The Swagger UI for InnerMatch is now publicly available. Fire live API calls directly in your browser — no SDK, no API key, no setup. Everything the UI demo does under the hood, exposed and documented.
May 16, 2026 · Behrang Mehrparvar
Since we launched the InnerMatch interactive UI demo, one question kept coming up from developers: "Can I call the API directly?" Today the answer is yes — and you don't even need a terminal.
The InnerMatch API Explorer is a full Swagger UI, publicly available at synaptosearch.com/innermatch_demo/swagger.php. It exposes the complete OpenAPI specification for the demo API, with a built-in request builder that fires real calls against the running InnerMatch engine.
Open the API Explorer now
No account required. No API key. Click "Try it out" on any endpoint and fire a live request.
The API is organised into four groups. Each group is a distinct concern in the InnerMatch lifecycle — ranking, learning, introspection, and session management.
Three endpoints handle all retrieval:
POST /api/v1/rank — Rank a set of items for a user. You supply the candidate items; the engine scores and orders them according to what it has learned about this user so far.POST /api/v1/follow — Get the engine's best current recommendation. Returns the single item InnerMatch is most confident about given the current user model — the exploitation side of the explore/exploit balance.POST /api/v1/alternate — Get a deliberately contrasting recommendation. Returns an item that is intentionally different from the emerging preference pattern — the exploration side, used to prevent the model from narrowing too quickly.The UI demo calls /follow and /alternate together to form each round's pair: one confident pick and one deliberate contrast. You can call them independently in the Explorer to observe each strategy in isolation.
POST /api/v1/feedback — Submit feedback to update the user model. Each call is processed immediately and the model is updated before the next request. No batch step, no retraining cycle.Call this endpoint after each ranking result to signal whether the shown item was relevant. The more feedback events you submit in sequence, the more the model refines its understanding of the user's preference structure.
POST /api/v1/features — Inspect learned feature weights. Returns the engine's current internal state — which feature dimensions it has learned to weight positively and which it has learned to suppress based on the feedback received so far.This endpoint is particularly valuable for researchers and technical evaluators: it makes the engine's reasoning transparent at every step of the session, not just at the end.
POST /api/v1/remove_model — Wipe the user model for a session. Resets all learned weights for the session, returning the engine to its baseline state. Useful for starting a fresh learning trajectory within the same session context.The interactive UI demo walks you through the experience as an end user: select skills, rate a baseline, guide five rounds of learning, and compare results side by side. It is designed to make the adaptive learning effect felt rather than just understood.
The API Explorer is for a different audience. It is designed for:
The demo API runs on the same InnerMatch engine available on the AWS Marketplace. The demo tier is constrained to a job-matching dataset and short session windows, but the API surface — endpoints, request shapes, response schemas — is representative of the production API.
When you are ready to move to production — with your own feature domain, your own item catalogue, and continuous long-horizon learning loops — request API access and we will provision credentials with the same endpoints you have already explored.
| Group | Endpoint | What it does |
|---|---|---|
| Ranking | POST /api/v1/rank | Rank a set of items for a user |
| Ranking | POST /api/v1/follow | Get the engine's best current recommendation |
| Ranking | POST /api/v1/alternate | Get a deliberately contrasting recommendation |
| Learning | POST /api/v1/feedback | Submit feedback to update the user model |
| Introspect | POST /api/v1/features | Inspect learned feature weights |
| Session | POST /api/v1/remove_model | Wipe the user model for a session |