Advisory source health: review flags, never auto-deactivate

- Add source health columns (last_success_at, last_error_at, last_error,
  consecutive_failures, review_flag, review_reason) via SCHEMA + migration.
- poll_source maintains them: success resets the failure streak and records the
  success time; failure increments it and stores the latest error.
- review_sources() flags active sources that are stale, repeatedly failing,
  low-acceptance, duplicate-heavy, or doom-skewed (high cortisol/ragebait) over
  a recent window. It is purely advisory: it sets review_flag/review_reason and
  never changes the active column (human stays in the loop), clearing the flag
  when a source recovers.
- CLI review-sources; cycle runs it as a final step (--no-review to skip);
  source-report shows a review line for flagged feeds.
- Tests: healthy/failing/stale/low-acceptance/recovery and never-deactivates.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This commit is contained in:
jay
2026-05-30 20:28:18 +00:00
parent aa4125ddec
commit 1e190c5e88
6 changed files with 243 additions and 4 deletions
+80
View File
@@ -3,6 +3,7 @@ from __future__ import annotations
import json
import sqlite3
import tomllib
from datetime import datetime, timezone
from pathlib import Path
from urllib.parse import urlsplit
@@ -153,3 +154,82 @@ def promote_candidate(
row = conn.execute("SELECT id FROM sources WHERE feed_url = ?", (cand["feed_url"],)).fetchone()
return int(row["id"])
# --- Advisory source health: flag for review, never auto-deactivate -----------
def review_sources(
conn: sqlite3.Connection,
stale_days: int = 14,
min_recent: int = 15,
recent_window: int = 40,
) -> list[dict]:
"""Recompute advisory review flags for active sources.
Sets review_flag/review_reason but NEVER changes `active` — the human stays
in the loop. Returns the list of newly-flagged sources.
"""
now = datetime.now(timezone.utc)
flagged = []
sources = conn.execute(
"SELECT id, name, consecutive_failures FROM sources WHERE active = 1"
).fetchall()
for s in sources:
reasons: list[str] = []
if (s["consecutive_failures"] or 0) >= 3:
reasons.append(f"failing ({s['consecutive_failures']} consecutive)")
recent = conn.execute(
"""
SELECT sc.accepted, sc.cortisol_score, sc.ragebait_score, a.duplicate_of,
COALESCE(a.published_at, a.discovered_at) AS dt
FROM articles a
JOIN article_scores sc ON sc.article_id = a.id
WHERE a.source_id = ?
ORDER BY COALESCE(a.published_at, a.discovered_at) DESC
LIMIT ?
""",
(s["id"], recent_window),
).fetchall()
n = len(recent)
if n == 0:
reasons.append("no articles yet")
else:
try:
newest = datetime.fromisoformat(recent[0]["dt"])
if newest.tzinfo is None:
newest = newest.replace(tzinfo=timezone.utc)
age = (now - newest).days
if age > stale_days:
reasons.append(f"stale (newest {age}d ago)")
except (ValueError, TypeError):
pass
if n >= min_recent:
acc = sum(r["accepted"] or 0 for r in recent) / n
if acc < 0.10:
reasons.append(f"low acceptance ({acc * 100:.0f}%)")
dup = sum(1 for r in recent if r["duplicate_of"] is not None) / n
if dup > 0.5:
reasons.append(f"duplicate-heavy ({dup * 100:.0f}%)")
avg_cort = sum(r["cortisol_score"] or 0 for r in recent) / n
if avg_cort > 5:
reasons.append(f"high cortisol (avg {avg_cort:.1f})")
avg_rage = sum(r["ragebait_score"] or 0 for r in recent) / n
if avg_rage > 3:
reasons.append(f"high ragebait (avg {avg_rage:.1f})")
flag = 1 if reasons else 0
reason = "; ".join(reasons) if reasons else None
conn.execute(
"UPDATE sources SET review_flag = ?, review_reason = ? WHERE id = ?",
(flag, reason, s["id"]),
)
if flag:
flagged.append({"id": s["id"], "name": s["name"], "reason": reason})
conn.commit()
return flagged