Add concise demo entry point

README.md

@@ -9,16 +9,10 @@ tags:
   - research
 ---
 
-# SwarmFL: authorized specialist consultation over Flower
+# SwarmFL
 
-SwarmFL is a research framework for federated multi-agent inference. A specialist
-keeps its model and feature view on its own Flower client; an origin client may
-request a prediction from an authorized peer, then fuse only the predictions it
-actually received.
-
-This Hub card contains a small CPU-friendly synthetic quickstart for inspecting the
-communication and authorization path. It is an API demonstration, not a trained
-benchmark model or a clinical system.
+Small Flower demo for authorized specialist consultation. Synthetic only; not a
+benchmark or clinical model.
 
 ## Quickstart
 
@@ -33,33 +27,19 @@ a selected one-peer consultation on a new case, and writes the fused probability
 and audit artifacts under `demo/`. The synthetic specialists use the ordered input
 views `(x0,x1)`, `(x2,x3)`, and `(x4,x5)`.
 
-## What this demonstrates
-
-- isolated specialist identities and feature declarations;
-- authorization-aware, selected-peer consultation;
-- prediction fusion from observed responses only;
-- request-level feedback/audit output; and
-- deterministic local generation of the tiny demo adapters.
-
-It does not claim benchmark improvement, privacy, secure aggregation, differential
-privacy, clinical validity, or production readiness. The examples use synthetic
-inputs and labels. Do not infer performance on Wisconsin, CDC Diabetes, Adult, or
-any other dataset from this card.
+The demo keeps specialist views separate, makes one selected peer request, and
+fuses the observed predictions.
 
-## Scope and provenance
-
-The package and quickstart are released under Apache-2.0. This Hub publication
-intentionally excludes frozen test sets, private CDC artifacts, research checkpoints,
-raw experiment outputs, and paper-only figures. Those materials remain governed by
-their repository-specific access and provenance rules.
+```bash
+python demo.py --feedback
+```
 
-For implementation details and the full research protocol, see the
-[SwarmFL repository](https://github.com/saadbinather/SwarmFL). Contributions and
-reproduction reports are welcome through GitHub issues.
+If needed, install the package first:
 
-## Hugging Face release
+```bash
+pip install swarmfl==0.1.0
+```
 
-This card is the lightweight SwarmFL Hub release: a small, reproducible entry
-point for trying the Flower consultation path without downloading research
-datasets or private checkpoints. The full research artifacts remain in the
-repository and are not part of this public demo.
+The public Hub release contains no research datasets, frozen tests, private
+checkpoints, or experiment outputs. See the [repository](https://github.com/saadbinather/SwarmFL)
+for source and research context.