SCROLL TO CONTINUE

STOP 01

The Context Engine

Documents flow in. The engine addresses every passage as a byte-exact row — handing row ids to the vector store and writing content rows to .spdb files, the data store.

  1. 01 · Ingest document → engine
  2. 02 · Index engine → vector store
  3. 03 · Persist engine → .spdb files

STOP 02

Runs in the browser

A local agent ingests and retrieves against on-device .spdb storage — no server. When the device fills up, SpeedyDb spills to a cloud instance for storage and retrieval. A policy gate decides what leaves the device: sync everything, or keep sensitive files on-device.

Policy
browser agent · localhost
agent · indexing
on-device .spdb
STORAGE FULL · SPILLING TO CLOUD
policy gate
spill →
← retrieval
CLOUD INSTANCE
storage · retrieval
  1. 01 · Local agent + on-device .spdb
  2. 02 · Overflow device storage fills
  3. 03 · Spill policy-gated → cloud

STOP 03

Embedded in applications

The same engine drops into your app as a ~0.7 MB library — no browser quota to fight. The application (or its user) sets the storage budget; SpeedyDb keeps data within it, reserving model + index space first and evicting cloud-backed cold data before anything local-only.

Storage size
Browser
bounded by the device
~2 GB · OS quota
Application SpeedyDb ~0.7 MB
you set the budget
budget 16 GB · 95% free
reserve model + index · evict cloud-backed cold · never drop local-only
  1. 01 · Embed ~0.7 MB library
  2. 02 · Configure you set the size
  3. 03 · Fit budget-managed, no forced spill

STOP 04

Shared context

One cloud-backed context, linked across your browser, your application, and other people. Access is scoped by a privacy policy — grant or restrict per user, or per group.

Scope by
  1. 01 · Link browser + application
  2. 02 · Invite other users join
  3. 03 · Scope grant / restrict by policy

STOP 05

Stays out of RAM

No process-resident cache — reads are served from the reclaimable OS page cache, and the binary is ~0.7 MB. So the memory budget stays free for the things that actually need it: local models, vector DBs, and your coding tools.

Store
16 GB SYSTEM RAM resident ~0.7 MB
local model vector DB coding free OVER BUDGET · SWAPPING
SpeedyDb ~0.7 MB local model vector DB coding · IDE free
reads served from the reclaimable OS page cache — no in-process row cache to hold RAM hostage
  1. 01 · Footprint ~0.7 MB resident
  2. 02 · Headroom RAM left on the table
  3. 03 · Coexist models · DBs · tools run

STOP 06

Bring your pandas coming soon

Point SpeedyDb's lazy scan_spdb engine at a columnar .spdb and run the pandas you already write — filter, groupby, sort, value_counts, assign. Same answer, often faster, at a fraction of the memory and CPU: it streams from disk and decodes only the columns each op touches, so it never materializes the whole frame the way pandas must to be quick.

The same streaming discipline is already serving agents over MCP — combine, filter, aggregate, pivot and outlier removal, each bounded, with peak memory measured well under the size of the input. Try it on a file of your own.

  1. 01 · Faster beats pandas reading a store
  2. 02 · Less RAM a bounded working set, not the whole frame resident
  3. 03 · Less CPU only the touched columns are ever decoded

STOP 07 · LAST STOP

Try it in your browser

Load a file or our samples — chunked and stored by the real SpeedyDb engine compiled to WebAssembly, entirely on-device and saved in your browser's own storage. Then hit next and ask: RAG search returns the exact stored passages; Agent search has an on-device model write a grounded answer from them. Nothing ever leaves the page.

Try RAG in your browser →

ingest files, search your sources, and ask an on-device agent — all client-side, nothing leaves the page.

Transform a table →

upload a CSV, ask in plain English, download the result — an on-device model picks the operation and the MCP server streams it, so a file larger than memory is answered slowly rather than not at all.

  1. 01 · Load files & samples, on-device
  2. 02 · RAG search exact stored passages
  3. 03 · Agent search grounded on-device answer