pure C, zero-dependency engine where columnar aggregation and graph traversal share a single operation DAG. Rayforce Cloud runs it for you — an isolated instance with easy access over TLS.
; aggregate by symbol, in the engine
Awaiting run
The platform is on us.The query is on you.
Provisioning, isolation, transport, credentials, metrics and snapshots are our problem. Yours is the query — so spend the day on the thing that actually pays, not on the plumbing under it.
Web-transport
ClientInstance
One socket · held open
The browser workbench and any programmatic client share one persistent TLS WebSocket. Jobs, progress, result pages and cancellations all travel over the single socket.
Credentials
rf_xi4m••••••••fp2
Issued per instance · revoked in one click
Credit the instance access using instance tokens you issue in the console. Revocable at any moment.
Monitoring the instances
Every instance reports what it is doing: memory, running jobs and queue depth sampled continuously and drawn on its page, execution logs you can filter and export, snapshots on the interval you choose and a restore that takes one action.
42%Memory
18Jobs
03Queue
InstancevCPUStorageState
prod-analytics2.0020 GBRunning
events-ingest1.5010 GBRunning
staging-graph0.755 GBProvisioning
submitthe Rayfall job
progresswhile it runs
resulttyped, first page
fetch_result_pagethe next page
bind_resultname it, aggregate in-engine
JSON or native binaryNo polling
A workbench that speaks straight to engine
A working surface that holds a WebSocket open to the runtime inside your instance and speaks its job protocol directly: submit, progress, cancel, fetch a page, bind a result.
Build visual charts of the whole answer.
A result that does not fit in a page stays retained inside the instance. Naming it hands the aggregation back to the engine, which runs it over every row and returns the few numbers a chart actually needs.
The browser is left doing the only thing a browser is good at: drawing. It never has to hold a million rows to plot them, and the picture never quietly becomes a picture of the first hundred.
Browser holds
100
rows · one page
Engine aggregates
every row
of the retained result
(select {y: (sum notional) from: .wb.r1 by: symbol})
A small engine that refuses to be two systems.
RayforceDB is an embeddable database written in pure C with no dependencies. Columnar analytics and graph traversal are not two engines bolted together in it — they share one operation DAG, pass through one optimizer, and execute as bytecode compiled at execution time. That is why a query can aggregate a table and walk a graph in the same pass, with no export step in the middle and no second store to keep in sync.
Pure C
Zero third-party dependencies, and no malloc — the engine manages its own memory.
One DAG
Columnar aggregation and graph traversal are the same operation graph, through the same multi-pass optimizer.
JIT bytecode
00load trades
01filter qty > 100
02group symbol
03reduce notional
Queries execute as morsel-driven bytecode compiled at execution time.
MIT licensed
MIT
LICENSE · rayforce
The engine is open source. The cloud runs it — it does not fork it.
Small enough to embed almost anywhere.
The core is a zero-dependency C library, so it goes where your code already runs: inside a Python process, a Rust service, a browser tab, your editor. The bindings are not wrappers around a server — they are the engine, in-process, with the cloud available when the data outgrows the laptop.
Python↗
rayforce-py
A Pythonic ORM with a chainable query API, talking to the core through its C API rather than a marshalling shim. Zero dependencies.
Rust↗
rayforce-rs
Bindings that call the C API directly, so reads are zero-copy where it counts — a numeric column arrives as a slice, not element by element.
VS Code↗
rayforce-vscode
Instance manager, interactive REPL, environment inspector, and full Rayfall language support with autocomplete and highlighting.
WebAssembly↗
rayforce-wasm
The engine in the browser, with TypedArray views straight over WASM memory and SIMD where the host supports it.
Q↗
rayforce-q
A Q IPC wire-format client exposed as Rayfall builtins, so an existing Q server can be queried from a Rayfall session.
Embedded C↗
rayforce.h
One header and a binary under a megabyte. If it links C, it can carry the whole database — which is the point.
Documentation
The cloud documentation covers the whole path — from provisioning an instance to running your first query. The Rayfall language has its own documentation: the type system, the builtins, and the C API you embed against.
Cloud docs · contents
Quickstart
Provision an instance, mint a token, run your first Rayfall query from the workbench.
Workbench
Typed results, paging, engine-side charts, progress and cancellation.
Client access
The job protocol over WSS: auth, submit, progress, result pages, bind, cancel.
Instances
Sizing, metrics, execution logs, snapshots and restore.
Teams
Members, roles, and what each permission actually gates.
Bring us your worst query.
Tell us the shape and dimensions of your data and the queries you keep working around. We will do our best to supply you with an instance and help you run it.