feat: Streamlit UI

requirements.txt

@@ -3,4 +3,5 @@ pandas
 scikit-learn
 pytest
 ruff
+streamlit
 python-dotenv

src/lysis/app.py

@@ -0,0 +1,44 @@
+"""Streamlit UI: streamlit run src\\lysis\\app.py"""
+
+import httpx
+import streamlit as st
+from dotenv import load_dotenv
+
+from lysis.cli import rank_repo
+from lysis.ingestion.github_client import GitHubClient
+from lysis.models.evaluate import load_dataset
+from lysis.models.gbm import make_gbm_ranker
+
+load_dotenv()
+st.set_page_config(page_title="Lysis", page_icon="🧭")
+
+
+@st.cache_resource
+def get_model():
+    df = load_dataset()
+    return make_gbm_ranker(df), len(df), df["first_timer"].mean()
+
+
+ranker, n, base = get_model()
+st.title("Lysis")
+st.caption("Beginner-friendly GitHub issue ranker")
+st.sidebar.write(f"Model trained on {n:,} closed issues. Base rate: {base:.2f}")
+repo = st.text_input("Repository", "psf/requests", placeholder="owner/repo")
+top = st.slider("Top N", 5, 30, 10)
+
+if st.button("Rank issues") and "/" in repo:
+    try:
+        with st.spinner("Fetching and ranking..."):
+            out = rank_repo(GitHubClient(), repo.strip(), ranker, top)
+    except httpx.HTTPError as e:
+        st.error(f"GitHub error: {e}")
+    else:
+        if out.empty:
+            st.warning("No open, unassigned issues found.")
+        else:
+            st.bar_chart(out.set_index("number")["score"])
+            st.dataframe(
+                out,
+                column_config={"url": st.column_config.LinkColumn("link")},
+                hide_index=True,
+            )