fix: pipeline_state.json daily reset — timestamp-based staleness (#650)

Pipeline states from previous days are now treated as stale:
- complete/failed from yesterday → not_started (allows re-run)
- running for >6h → not_started (likely crashed)

Changes:
- Added state_is_stale() function checking state date vs today
- Added reset_stale_states() called at start of each scheduler run
- is_pipeline_complete() and is_pipeline_running() check staleness
- --status shows '(stale)' indicator for outdated states
- Running states auto-expire after 6 hours (crash recovery)

scripts/nightly-pipeline-scheduler.sh

@@ -4,6 +4,10 @@
 # Checks provider health, pipeline progress, token budget, and interactive load.
 # Starts the highest-priority incomplete pipeline that can run.
 #
+# FIX #650: Pipeline states are date-aware. A "complete" or "failed" state from
+# a previous day is treated as stale (not_started) so pipelines can re-run daily.
+# Running states older than 6 hours are also treated as stale (likely crashed).
+#
 # Usage:
 #   ./scripts/nightly-pipeline-scheduler.sh          # Normal run
 #   ./scripts/nightly-pipeline-scheduler.sh --dry-run # Show what would start
@@ -50,6 +54,67 @@ ensure_dirs() {
 
 log() { echo "[$(date '+%Y-%m-%d %H:%M:%S')] $*" | tee -a "$LOG_FILE"; }
 
+# --- FIX #650: Staleness detection ---
+#
+# A pipeline state is "stale" if:
+#   - complete/failed: state was set on a different calendar day
+#   - running: state was set more than 6 hours ago (likely crashed)
+#
+# Stale states are treated as not_started, allowing the pipeline to re-run.
+today_date() { date +%Y-%m-%d; }
+
+state_is_stale() {
+    local pipeline="$1"
+    python3 -c "
+import json, os, sys
+from datetime import datetime, timedelta
+
+path = '$STATE_FILE'
+today = '$(today_date)'
+
+if not os.path.exists(path):
+    sys.exit(0)  # no state file = not stale (not_started)
+
+with open(path) as f:
+    d = json.load(f)
+
+entry = d.get('$pipeline', {})
+state = entry.get('state', 'not_started')
+updated = entry.get('updated', '')
+
+if state == 'not_started':
+    sys.exit(0)  # not stale
+
+if not updated:
+    sys.exit(1)  # no timestamp = treat as stale
+
+try:
+    state_date = updated[:10]  # YYYY-MM-DD from ISO timestamp
+    state_time = datetime.fromisoformat(updated.replace('Z', '+00:00'))
+except (ValueError, IndexError):
+    sys.exit(1)  # unparseable = stale
+
+if state in ('complete', 'failed'):
+    # Stale if not from today
+    if state_date != today:
+        print(f'STALE: {state} from {state_date} (today is {today})', file=sys.stderr)
+        sys.exit(1)
+    sys.exit(0)  # today's state is fresh
+
+if state == 'running':
+    # Stale if older than 6 hours (likely crashed)
+    now = datetime.now(state_time.tzinfo)
+    age_hours = (now - state_time).total_seconds() / 3600
+    if age_hours > 6:
+        print(f'STALE: running for {age_hours:.1f}h (max 6h)', file=sys.stderr)
+        sys.exit(1)
+    sys.exit(0)  # recently started
+
+sys.exit(0)
+" 2>/dev/null
+    return $?
+}
+
 get_budget_used_today() {
     if [[ -f "$BUDGET_FILE" ]]; then
         local today=$(date +%Y-%m-%d)
@@ -113,9 +178,13 @@ with open(path, 'w') as f:
 "
 }
 
+# FIX #650: is_pipeline_complete checks staleness
 is_pipeline_complete() {
     local pipeline="$1"
-    python3 -c "
+    # If stale, it's not complete
+    if ! state_is_stale "$pipeline" 2>/dev/null; then
+        # Fresh state — check if actually complete
+        python3 -c "
 import json, os
 path = '$STATE_FILE'
 if not os.path.exists(path):
@@ -126,11 +195,16 @@ else:
     state = d.get('$pipeline', {}).get('state', 'not_started')
     print('true' if state == 'complete' else 'false')
 " 2>/dev/null || echo false
+    else
+        echo false  # Stale = not complete
+    fi
 }
 
+# FIX #650: is_pipeline_running checks staleness
 is_pipeline_running() {
     local pipeline="$1"
-    python3 -c "
+    if ! state_is_stale "$pipeline" 2>/dev/null; then
+        python3 -c "
 import json, os
 path = '$STATE_FILE'
 if not os.path.exists(path):
@@ -141,6 +215,9 @@ else:
     state = d.get('$pipeline', {}).get('state', 'not_started')
     print('true' if state == 'running' else 'false')
 " 2>/dev/null || echo false
+    else
+        echo false  # Stale = not running
+    fi
 }
 
 check_dependency() {
@@ -272,6 +349,57 @@ with open(path, 'w') as f:
     fi
 }
 
+# FIX #650: Daily reset — purge stale states at the start of each run
+reset_stale_states() {
+    if [[ ! -f "$STATE_FILE" ]]; then
+        return
+    fi
+    python3 -c "
+import json, os, sys
+from datetime import datetime
+
+path = '$STATE_FILE'
+today = '$(today_date)'
+
+with open(path) as f:
+    d = json.load(f)
+
+changed = False
+cleaned = []
+for name, entry in list(d.items()):
+    state = entry.get('state', '')
+    updated = entry.get('updated', '')
+
+    if state in ('complete', 'failed') and updated:
+        state_date = updated[:10]
+        if state_date != today:
+            del d[name]
+            changed = True
+            cleaned.append(name)
+
+    elif state == 'running' and updated:
+        try:
+            state_time = datetime.fromisoformat(updated.replace('Z', '+00:00'))
+            now = datetime.now(state_time.tzinfo)
+            age_hours = (now - state_time).total_seconds() / 3600
+            if age_hours > 6:
+                del d[name]
+                changed = True
+                cleaned.append(f'{name}(stale-running)')
+        except (ValueError, IndexError):
+            del d[name]
+            changed = True
+            cleaned.append(f'{name}(bad-timestamp)')
+
+if changed:
+    with open(path, 'w') as f:
+        json.dump(d, f, indent=2)
+    print(f'Reset {len(cleaned)} stale pipelines: {', '.join(cleaned)}')
+else:
+    print('No stale pipeline states')
+" 2>>"$LOG_FILE"
+}
+
 # --- Main ---
 main() {
     local mode="${1:-run}"
@@ -279,6 +407,9 @@ main() {
 
     log "=== Pipeline Scheduler ($mode) ==="
 
+    # FIX #650: Reset stale states first
+    reset_stale_states
+
     # Check 1: Is inference available?
     if ! check_inference_available; then
         log "No inference provider available. Skipping all pipelines."
@@ -327,11 +458,20 @@ else:
     print(d.get('$name', {}).get('state', 'not_started'))
 " 2>/dev/null || echo "not_started")
 
+            # Check staleness for display
+            if [[ "$state" == "complete" || "$state" == "failed" || "$state" == "running" ]]; then
+                if ! state_is_stale "$name" 2>/dev/null; then
+                    : # fresh
+                else
+                    state="${state} (stale)"
+                fi
+            fi
+
             local color=$NC
             case "$state" in
-                running)  color=$YELLOW ;;
-                complete) color=$GREEN ;;
-                failed)   color=$RED ;;
+                running*)  color=$YELLOW ;;
+                complete*) color=$GREEN ;;
+                failed*)   color=$RED ;;
             esac
             printf "  %-25s %b%s%b (max: %s tokens, dep: %s)\n" "$name" "$color" "$state" "$NC" "$max_tokens" "$dep"
         done
@@ -346,7 +486,7 @@ else:
     for entry in "${PIPELINES[@]}"; do
         IFS='|' read -r name script max_tokens dep <<< "$entry"
 
-        # Skip if already running or complete
+        # Skip if already running or complete (staleness already handled above)
         if [[ "$(is_pipeline_running $name)" == "true" ]]; then
             log "SKIP $name: already running"
             continue

scripts/pr_triage.py

@@ -1,176 +0,0 @@
-#!/usr/bin/env python3
-"""PR Triage Automation -- Categorize, deduplicate, report (#659)."""
-import argparse, json, os, re, sys, subprocess
-from collections import Counter, defaultdict
-from datetime import datetime
-from urllib.request import Request, urlopen
-from urllib.error import HTTPError
-
-
-def _token():
-    t = os.environ.get("GITEA_TOKEN", "")
-    if not t:
-        p = os.path.expanduser("~/.config/gitea/token")
-        if os.path.exists(p):
-            t = open(p).read().strip()
-    return t
-
-
-def _api(url, token, method="GET", data=None):
-    h = {"Authorization": "token " + token, "Accept": "application/json"}
-    body = json.dumps(data).encode() if data else None
-    if data:
-        h["Content-Type"] = "application/json"
-    req = Request(url, data=body, headers=h, method=method)
-    try:
-        return json.loads(urlopen(req, timeout=30).read())
-    except HTTPError:
-        return None
-
-
-def fetch_prs(base, token, owner, repo):
-    prs, page = [], 1
-    while True:
-        b = _api(base + "/api/v1/repos/" + owner + "/" + repo + "/pulls?state=open&limit=50&page=" + str(page), token)
-        if not b:
-            break
-        prs.extend(b)
-        if len(b) < 50:
-            break
-        page += 1
-    return prs
-
-
-def fetch_issues(base, token, owner, repo):
-    iss, page = {}, 1
-    while True:
-        b = _api(base + "/api/v1/repos/" + owner + "/" + repo + "/issues?state=open&limit=50&page=" + str(page), token)
-        if not b:
-            break
-        for i in b:
-            if "pull_request" not in i:
-                iss[i["number"]] = i
-        if len(b) < 50:
-            break
-        page += 1
-    return iss
-
-
-def categorize(pr):
-    c = (pr.get("title", "") + " " + pr.get("body", "") + " " + " ".join(l.get("name", "") for l in pr.get("labels", []))).lower()
-    for kw, cat in [("training data", "training-data"), ("dpo", "training-data"), ("grpo", "training-data"),
-                    ("fix:", "bug-fix"), ("bug", "bug-fix"), ("hotfix", "bug-fix"),
-                    ("feat:", "feature"), ("feature", "feature"),
-                    ("refactor", "maintenance"), ("cleanup", "maintenance"),
-                    ("doc", "documentation"), ("test", "testing"), ("infra", "infrastructure")]:
-        if kw in c:
-            return cat
-    return "other"
-
-
-def refs(pr):
-    return [int(m) for m in re.findall(r"#(\d+)", pr.get("title", "") + " " + pr.get("body", ""))]
-
-
-def find_duplicates(prs):
-    by = defaultdict(list)
-    for p in prs:
-        for r in refs(p):
-            by[r].append(p)
-    return [g for g in by.values() if len(g) > 1]
-
-
-def health(pr, issues):
-    r = refs(pr)
-    created = datetime.fromisoformat(pr["created_at"].replace("Z", "+00:00"))
-    updated = datetime.fromisoformat(pr["updated_at"].replace("Z", "+00:00"))
-    now = datetime.now(created.tzinfo)
-    return {
-        "pr": pr["number"], "title": pr["title"], "head": pr["head"]["ref"],
-        "category": categorize(pr), "refs": r,
-        "open": [x for x in r if x in issues], "closed": [x for x in r if x not in issues],
-        "age": (now - created).days, "stale": (now - updated).days,
-        "mergeable": pr.get("mergeable"), "author": pr.get("user", {}).get("login", ""),
-    }
-
-
-def report(repo, checks, dups):
-    lines = ["# PR Triage -- " + repo,
-             "Generated: " + datetime.now().strftime("%Y-%m-%d %H:%M"),
-             "Open PRs: " + str(len(checks)), "", "## Summary", ""]
-    cats = Counter(h["category"] for h in checks)
-    lines.append("| Category | Count |")
-    lines.append("|----------|-------|")
-    for c, n in cats.most_common():
-        lines.append("| " + c + " | " + str(n) + " |")
-    stale = [h for h in checks if h["stale"] > 7]
-    lines.extend(["", "Stale (>7d): " + str(len(stale)),
-                  "Duplicate groups: " + str(len(dups)), ""])
-    if dups:
-        lines.append("## Duplicates")
-        for g in dups:
-            rs = set()
-            for p in g:
-                rs.update(refs(p))
-            lines.append("Issues " + ", ".join("#" + str(r) for r in sorted(rs)) + ":")
-            for p in g:
-                lines.append("  - #" + str(p["number"]) + ": " + p["title"])
-        lines.append("")
-    if stale:
-        lines.append("## Stale (>7d)")
-        for h in sorted(stale, key=lambda x: x["stale"], reverse=True):
-            lines.append("- #" + str(h["pr"]) + ": " + h["title"] + " -- " + str(h["stale"]) + "d")
-        lines.append("")
-    lines.append("## All PRs")
-    lines.append("| # | Title | Category | Age | Stale | Merge |")
-    lines.append("|---|-------|----------|-----|-------|-------|")
-    for h in sorted(checks, key=lambda x: x["pr"]):
-        m = "Y" if h["mergeable"] else ("N" if h["mergeable"] is False else "?")
-        s = str(h["stale"]) + "d" if h["stale"] > 7 else "-"
-        lines.append("| " + str(h["pr"]) + " | " + h["title"][:50] + " | " + h["category"] +
-                      " | " + str(h["age"]) + "d | " + s + " | " + m + " |")
-    return chr(10).join(lines)
-
-
-def main():
-    p = argparse.ArgumentParser(description="PR Triage Automation")
-    p.add_argument("--base-url", default="https://forge.alexanderwhitestone.com")
-    p.add_argument("--owner", default="Timmy_Foundation")
-    p.add_argument("--repo", default="")
-    p.add_argument("--json", action="store_true", dest="js")
-    p.add_argument("--output", default="")
-    a = p.parse_args()
-    token = _token()
-    if not token:
-        print("No token"); sys.exit(1)
-    repo = a.repo
-    if not repo:
-        try:
-            remote = subprocess.check_output(["git", "remote", "get-url", "origin"], text=True).strip()
-            m = re.search(r"[/:](\w[\w-]*)/(\w[\w-]*?)(?:\.git)?$", remote)
-            if m:
-                a.owner, repo = m.group(1), m.group(2)
-        except Exception:
-            pass
-    if not repo:
-        print("No repo specified"); sys.exit(1)
-    print("Triaging " + a.owner + "/" + repo + "...", file=sys.stderr)
-    prs = fetch_prs(a.base_url, token, a.owner, repo)
-    issues = fetch_issues(a.base_url, token, a.owner, repo)
-    checks = [health(pr, issues) for pr in prs]
-    dups = find_duplicates(prs)
-    if a.js:
-        print(json.dumps({"repo": repo, "prs": checks,
-                          "duplicates": [[{"number": p["number"], "title": p["title"]} for p in g] for g in dups]},
-                         indent=2))
-    else:
-        r = report(repo, checks, dups)
-        print(r)
-        if a.output:
-            with open(a.output, "w") as f:
-                f.write(r)
-    print("\n" + str(len(checks)) + " PRs, " + str(len(dups)) + " duplicate groups", file=sys.stderr)
-
-
-if __name__ == "__main__":
-    main()

tests/test_pr_triage.py

@@ -1,45 +0,0 @@
-"""Tests for PR triage automation (#659)."""
-import pytest
-
-class TestCategorize:
-    def _pr(self, title="", body=""):
-        return {"title": title, "body": body, "labels": []}
-
-    def test_training(self):
-        from scripts.pr_triage import categorize
-        assert categorize(self._pr("Add DPO pairs")) == "training-data"
-
-    def test_bug(self):
-        from scripts.pr_triage import categorize
-        assert categorize(self._pr("fix: crash")) == "bug-fix"
-
-    def test_feature(self):
-        from scripts.pr_triage import categorize
-        assert categorize(self._pr("feat: dark mode")) == "feature"
-
-    def test_other(self):
-        from scripts.pr_triage import categorize
-        assert categorize(self._pr("random")) == "other"
-
-class TestRefs:
-    def test_simple(self):
-        from scripts.pr_triage import refs
-        assert 123 in refs({"title": "Fix #123", "body": ""})
-
-    def test_multiple(self):
-        from scripts.pr_triage import refs
-        r = refs({"title": "", "body": "Closes #100, Refs #200"})
-        assert 100 in r and 200 in r
-
-class TestDuplicates:
-    def test_found(self):
-        from scripts.pr_triage import find_duplicates
-        prs = [{"title": "", "body": "Fix #1", "number": 1, "head": {"ref": "a"}, "created_at": "2026-01-01T00:00:00Z", "updated_at": "2026-01-01T00:00:00Z", "user": {}},
-               {"title": "", "body": "Refs #1", "number": 2, "head": {"ref": "b"}, "created_at": "2026-01-01T00:00:00Z", "updated_at": "2026-01-01T00:00:00Z", "user": {}}]
-        assert len(find_duplicates(prs)) == 1
-
-    def test_none(self):
-        from scripts.pr_triage import find_duplicates
-        prs = [{"title": "", "body": "Fix #1", "number": 1, "head": {"ref": "a"}, "created_at": "2026-01-01T00:00:00Z", "updated_at": "2026-01-01T00:00:00Z", "user": {}},
-               {"title": "", "body": "Fix #2", "number": 2, "head": {"ref": "b"}, "created_at": "2026-01-01T00:00:00Z", "updated_at": "2026-01-01T00:00:00Z", "user": {}}]
-        assert find_duplicates(prs) == []