Add Gradio UI wiring checks to the reviewer model

Model + space pickers (org spaces fetched live, free text allowed),
instant verdict/checklist/opportunities from the deterministic layer,
then the streamed LLM review.

app.py

@@ -0,0 +1,101 @@
+"""Submit Eval — is your Build Small hackathon Space ready to submit?
+
+Pick a model, pick a Space in the build-small-hackathon org, and get a
+grounded rule checklist plus model-written recommendations.
+"""
+
+import gradio as gr
+
+import checks
+import llm
+from guide import ORG, DEADLINE
+
+VERDICT_STYLE = {
+    "READY TO SUBMIT": ("🟢", "All hard rules pass — submit it!"),
+    "ALMOST READY": ("🟡", "No rule failures, but check the warnings below."),
+    "NOT READY": ("🔴", "At least one entry rule is not met yet."),
+}
+
+
+def refresh_spaces():
+    names = checks.list_org_spaces()
+    return gr.Dropdown(choices=names)
+
+
+def run_checks(space_name: str):
+    if not (space_name or "").strip():
+        raise gr.Error("Pick or paste a Space first.")
+    ev = checks.evaluate_space(space_name)
+    if not ev.exists:
+        return (
+            f"## 🔴 NOT READY\n{ev.error}",
+            "",
+            "",
+            ev.to_dict(),
+        )
+    icon, blurb = VERDICT_STYLE[ev.verdict]
+    verdict_md = f"## {icon} {ev.verdict}\n**`{ev.space_id}`** — {blurb}"
+    opps = ev.facts.get("opportunities") or []
+    opps_md = (
+        "### 🏅 Prizes & badges you might be leaving on the table\n"
+        + "\n".join(f"- {o}" for o in opps)
+        if opps
+        else ""
+    )
+    return verdict_md, checks.checklist_markdown(ev), opps_md, ev.to_dict()
+
+
+def run_review(model_id: str, ev_dict: dict):
+    if not ev_dict or not ev_dict.get("exists"):
+        yield "*(No Space evaluated — nothing to review.)*"
+        return
+    yield "*Loading the model — the first run of a model can take a few minutes…*"
+    yield from llm.generate_review(model_id, ev_dict)
+
+
+with gr.Blocks(title="Submit Eval — Build Small") as demo:
+    gr.Markdown(
+        f"# 👁 Submit Eval\n"
+        f"Checks a Space in the **`{ORG}`** org against the official "
+        f"[Build Small field guide](https://build-small-hackathon-field-guide.hf.space/) "
+        f"rules — track tags, badges, demo video, social post, the 32B model cap — "
+        f"then a small model (your pick) writes grounded recommendations. "
+        f"Deadline: **{DEADLINE}**."
+    )
+    with gr.Row():
+        model_dd = gr.Dropdown(
+            choices=[(label, mid) for mid, label in llm.MODELS.items()],
+            value=llm.DEFAULT_MODEL,
+            label="Reviewer model (all under 32B, of course)",
+            scale=2,
+        )
+        space_dd = gr.Dropdown(
+            choices=[],
+            label=f"Space in {ORG} (pick, or paste a name/URL)",
+            allow_custom_value=True,
+            scale=2,
+        )
+        refresh_btn = gr.Button("🔄", scale=0, min_width=48)
+        eval_btn = gr.Button("Evaluate", variant="primary", scale=1)
+
+    verdict_md = gr.Markdown()
+    checklist_md = gr.Markdown()
+    opps_md = gr.Markdown()
+    ev_state = gr.State()
+    gr.Markdown("---")
+    review_md = gr.Markdown(label="Model recommendations")
+
+    demo.load(refresh_spaces, outputs=[space_dd])
+    refresh_btn.click(refresh_spaces, outputs=[space_dd])
+    eval_btn.click(
+        run_checks,
+        inputs=[space_dd],
+        outputs=[verdict_md, checklist_md, opps_md, ev_state],
+    ).then(
+        run_review,
+        inputs=[model_dd, ev_state],
+        outputs=[review_md],
+    )
+
+if __name__ == "__main__":
+    demo.launch()