n-seo 0.7.0 → 0.7.1

package/CHANGELOG.md

@@ -8,6 +8,27 @@ uses [Semantic Versioning](https://semver.org/).
 
 Nothing yet.
 
+## [0.7.1] - 2026-09-28
+
+### Fixed
+- **One HTTP/2 hiccup no longer fails the whole index-status step.** The
+  retry helper only retried the curl exits it knew about, and treated every
+  other one as a bad request. A dropped HTTP/2 connection (curl exit 16)
+  mid-run failed the first URL Inspection call it hit, which took the step
+  down, and the daily run's one step-level retry hit the same thing. Exits
+  16 and 92 (HTTP/2 connection and stream errors) and 55 (send failure) now
+  get the same per-request backoff as timeouts and resets.
+- **A curl failure now says why.** Requests ran with `-s`, which silences
+  curl's own error line along with the progress meter, so the failure read
+  `curl exit 16: ` with nothing after it. They run with `-sS` now.
+- **A slow first answer no longer reads as a regression.** The site probe
+  fetched each file once with a 15-second limit. A cold origin behind a CDN
+  can miss that on its first request of the morning, and the run then
+  alerted `robots.txt REGRESSED` for a site serving it fine. It happened
+  every few days to one site. The probe now tries once more, five seconds
+  later, when there is no answer or a 5xx. A site that is really down fails
+  both tries and is still reported.
+
 ## [0.7.0] - 2026-09-18
 
 ### Added
@@ -537,7 +558,9 @@ First public release.
   `n-seo upgrade` gate — without changing the reported test counts, and an
   in-place install ran those assertions against the owner's live data.
 
-[Unreleased]: https://github.com/en-dash-consulting/n-seo/compare/v0.6.1...HEAD
+[Unreleased]: https://github.com/en-dash-consulting/n-seo/compare/v0.7.1...HEAD
+[0.7.1]: https://github.com/en-dash-consulting/n-seo/compare/v0.7.0...v0.7.1
+[0.7.0]: https://github.com/en-dash-consulting/n-seo/compare/v0.6.1...v0.7.0
 [0.6.1]: https://github.com/en-dash-consulting/n-seo/compare/v0.6.0...v0.6.1
 [0.6.0]: https://github.com/en-dash-consulting/n-seo/compare/v0.5.0...v0.6.0
 [0.5.0]: https://github.com/en-dash-consulting/n-seo/compare/v0.4.1...v0.5.0

package/ingest/http_util.py

@@ -14,9 +14,10 @@ import subprocess
 import sys
 import time
 
-# curl exits worth another try: 6/7 resolve+connect, 28 timeout,
-# 35 TLS handshake, 52 empty reply, 56 recv error.
-RETRYABLE_EXITS = {6, 7, 28, 35, 52, 56}
+# curl exits worth another try: 6/7 resolve+connect, 16 HTTP/2 framing,
+# 28 timeout, 35 TLS handshake, 52 empty reply, 55 send error,
+# 56 recv error, 92 HTTP/2 stream reset.
+RETRYABLE_EXITS = {6, 7, 16, 28, 35, 52, 55, 56, 92}
 TIMEOUT = 180
 ATTEMPTS = 4
 
@@ -29,7 +30,8 @@ def curl_json(args, *, timeout=TIMEOUT, attempts=ATTEMPTS, label=""):
     """
     last = None
     for attempt in range(1, attempts + 1):
-        p = subprocess.run(["curl", "-s", "--max-time", str(timeout), *args],
+        # -sS: no progress meter, but curl still says why it failed.
+        p = subprocess.run(["curl", "-sS", "--max-time", str(timeout), *args],
                            capture_output=True, text=True)
         if p.returncode == 0:
             try:

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "n-seo",
-  "version": "0.7.0",
+  "version": "0.7.1",
   "description": "Agentic, local-first SEO / AEO / GEO control plane: pulls Search Console + GA4, probes your sites, and ranks the next moves. Your LLM writes the proposals; your coding agent works the queue over MCP.",
   "license": "MIT",
   "author": "En Dash Consulting (https://endash.us)",

package/probes/site_probe.py

@@ -13,6 +13,7 @@ Stdlib only — no pip installs needed.
 import json
 import re
 import sys
+import time
 from datetime import datetime, timezone
 from html.parser import HTMLParser
 from pathlib import Path
@@ -33,7 +34,15 @@ UA = "Mozilla/5.0 (compatible; n-seo-probe/0.1)"
 
 
 def fetch(url, timeout=15):
-    return fetch_text(url, timeout=timeout, ua=UA)
+    """One retry when the site doesn't answer or answers 5xx. A cold origin
+    behind a CDN can take longer than the timeout on its first request of the
+    morning, and without the retry that read as robots.txt or sitemap.xml
+    REGRESSED. A site that is really down fails both tries."""
+    status, body = fetch_text(url, timeout=timeout, ua=UA)
+    if status is None or status >= 500:
+        time.sleep(5)
+        status, body = fetch_text(url, timeout=timeout, ua=UA)
+    return status, body
 
 
 class MetaParser(HTMLParser):