AIRA Forge turns scattered, messy data into a living model of your organization—then lets anyone build dashboards, automate decisions, and ask an AI that answers from your real data and cites its sources. No cloud dependency. Nothing ever leaves your network.
Connect → Construct → ActCortex grounded AI18 industry templates100% on-premiseRuns on your own local LLM
01 Connect — the data layer
Bring in any source, and understand it instantly.
Upload spreadsheets, pull from live databases and REST APIs, or sync a Google Sheet on a schedule. The moment data lands, Forge profiles it for you.
Connect to anything — CSV/Excel, Postgres/MySQL, REST APIs, CSV & Sheets URLs, with scheduled auto-sync that keeps objects current.
Schema inference — Forge reads your columns and builds the object type for you; no manual modeling to get started.
Automatic profiling — types, missing values, ranges, outliers and correlations the instant data arrives.
ExamplePoint Forge at your warehouse database. Within seconds it flags that 12% of value_php is missing and that two shipments are statistical outliers—before you've built a single chart.
forge · connect · data profiling
Fig 1 — Connect ▸ Data profiling: every field characterised the moment data lands.
02 Construct — the semantic model
Turn tables into a living model of your business.
Forge maps raw data to real-world objects—a Shipment, a Patient, a Counterparty—linked into the Construct: one shared model everyone uses instead of arguing over SQL.
Object & relationship model — drag types around an ontology map and draw the relationships between them.
Link analysis — an interactive graph with degree centrality, community detection and shortest-path to surface hidden connections.
Explore & act — search, filter and export objects; open any one to inspect, analyse with Cortex, or stage an action.
ExampleAn investigator drops a person of interest onto the graph and runs shortest-path—Forge reveals the two intermediaries linking them to a flagged organization.
forge · construct · link analysis
Fig 2 — Construct ▸ Link analysis: shortest-path highlights the connection in amber.
03 Act — dashboards & automation
Operational apps, not static reports.
Forge's dashboards behave like applications: a shared filter drives every widget at once, charts cross-filter on click, and a background runtime evaluates your rules and acts—staging actions for approval and sending alerts.
Reactive dashboards — bind an object set, then a slider or filter updates every metric, chart, table and map live.
Flow builder + approvals — compose “when X, do Y” visually; actions route to an approvals queue with a full audit trail.
Automation runtime — workflows fire on their own and notify you in-app, by Slack, webhook or email.
ExampleDrag the risk slider to 7 and the board instantly narrows to 38 high-risk shipments; overnight, the runtime stages reroute actions for the three that turned delayed and pings the ops channel.
forge · act · workshop dashboard
Fig 3 — Act ▸ Reactive dashboard: one filter drives every widget; select a row to inspect and act.
04 Cortex — the AI layer
An AI that answers from your data—and cites its sources.
Cortex reasons over your live objects and your documents using a language model running on your hardware. It retrieves the actual passages, answers, and shows you exactly where each claim came from.
Document RAG with citations — upload a contract or report; ask a question; get an answer that cites each source passage [D#].
Proactive insights — Cortex scans everything on a schedule and surfaces risks, spikes and anomalies with a recommended action.
Ask for a chart — describe what you want in plain language and Cortex computes and draws it live.
Grounded, never invented — answers are tied to retrieved evidence, with role- and clearance-aware access.
Example“What are the payment terms?” → Cortex answers “Net 30 days; 2%/month interest on late payment [D1]”, with the exact clause shown beneath.
cortex · knowledge · ask your documents
Fig 4 — Cortex ▸ Knowledge: a retrieval-augmented answer with citations traced to the source clauses.
Insights that find you.
Cortex doesn't wait to be asked. It continuously scans the whole model, ranks what matters, and writes the next step—so the right information surfaces at the right time.
ExampleMonday morning: “5 counterparties crossed risk 8 this week—recommend staging compliance review for Sunrise Holdings (₱240M exposure, KYC pending).”
For a government agency, a bank, a hospital or a utility, adopting AI has never been only about capability—it's about control. Forge and Cortex are built to run entirely inside your own network.
🔒 Fully on-premise
Air-gap-capable, zero cloud dependency. Sensitive records never leave your infrastructure—supporting your data-residency and privacy obligations.
🧠 Your own LLM
Cortex runs on local models on your hardware. No external API calls, no data sent abroad, no per-token cloud bill.
🛡️ Governed access
Role-based permissions and data-classification clearances—and the AI inherits the same boundaries every user has.
05 Built for any industry
Start in minutes with an industry template.
Every industry is the same engine with a different model. Pick yours and Forge scaffolds a complete working app—objects, sample data, a dashboard, an automation and an AI function—in one click.