Skip to content
oramasearchPublic

About

The lake binary, where a customer designs, trains and ships AI on their own data: workspaces on Zetta, agents, apps, secure networking, forge, drive, identity, secrets, PKI, artifact registry and search, served as a web app, REST, OpenAPI and MCP. The hosted service is lake.live.

Resources

Code of conduct

Contributing

Security policy

Stars

2 stars

Watchers

0 watching

Forks

Latest commit

 

History

1,631 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

lake

The workspace where you build, train and ship AI on your own data.

lake is the server behind Orama Lake's console. It gives each account workspaces for its data, projects, pipelines, runs and models, and adds what a team needs around them: accounts and sign-in, agents, apps and services, version-controlled repositories, search, secrets and audit. Everything it does is one registry of capabilities, served as REST, OpenAPI and MCP, and the web console calls only that API, so an agent can do anything a person can.

Part of Orama Lake, the sovereign AI stack. MIT-licensed.

What it does

  • Workspaces on their own engine: each workspace has its own Zetta for datasets, projects, pipelines, runs and models, with Design, Train and Ship pages in the console. Docs
  • One API, three renderings: every capability is a REST call, an OpenAPI operation and an MCP tool, authorized, validated and audited in one place. Docs
  • Sign-in and identity: password, passkeys and TOTP, OpenID Connect and SAML single sign-on, SCIM provisioning, scoped API tokens, and identity pools for your own applications. Docs
  • Agents: library agents and your own, started from the console, a schedule or a stream, each with a credential bound to its workspace. Docs
  • Apps and services: deploy web and data apps and managed databases into a workspace. Docs
  • Repositories, drives and packages: liquid repositories with reviews, issues and branch rules; drives with sharing and history; OCI, npm and Python registries. Docs
  • Search, secrets and audit: repository search, a secret store with rotation and dynamic leases, a private certificate authority, and a signed, hash-linked audit log with a transparency log. Docs
  • Usage and billing: usage per workspace over time and limits; billing and prepaid balances are off unless LAKE_BILLING turns them on. Docs

What is implemented, tested and deployed is in the implementation status; every release, with what it includes and its known issues, is in the release notes.

Quickstart

To run the whole stack at once (lake, its engine, the agents and the operator), use the Orama Lake install guide.

To run lake alone from source, build it (below) and start it with the in-memory control plane:

./build/lake version
./build/lake --listen 127.0.0.1:4900 --node dev --memory-store --admin-token t

In another terminal:

curl -sS http://127.0.0.1:4900/v1/server_info

It answers the version and revision. Open http://127.0.0.1:4900/ in a browser for the console.

Build from source

You need CMake 3.28 or newer, Ninja, Clang 17 or newer (Homebrew LLVM on macOS), OpenSSL 3, libpq and Node. lake builds against two checkouts that are not its submodules, zetta and liquid, at the commits lake-platform pins; cloning the platform gets all three:

git clone --recurse-submodules https://github.com/oramasearch/lake-platform.git
cd lake-platform/modules/liquid-bridge/modules/lake
cmake -S . -B build -G Ninja \
  -DLAKE_ZETTA_SOURCE_DIR=$PWD/../../../zetta-bridge/modules/zetta \
  -DLAKE_LIQUID_SOURCE_DIR=$PWD/../liquid
cmake --build build -j 6
ctest --test-dir build --output-on-failure

Without LAKE_TEST_DB_URI the PostgreSQL-backed tests skip; with it, and LAKE_TEST_ZETTA_BINARY for the tests that start a Zetta, the whole suite runs. Every option is in docs/build/local.md and docs/build/testing.md.

How it fits the stack

lake is the control plane and the console. It keeps accounts, policy and audit in PostgreSQL, keeps each workspace's data in that workspace's Zetta, stores repositories with liquid's library, and hands what should run (workspaces, apps, services, builds) to the operator as desired state.

Piece What it does Relation to lake
lake-platform The whole stack, its install and its skill Pins and ships lake
zetta The engine for data, search and machine learning Runs each workspace; lake calls it
liquid Version control, builds, packages and deploys lake embeds its store and serves its sync
orama-operator Runs the stack on Kubernetes Applies lake's schema and runs what lake asks for

Known limits

  • The container image is built on Ubuntu, which bundles GPL and LGPL components; a static image is lake#1334.
  • Agent session keys do not yet meet liquid v0.3.0's delegation rules (lake#1346), and a new pod needs its liquid identity custody configured (lake#1352).
  • An unknown HTTP method on a routed path closes the connection without an answer (lake#1341).
  • The rest are in the open issues and each release's known issues.

Documentation

Community

Contributing

Contributions are welcome. CONTRIBUTING.md says how to build, test and open a pull request (against dev), and how contributions are licensed (CLA.md).

License

MIT; see LICENSE and COPYRIGHT. The components lake includes keep their own licences, listed in THIRD_PARTY_LICENSES.

About

The lake binary, where a customer designs, trains and ships AI on their own data: workspaces on Zetta, agents, apps, secure networking, forge, drive, identity, secrets, PKI, artifact registry and search, served as a web app, REST, OpenAPI and MCP. The hosted service is lake.live.

Resources

Code of conduct

Contributing

Security policy

Stars

2 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages