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>

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)