Private AI Module Kit
AI modules you embed in your clients' apps under your own brand — document reading, reply drafting with a human check, and summaries in English, Telugu and Hindi.
The modules inside Harbor Wealth, a fictional mutual-fund distributor's portal
The real screen, on synthetic dataRead: every field beside the words it came from
The real screen, on synthetic dataDraft: a reply about one account waits for a person
The real screen, on synthetic dataSummarise: a quarterly review in Telugu, its figures computed in code
The real screen, on synthetic dataThe whole demo as a video, with captions and chapters
Demo built on synthetic data.
1 of 5: The portal
The problem
Your clients want AI in the app you're building for them, and their security team wants to know where the data goes. Building each feature from scratch, with evaluation and a way to run it in the client's cloud, costs more than the feature earns you.
What it does
Three modules, shipped as containers with a REST API and a small JavaScript SDK — read a document into fields; draft a reply and hold it for review below a confidence line; write a summary in English, Telugu or Hindi.
Themeable components that take your client's brand. Stafflo's name doesn't appear.
Numbers in any summary are computed from your data; the model only writes the words around them.
Each module ships with a test report on the client's own examples.
Sample integrations for Laravel with Inertia and for Next.js.
The demo embeds all three modules in Harbor Wealth, a fictional mutual-fund distributor's portal.
Try to break it
On the call we'll trigger these, live:
- A client uploads a phone photo of a statement. The module returns what it could read and marks the rest for review.
- The model server stops. The module answers "unavailable, retry in 30 seconds" and your app shows its own fallback.
- A number in a summary doesn't match the portfolio. The summary is blocked before anyone sees it.
Where it runs
In your client's cloud or on their servers.
Who else does something like this
Sarvam offers Indian-language AI models with private cloud and on-premise deployment.
The Module Kit is made for agencies — white-label, tested per client, and built to sit inside apps you already ship.
The rules it respects
Modules run inside your client's boundary and write an audit log to their database.
What our tests measured
How we testThese figures come from test runs on Google's Gemini (gemini-3.5-flash-lite), a hosted model, on synthetic data. We'll add runs on a local model.
- Run on 4 October 2026
- Model: gemini:gemini-3.5-flash-lite
- 252 examples
For your engineers
- Engines
- Extractor, Gatekeeper, Proof Console
- Sealed mode
- Test passed: with sealed mode on, it refuses every outbound connection except to the model you've approved.
Data flow