Skip to content
iblaiPublic

About

On-premise knowledge layer that makes an organization's existing systems (PeopleSoft, Snowflake, Canvas, Salesforce, …) queryable by AI agents over MCP — no data extraction, your data stays in your network.

Topics

Resources

Stars

12 stars

Watchers

0 watching

Forks

Latest commit

 

History

4 Commits

Folders and files

Repository files navigation

Ontology

One ontology of the institution, exposed to agents over MCP. Connect the systems a campus already runs, dissect them into business terms and relationships, certify the result, and publish it as MCP servers that ibl.ai agents, Claude, Codex and any other MCP client can call. Runs inside your network, signs in with ibl.ai, and keeps no database of its own.

Status: the console works against the three systems it is built on: ibl.ai (sign-in, roles, API keys, agents), the NVIDIA auto-ontology engine and Google MCP Toolbox. It has not yet been run against a live platform, engine or runtime; docs/plan.md says what is unverified. Demo mode still serves the fixed data in lib/mock.ts.

Map

Quick start

On one host, as a developer runs it: the console's pages and its server half (the host API, the MCP gateway, Toolbox and their files) on one origin.

  1. Start the engine from a checkout of NVIDIA/auto-ontology. Fill in its .env as its README says, but set APP_URL=http://localhost:3100 and leave CONNECTION_STRINGS empty. The override runs the engine on the host's network (a Linux host), with its web app on 127.0.0.1:3100 and every service listening on loopback only, so a database on this machine is reached at 127.0.0.1 by the engine and by the console alike (Docker Compose 2.24.4 or later):

    docker compose -f docker-compose.yml -f <this repo>/deploy/auto-ontology.override.yml up -d --build
  2. Sign in at http://localhost:3100 as the engine's admin, open the user menu → API Tokens → New token, and copy it. It is shown once. An admin's token is needed for the model export.

  3. Start the console:

    pnpm install
    cp .env.example .env.local    # set AO_API_KEY
    pnpm dev                      # http://localhost:3000 → signs in with ibl.ai

    No setting names the platform: the console works on the one you sign in to, and the header's platform menu switches to another.

One engine serves one platform. The engine from step 1 is the host's own and serves the first platform whose admin uses the console. To serve another platform from the same host, start one more engine stack for it and write its address and service key to $ONTOLOGY_HOME/platforms/<platform>/engine.json as {"url": "...", "key": "..."}; until then that platform is told the host's engine serves another platform (docs/network.md).

Without an ibl.ai account, run in demo mode (fixed session, SDK panels show a notice):

NEXT_PUBLIC_DEMO=1 pnpm dev
node scripts/screenshots.mjs      # regenerate docs/screenshots (demo server running)
node scripts/audit-layout.mjs     # check every page at 375–1920px for broken layout

Sign-in returns only to origins on your platform's allowed redirect origins on ibl.ai; add http://localhost:3000 and the deployed hostname there.

Hosted: the pages on Vercel, the server on yours

The same code runs split in two. The console's pages are served from ibl.ai's Vercel (ontology.iblai.me for staging), built with NEXT_PUBLIC_ONTOLOGY_RELAY=1; they forward every /api/host/* call to the platform's own server. That server runs on a client's machine, installed with one command: the engine, the server half and Caddy for TLS, with the server half under https://<host>/iblai/ and NVIDIA's web app at the root. Opening the link the installer prints sets it up for one platform. Agents call https://<host>/iblai/<platform>/<slug>/mcp.

The install is in docs/install.md; what talks to what is in docs/network.md.

Documentation

Using the console

Document What it covers
docs/console.md Every page with screenshots, and what each control does
docs/guides/connect-data-sources.md Adding databases and HTTP APIs: read-only logins, scope, PII, what Connect & Dissect does, troubleshooting
docs/guides/permissions.md Policies and toolsets, sign-in methods, row-level restriction, PII, writes, origins, audit
docs/guides/connect-agents.md Attaching ibl.ai OS agents; Claude Desktop and Claude Code, Codex, Cursor, Python, cURL

Architecture and operations

Document What it covers
docs/install.md Installing a platform's server on a fresh Ubuntu with one command, and setting it up from its link
docs/auth.md ibl.ai SSO: the session as iblai/os implements it, how this console wires it, configuration, demo mode
docs/network.md Topology, inbound, outbound to ext.ibl.ai and data sources, what runs where, hardening
docs/data-sources.md Where every value on screen comes from, and where each thing is saved
docs/plan.md What is built and what is not yet verified

Screenshots

All in docs/screenshots/, one per page and tab: Dashboard · Systems · Systems · HTTP API · Map · Ontology · MCP Servers · New MCP Server · Server · Connect · Server · Playground · Server · Activity · Server · Tokens & Origins · Agents · Roles · Access · Catalog · Compilation · Config · Notifications · Collapsed Sidebar · Organization Menu · Management Dialog · Phone · Phone Menu

Repository layout

The repository root is the Next.js 16 console in the ibl.ai shell (os.ibl.ai header and sidebar, Open Sans, #0058cc).

Path What it is
app/(app)/* One folder per console route: dashboard, systems, map, ontology, mcp, mcp/new, mcp/[slug], agents, roles, access, catalog, sync (Compilation), config
components/shell/ Sidebar, header, page primitives, shell state, the OS-style footer dialogs hosting SDK tabs and the prototype walkthrough stepper
components/graph/ontology-graph.tsx Canvas force graph used on the Map
components/ui/ shadcn components (base-nova on Base UI)
lib/iblai/, store/, app/(auth)/ ibl.ai SSO: session, SDK store, sign-in callback
lib/host/, app/api/host/ The host layer: checks the caller with ibl.ai, proxies the platform's engine, writes its server configs and runs one Toolbox process per server
lib/host/setup.ts A server installed from a link: the setup code check and the one platform it serves
lib/host/relay.ts On the pages-only console, forwards each /api/host/* call to the server the browser names
lib/server-address.ts The x-ontology-server header, a server's address https://<host>/iblai, and the browser's store of it per platform
app/[platform]/[slug]/mcp/ The gateway agents call: origin, credential and toolset checks, then Toolbox
lib/data.ts, lib/ibl-flows.ts What each page reads and does; the ibl.ai writes (roles, policies, API keys, agents)
lib/mock.ts The fixed data demo mode shows
deploy/auto-ontology.override.yml Compose override that runs the engine on the host's network, on loopback, its web app on 3100
deploy/server.yml Compose file for a client's server, on top of the engine's: the server half on 127.0.0.1:3000 and Caddy with the /iblai routes
deploy/install.sh The one-command install of a platform's server on Ubuntu (docs/install.md)
Dockerfile The server half's image, ghcr.io/iblai/ontology: Next's standalone build with the Toolbox binary
.env.example Environment settings
scripts/screenshots.mjs Regenerates every screenshot in docs/screenshots/
scripts/audit-layout.mjs Checks every page at six widths for overflow, overlap and sideways scrolling
providers/iblai-providers.tsx ibl.ai SSO provider chain (see docs/auth.md)
AGENTS.md Rules for coding agents, written by next dev; keep it committed
docs/ The documents above
.claude/launch.json Dev-server definition for the Claude desktop preview

About

On-premise knowledge layer that makes an organization's existing systems (PeopleSoft, Snowflake, Canvas, Salesforce, …) queryable by AI agents over MCP — no data extraction, your data stays in your network.

Topics

Resources

Stars

12 stars

Watchers

0 watching

Forks

Releases

Packages

Used by

Contributors

Languages