Promote-candidate UI: add-a-source pipeline in the admin console

Bring the supervised source-candidate flow into Sources (Codex's v1 scope), so
adding feeds no longer needs the CLI.

* feeds.safe_fetch_feed: SSRF-safe fetch for UNTRUSTED (admin-pasted) URLs —
  http(s) only, every redirect hop re-validated via enrich._host_is_public,
  body size-capped, bounded redirects, no cookies. preview_feed gains a
  `fetcher` param; the API path passes safe_fetch_feed (NOT the raw fetch_feed
  used for already-vetted polling).
* API (admin-gated): GET /candidates; POST /candidates (suggest+preview, gated
  before the outbound fetch, no DB conn held during network); /{id}/preview
  (explicit re-preview); /{id}/promote (paused by default, returns the new
  source + updated candidate); /{id}/reject. rejected stays on candidates only.
* Admin Sources tab: "Add a source" field + a candidate queue showing the
  preview (pass rate, recent count, example headlines) with Promote (as paused,
  or Activate immediately) / Re-preview / Reject.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

frontend/src/routes/admin/+page.svelte

@@ -30,6 +30,7 @@
     try {
       await loadStats();
       feedback = await getJSON('/api/admin/feedback');
+      candidates = await getJSON('/api/admin/candidates');
     } catch {
       error = "Couldn't load stats.";
     }
@@ -116,6 +117,51 @@
     catch { s.review_flag = 1; s.review_reason = prevR; }
   }
 
+  // --- Source candidates: supervised "add a source" pipeline ---
+  let candidates = $state([]);
+  let newFeedUrl = $state('');
+  let newFeedName = $state('');
+  let addBusy = $state(false);
+  let addErr = $state('');
+  let pendingCandidates = $derived(candidates.filter((c) => c.status !== 'promoted' && c.status !== 'rejected'));
+
+  async function addCandidate() {
+    const url = newFeedUrl.trim();
+    if (!url || addBusy) return;
+    addBusy = true; addErr = '';
+    try {
+      const c = await postJSON('/api/admin/candidates', { feed_url: url, name: newFeedName.trim() || null });
+      candidates = [c, ...candidates.filter((x) => x.id !== c.id)];
+      newFeedUrl = ''; newFeedName = '';
+    } catch (e) {
+      addErr = e?.message || "Couldn't preview that feed.";
+    } finally {
+      addBusy = false;
+    }
+  }
+  async function repreviewCandidate(c) {
+    c._err = '';
+    try { const u = await postJSON(`/api/admin/candidates/${c.id}/preview`); c.preview = u.preview; }
+    catch (e) { c._err = e?.message || 'Re-preview failed.'; }
+  }
+  async function promoteCandidate(c) {
+    c._err = '';
+    try {
+      await postJSON(`/api/admin/candidates/${c.id}/promote`, {
+        default_category: (c._cat || '').trim() || null,
+        active: !!c._activate,
+      });
+      c.status = 'promoted';
+      await loadStats(); // surface the new (paused) source in the table below
+    } catch (e) {
+      c._err = e?.message || 'Promote failed.';
+    }
+  }
+  async function rejectCandidate(c) {
+    try { await postJSON(`/api/admin/candidates/${c.id}/reject`); c.status = 'rejected'; }
+    catch { /* leave as-is */ }
+  }
+
   // Feedback inbox: filter + read/unread + delete.
   let fbCat = $state('all'); // 'all' | 'unread' | a category
   let shownFeedback = $derived(
@@ -338,6 +384,48 @@
       </div>
 
     {:else if section === 'sources'}
+      <section class="addsrc">
+        <h2>Add a source</h2>
+        <div class="addrow">
+          <input type="url" placeholder="Feed URL (RSS / Atom)" bind:value={newFeedUrl} onkeydown={(e) => e.key === 'Enter' && addCandidate()} />
+          <input type="text" placeholder="Name (optional)" bind:value={newFeedName} />
+          <button class="csend" onclick={addCandidate} disabled={addBusy || !newFeedUrl.trim()}>{addBusy ? 'Previewing…' : 'Preview & add'}</button>
+        </div>
+        {#if addErr}<p class="cerr">{addErr}</p>{/if}
+        <p class="legend2">Fetches a sample (safely) and stages it as a candidate — nothing is added live until you Promote.</p>
+      </section>
+
+      {#if pendingCandidates.length}
+        <section>
+          <h2>Candidate queue <span class="count">({pendingCandidates.length})</span></h2>
+          <ul class="candlist">
+            {#each pendingCandidates as c (c.id)}
+              <li>
+                <div class="chead">
+                  <span class="cname">{c.name || c.feed_url}</span>
+                  <span class="cstatus">{c.status}</span>
+                </div>
+                <div class="curl">{c.feed_url}</div>
+                {#if c.preview}
+                  <div class="cprev">
+                    {c.preview.accepted ?? 0}/{c.preview.sampled ?? 0} sampled would pass{#if c.preview.acceptance_rate != null} · {Math.round(c.preview.acceptance_rate * 100)}% accept{/if}{#if c.preview.recent_7d != null} · {c.preview.recent_7d} in last 7d{/if}
+                    {#if c.preview.examples_accepted?.length}<div class="cex">e.g. {c.preview.examples_accepted.slice(0, 3).join(' · ')}</div>{/if}
+                  </div>
+                {/if}
+                {#if c._err}<p class="cerr">{c._err}</p>{/if}
+                <div class="cactions">
+                  <input class="ccat" type="text" placeholder="category (optional)" bind:value={c._cat} />
+                  <label class="cchk"><input type="checkbox" bind:checked={c._activate} /> Activate immediately</label>
+                  <button class="csend" onclick={() => promoteCandidate(c)}>Promote{c._activate ? '' : ' as paused'}</button>
+                  <button class="act" onclick={() => repreviewCandidate(c)}>Re-preview</button>
+                  <button class="act del" onclick={() => rejectCandidate(c)}>Reject</button>
+                </div>
+              </li>
+            {/each}
+          </ul>
+        </section>
+      {/if}
+
       <h2>Sources</h2>
       <p class="sub2">{healthy} healthy · {resting} resting · {flagged} flagged · {paused} paused · {retired} retired · {sources.length} total</p>
       <div class="filterchips">
@@ -664,6 +752,23 @@
   .filterchips .chip:hover { border-color: var(--accent); }
   .filterchips .chip.on { background: var(--accent); border-color: var(--accent); color: #fff; }
 
+  /* Add a source + candidate queue */
+  .addsrc { background: var(--surface); border: 1px solid var(--line); border-radius: 14px; padding: 14px 16px; margin-bottom: 6px; }
+  .addrow { display: flex; gap: 8px; flex-wrap: wrap; }
+  .addrow input { flex: 1 1 200px; box-sizing: border-box; font: inherit; font-size: 0.9rem; padding: 8px 11px; border: 1px solid var(--line); border-radius: 9px; background: var(--bg); color: var(--ink); }
+  .addrow input:focus { outline: none; border-color: var(--accent); }
+  ul.candlist { list-style: none; margin: 0; padding: 0; display: flex; flex-direction: column; gap: 10px; }
+  ul.candlist li { background: var(--surface); border: 1px solid var(--line); border-radius: 12px; padding: 12px 14px; }
+  .chead { display: flex; align-items: baseline; gap: 10px; }
+  .chead .cname { font-weight: 600; color: var(--ink); }
+  .chead .cstatus { font-size: 0.72rem; color: var(--muted); text-transform: capitalize; }
+  .curl { font-size: 0.76rem; color: var(--muted); word-break: break-all; margin-top: 2px; }
+  .cprev { font-size: 0.84rem; color: var(--ink); margin-top: 7px; }
+  .cprev .cex { color: var(--muted); font-size: 0.8rem; margin-top: 2px; font-style: italic; }
+  .cactions { display: flex; gap: 9px; align-items: center; flex-wrap: wrap; margin-top: 10px; }
+  .cactions .ccat { font: inherit; font-size: 0.8rem; padding: 5px 9px; border: 1px solid var(--line); border-radius: 8px; background: var(--bg); color: var(--ink); width: 150px; }
+  .cactions .cchk { font-size: 0.8rem; color: var(--muted); display: inline-flex; align-items: center; gap: 5px; }
+
   /* Source health table */
   .tablewrap { overflow-x: auto; }
   .srctable { width: 100%; border-collapse: collapse; font-size: 0.86rem; min-width: 560px; }

goodnews/api.py

@@ -31,7 +31,7 @@ from fastapi.responses import HTMLResponse, RedirectResponse
 from fastapi.staticfiles import StaticFiles
 from pydantic import BaseModel
 
-from . import auth, email_send, feeds, oauth_google, queries, share, summarize
+from . import auth, email_send, feeds, oauth_google, queries, share, sources, summarize
 from .markup import reply_html_to_text, sanitize_reply_html
 from .db import connect
 from .filters import filter_articles, prefs_from_json
@@ -394,6 +394,19 @@ class SourceVisibilityBody(BaseModel):
     visible: bool = True
 
 
+class CandidateSuggestBody(BaseModel):
+    feed_url: str = ""
+    name: str | None = None
+
+
+class CandidatePromoteBody(BaseModel):
+    default_category: str | None = None
+    active: bool = False  # promote-as-paused by default; opt in to activate
+    trust_score: int = 5
+    pr_risk_score: int = 3
+    poll_interval_minutes: int = 180
+
+
 class SourceReviewBody(BaseModel):
     flag: bool = False
     reason: str | None = None
@@ -918,6 +931,85 @@ def create_app() -> FastAPI:
             conn.commit()
         return {"ok": True, "visible": body.visible}
 
+    # --- Source candidates (supervised add-a-source pipeline) ----------------
+
+    def _candidate_dict(row) -> dict:
+        d = dict(row)
+        raw = d.pop("preview_json", None)
+        try:
+            d["preview"] = json.loads(raw) if raw else None
+        except (ValueError, TypeError):
+            d["preview"] = None
+        return d
+
+    @app.get("/api/admin/candidates")
+    def admin_candidates(request: Request) -> list[dict]:
+        with get_conn() as conn:
+            _require_admin(conn, request)
+            rows = sources.list_candidates(conn)
+        return [_candidate_dict(r) for r in rows]
+
+    def _preview_or_502(url: str) -> dict:
+        # SSRF-safe fetch (admin-pasted URL is untrusted); heuristic-only (fast).
+        try:
+            return feeds.preview_feed(url, sample=20, fetcher=feeds.safe_fetch_feed)
+        except Exception as exc:  # noqa: BLE001 — surface a readable reason
+            raise HTTPException(status_code=502, detail=f"Couldn't preview that feed: {exc}")
+
+    @app.post("/api/admin/candidates")
+    def admin_candidate_suggest(body: CandidateSuggestBody, request: Request) -> dict:
+        url = (body.feed_url or "").strip()
+        if not url:
+            raise HTTPException(status_code=422, detail="feed_url is required")
+        with get_conn() as conn:  # gate BEFORE the outbound fetch
+            _require_admin(conn, request)
+        preview = _preview_or_502(url)  # no DB connection held during network I/O
+        with get_conn() as conn:
+            _require_admin(conn, request)
+            row = sources.save_candidate(conn, url, preview=preview, name=(body.name or None), status="suggested")
+        return _candidate_dict(row)
+
+    @app.post("/api/admin/candidates/{cid}/preview")
+    def admin_candidate_repreview(cid: int, request: Request) -> dict:
+        with get_conn() as conn:
+            _require_admin(conn, request)
+            cand = conn.execute("SELECT feed_url FROM source_candidates WHERE id = ?", (cid,)).fetchone()
+            if not cand:
+                raise HTTPException(status_code=404, detail="candidate not found")
+            url = cand["feed_url"]
+        preview = _preview_or_502(url)
+        with get_conn() as conn:
+            _require_admin(conn, request)
+            row = sources.save_candidate(conn, url, preview=preview)
+        return _candidate_dict(row)
+
+    @app.post("/api/admin/candidates/{cid}/promote")
+    def admin_candidate_promote(cid: int, body: CandidatePromoteBody, request: Request) -> dict:
+        with get_conn() as conn:
+            _require_admin(conn, request)
+            try:
+                source_id = sources.promote_candidate(
+                    conn, cid, active=body.active, default_category=body.default_category,
+                    trust_score=body.trust_score, pr_risk_score=body.pr_risk_score,
+                    poll_interval_minutes=body.poll_interval_minutes,
+                )
+            except ValueError:
+                raise HTTPException(status_code=404, detail="candidate not found")
+            src = conn.execute(
+                "SELECT id, name, status, active, content_visible FROM sources WHERE id = ?", (source_id,)
+            ).fetchone()
+            cand = conn.execute("SELECT * FROM source_candidates WHERE id = ?", (cid,)).fetchone()
+        return {"ok": True, "source_id": source_id, "source": dict(src), "candidate": _candidate_dict(cand)}
+
+    @app.post("/api/admin/candidates/{cid}/reject")
+    def admin_candidate_reject(cid: int, request: Request) -> dict:
+        with get_conn() as conn:
+            _require_admin(conn, request)
+            if not sources.reject_candidate(conn, cid):
+                raise HTTPException(status_code=404, detail="candidate not found")
+            cand = conn.execute("SELECT * FROM source_candidates WHERE id = ?", (cid,)).fetchone()
+        return _candidate_dict(cand)
+
     @app.post("/api/admin/sources/{sid}/review")
     def admin_source_review(sid: int, body: SourceReviewBody, request: Request) -> dict:
         with get_conn() as conn:

goodnews/feeds.py

@@ -9,12 +9,15 @@ import xml.etree.ElementTree as ET
 from collections import Counter
 from dataclasses import dataclass
 from datetime import UTC, datetime
+from urllib.parse import urljoin, urlsplit
 
+from .enrich import MAX_REDIRECTS, _NoRedirect, _host_is_public
 from .scoring import score_article
 from .text import canonicalize_url, clean_text, sha256_text
 
 
 USER_AGENT = "goodNews/0.1 (+local constructive news prototype)"
+FEED_MAX_BYTES = 2_000_000  # cap on a fetched feed body (SSRF-safe preview path)
 
 
 @dataclass
@@ -176,14 +179,15 @@ def poll_source(conn: sqlite3.Connection, source: sqlite3.Row) -> dict:
         }
 
 
-def preview_feed(url: str, sample: int = 25, pr_risk_default: int = 3, client=None) -> dict:
+def preview_feed(url: str, sample: int = 25, pr_risk_default: int = 3, client=None, fetcher=None) -> dict:
     """Fetch and score a sample of a feed WITHOUT persisting anything.
 
     Read-only: lets a user vet a candidate source before it is ever added. By
     default it uses the fast heuristic; pass an LLM client to also get the
-    topic/flavor mix and the model's acceptance view (slower).
+    topic/flavor mix and the model's acceptance view (slower). Pass
+    fetcher=safe_fetch_feed for untrusted (admin-pasted) URLs.
     """
-    items = parse_feed(fetch_feed(url))
+    items = parse_feed((fetcher or fetch_feed)(url))
     rows = []
     for item in items[:sample]:
         title = clean_text(item.title, max_len=500)
@@ -287,6 +291,40 @@ def fetch_feed(url: str, timeout: int = 20) -> bytes:
         raise RuntimeError(f"failed fetching {url}: {exc.reason}") from exc
 
 
+def safe_fetch_feed(url: str, timeout: int = 10) -> bytes:
+    """SSRF-safe feed fetch for UNTRUSTED (admin-pasted) URLs.
+
+    Unlike fetch_feed (used for already-vetted, curated feeds), this re-validates
+    every redirect hop against public IPs (reusing enrich._host_is_public),
+    allows only http(s), caps the body, follows a bounded number of redirects,
+    and sends no cookies/credentials. Use this for the preview/candidate path.
+    """
+    opener = urllib.request.build_opener(_NoRedirect)
+    current = url
+    for _ in range(MAX_REDIRECTS + 1):
+        parts = urlsplit(current)
+        if parts.scheme not in ("http", "https") or not _host_is_public(parts.hostname):
+            raise RuntimeError("refusing to fetch a non-public or non-http(s) URL")
+        request = urllib.request.Request(current, headers={"User-Agent": USER_AGENT})
+        try:
+            response = opener.open(request, timeout=timeout)
+        except (urllib.error.URLError, OSError, ValueError) as exc:
+            raise RuntimeError(f"failed fetching {current}: {exc}") from exc
+        status = getattr(response, "status", 200) or 200
+        if status in (301, 302, 303, 307, 308):
+            location = response.headers.get("Location")
+            response.close()
+            if not location:
+                raise RuntimeError("redirect without a location")
+            current = urljoin(current, location)
+            continue
+        try:
+            return response.read(FEED_MAX_BYTES)
+        finally:
+            response.close()
+    raise RuntimeError("too many redirects")
+
+
 def parse_feed(xml: bytes) -> list[FeedItem]:
     root = ET.fromstring(xml)
     root_name = _local_name(root.tag)

tests/test_admin.py

@@ -123,3 +123,40 @@ def test_admin_stats_days_param_clamped(tmp_path, monkeypatch):
     assert tc.get("/api/admin/stats?days=90").json()["days"] == 90
     assert tc.get("/api/admin/stats?days=999").json()["days"] == 30   # clamped
     assert tc.get("/api/admin/stats").json()["days"] == 30            # default
+
+
+def test_candidate_suggest_promote_paused(tmp_path, monkeypatch):
+    app, api = _make(tmp_path, monkeypatch, admin_email="boss@x.com")
+    monkeypatch.setattr(api.feeds, "preview_feed",
+                        lambda url, **k: {"url": url, "sampled": 5, "accepted": 4, "examples_accepted": ["A", "B"]})
+    tc = _signin(app, api, "boss@x.com")
+    cand = tc.post("/api/admin/candidates", json={"feed_url": "http://good/feed", "name": "Good Feed"}).json()
+    assert cand["status"] == "suggested" and cand["preview"]["accepted"] == 4
+    cid = cand["id"]
+    assert any(c["id"] == cid for c in tc.get("/api/admin/candidates").json())
+    # promote defaults to paused (active-on-approval off) — no mirror drift
+    res = tc.post(f"/api/admin/candidates/{cid}/promote", json={}).json()
+    assert res["source"]["status"] == "paused" and res["source"]["active"] == 0
+    assert res["candidate"]["status"] == "promoted"
+    assert any(s["name"] == "Good Feed" for s in tc.get("/api/admin/stats").json()["sources"])
+
+
+def test_candidate_reject_and_gating(tmp_path, monkeypatch):
+    app, api = _make(tmp_path, monkeypatch, admin_email="boss@x.com")
+    monkeypatch.setattr(api.feeds, "preview_feed", lambda url, **k: {"url": url, "sampled": 1, "accepted": 0})
+    tc = _signin(app, api, "boss@x.com")
+    cand = tc.post("/api/admin/candidates", json={"feed_url": "http://meh/feed"}).json()
+    assert tc.post(f"/api/admin/candidates/{cand['id']}/reject").json()["status"] == "rejected"
+    anon = TestClient(app)
+    assert anon.get("/api/admin/candidates").status_code == 401
+    assert anon.post("/api/admin/candidates", json={"feed_url": "http://x/f"}).status_code == 401
+    assert anon.post("/api/admin/candidates/1/promote", json={}).status_code == 401
+
+
+def test_safe_fetch_feed_blocks_ssrf():
+    import pytest
+    from goodnews.feeds import safe_fetch_feed
+    for bad in ("http://127.0.0.1/x", "http://localhost/x", "file:///etc/passwd",
+                "http://169.254.169.254/latest", "ftp://x/y"):
+        with pytest.raises(RuntimeError):
+            safe_fetch_feed(bad, timeout=2)