fix: 4 small bugs found during health check

1. orchestrate line 534: 'local' outside function — replaced with bare assignment (no functional impact, just stderr noise)
2. status server /logs allowlist: add surrogate-dev-loop, scrape-keyword-tuner, hermes-discord-bot, ollama-pull-* (was missing 8 logs)
3. status server _ollama_models(): cache 30s + non-blocking timeout 1.5s — was causing /status to hang when ollama busy
4. domain-scrape-loop.sh:
- init scrape-ledger.db if missing (was producing empty BEFORE_LEDGER → arithmetic errors)
- default 0 for all sqlite3 queries (was breaking iter progress lines)
- Linux-compat 'load average:' parsing (was macOS-only)
- update training-jsonl path to ~/.surrogate/training-pairs.jsonl

VERIFIED working today:
- Pipeline: 50+ commits to dataset (last 27 min: ~3411 pairs uploaded)
- Dev agent committed b10dd74 to arkashira/vanguard:
backend/app/api/v1/endpoints/cloud_rules.py (5588 bytes via DeepSeek-V3.1 HF Router)
- Agentic crawler: 40735 URLs visited, 17695 in frontier
- 5 Ollama models pulling
- 25 daemons logging

bin/domain-scrape-loop.sh

@@ -10,8 +10,18 @@ DUR="${1:-900}"
 PARALLEL="${2:-3}"
 LOG="$HOME/.surrogate/logs/domain-scrape-loop.log"
 START=$(date +%s)
-BEFORE_PAIRS=$(wc -l ~/axentx/surrogate/data/training-jsonl/*.jsonl 2>/dev/null | tail -1 | awk '{print $1}')
-BEFORE_LEDGER=$(sqlite3 ~/.surrogate/state/scrape-ledger.db "SELECT COUNT(*) FROM scraped" 2>/dev/null)
+
+# Initialize ledger if missing (creates the 'scraped' table)
+LEDGER_DB="$HOME/.surrogate/state/scrape-ledger.db"
+if [[ ! -f "$LEDGER_DB" ]] || ! sqlite3 "$LEDGER_DB" "SELECT 1 FROM scraped LIMIT 1" >/dev/null 2>&1; then
+    bash "$HOME/.surrogate/bin/scrape-ledger-init.sh" 2>>"$LOG"
+fi
+
+# Default 0 if query fails (was causing empty arithmetic in iter logs)
+BEFORE_PAIRS=$(wc -l "$HOME/.surrogate/training-pairs.jsonl" 2>/dev/null | awk '{print $1}')
+BEFORE_PAIRS=${BEFORE_PAIRS:-0}
+BEFORE_LEDGER=$(sqlite3 "$LEDGER_DB" "SELECT COUNT(*) FROM scraped" 2>/dev/null)
+BEFORE_LEDGER=${BEFORE_LEDGER:-0}
 
 echo "═══ LOOP START $(date +%H:%M:%S) duration=${DUR}s parallel=$PARALLEL" | tee -a "$LOG"
 echo "   before: pairs=$BEFORE_PAIRS ledger_repos=$BEFORE_LEDGER" | tee -a "$LOG"
@@ -22,8 +32,9 @@ while true; do
     [[ $((NOW - START)) -gt $DUR ]] && break
     ITER=$((ITER + 1))
 
-    # Health check — break if load > 10
-    LOAD=$(uptime | awk -F'load averages:' '{print $2}' | awk '{print int($1)}')
+    # Health check — break if load > 10 (Linux: "load average:", macOS: "load averages:")
+    LOAD=$(uptime | sed -E 's/.*load average[s]?:[[:space:]]*//' | awk -F',' '{print int($1)}')
+    LOAD=${LOAD:-0}
     if [[ $LOAD -gt 10 ]]; then
         echo "  [iter=$ITER] load=$LOAD pause 30s" | tee -a "$LOG"
         sleep 30
@@ -43,14 +54,18 @@ while true; do
 
     # Progress every 5 iters
     if (( ITER % 5 == 0 )); then
-        PAIRS=$(wc -l ~/axentx/surrogate/data/training-jsonl/*.jsonl 2>/dev/null | tail -1 | awk '{print $1}')
-        LEDGER=$(sqlite3 ~/.surrogate/state/scrape-ledger.db "SELECT COUNT(*) FROM scraped" 2>/dev/null)
+        PAIRS=$(wc -l "$HOME/.surrogate/training-pairs.jsonl" 2>/dev/null | awk '{print $1}')
+        PAIRS=${PAIRS:-0}
+        LEDGER=$(sqlite3 "$LEDGER_DB" "SELECT COUNT(*) FROM scraped" 2>/dev/null)
+        LEDGER=${LEDGER:-0}
         echo "  [iter=$ITER $((NOW - START))s] pairs=$PAIRS (+$((PAIRS - BEFORE_PAIRS))) ledger=$LEDGER (+$((LEDGER - BEFORE_LEDGER)))" | tee -a "$LOG"
     fi
 done
 
-AFTER_PAIRS=$(wc -l ~/axentx/surrogate/data/training-jsonl/*.jsonl 2>/dev/null | tail -1 | awk '{print $1}')
-AFTER_LEDGER=$(sqlite3 ~/.surrogate/state/scrape-ledger.db "SELECT COUNT(*) FROM scraped" 2>/dev/null)
+AFTER_PAIRS=$(wc -l "$HOME/.surrogate/training-pairs.jsonl" 2>/dev/null | awk '{print $1}')
+AFTER_PAIRS=${AFTER_PAIRS:-0}
+AFTER_LEDGER=$(sqlite3 "$LEDGER_DB" "SELECT COUNT(*) FROM scraped" 2>/dev/null)
+AFTER_LEDGER=${AFTER_LEDGER:-0}
 echo "═══ LOOP DONE $(date +%H:%M:%S)" | tee -a "$LOG"
 echo "   iters: $ITER" | tee -a "$LOG"
 echo "   pairs added:  $((AFTER_PAIRS - BEFORE_PAIRS))" | tee -a "$LOG"

bin/hermes-status-server.py

@@ -102,10 +102,23 @@ def _agentic_visited() -> int:
 
 
 def _ollama_models() -> list[str]:
+    """Quick (non-blocking) check of loaded Ollama models. Caches for 30s."""
+    cache = HOME / ".surrogate/state/.ollama-models-cache.json"
     try:
-        import urllib.request, json as _json
-        with urllib.request.urlopen("http://127.0.0.1:11434/api/tags", timeout=2) as r:
-            return [m["name"] for m in _json.load(r).get("models", [])]
+        import json as _json, time
+        if cache.exists():
+            cached = _json.loads(cache.read_text())
+            if time.time() - cached.get("ts", 0) < 30:
+                return cached.get("models", [])
+    except Exception:
+        pass
+    try:
+        import urllib.request, json as _json, time
+        with urllib.request.urlopen("http://127.0.0.1:11434/api/tags", timeout=1.5) as r:
+            models = [m["name"] for m in _json.load(r).get("models", [])]
+        cache.parent.mkdir(parents=True, exist_ok=True)
+        cache.write_text(_json.dumps({"ts": time.time(), "models": models}))
+        return models
     except Exception:
         return []
 
@@ -136,10 +149,14 @@ def health() -> dict:
 def log_tail(name: str, lines: int = 100) -> PlainTextResponse:
     """Tail a specific log file. Allowlist for security."""
     allowed = {
-        "boot", "cron", "scrape-continuous", "scrape-daemon",
-        "agentic-crawler", "skill-synthesis", "auto-orchestrate-loop",
-        "training-push", "ollama", "discord-bot", "surrogate-research-loop",
-        "domain-scrape", "qwen-coder", "git-clone", "git-pull",
+        "boot", "cron", "cron-master", "scrape-continuous", "scrape-daemon",
+        "scrape-keyword-tuner", "agentic-crawler", "skill-synthesis",
+        "auto-orchestrate-loop", "training-push", "ollama", "discord-bot",
+        "hermes-discord-bot", "surrogate-research-loop", "surrogate-research-apply",
+        "surrogate-dev-loop", "domain-scrape-loop", "github-domain-scrape",
+        "qwen-coder", "git-clone", "git-pull", "redis",
+        "ollama-pull-coder", "ollama-pull-devstral", "ollama-pull-fallback",
+        "ollama-pull-yicoder", "ollama-pull-embed", "ollama-pull-light",
     }
     if name not in allowed:
         raise HTTPException(404, f"Unknown log: {name}. Allowed: {sorted(allowed)}")

bin/surrogate-orchestrate.sh

@@ -531,8 +531,8 @@ if echo "$VERDICT_TEXT" | grep -qi "APPROVE"; then
     echo "${GR}${B}▸ Reviewer approved — committing changes${R}"
     if ! git -C "$(pwd)" diff --quiet 2>/dev/null || ! git -C "$(pwd)" diff --cached --quiet 2>/dev/null; then
         git -C "$(pwd)" add -A 2>/dev/null
-        local short_task; short_task=$(echo "$TASK" | head -c 72)
-        if git -C "$(pwd)" commit -m "feat: $short_task
+        SHORT_TASK=$(echo "$TASK" | head -c 72)
+        if git -C "$(pwd)" commit -m "feat: $SHORT_TASK
 
 [surrogate auto-dev session $SESSION_ID]
 [reviewed: APPROVE]" 2>&1 | tee -a "$WORKDIR/git-commit.log" | grep -q "master\|main\|\["; then