feat: add Runpod serverless image collector pools

2026-10-11 04:55:01 +00:00
parent 6f85129e45
commit 1093818327
42 changed files with 5811 additions and 238 deletions
@@ -0,0 +1,57 @@
+"""Private, atomic job-ID journal. Contains identifiers/status only, never payloads."""
+import json
+import os
+from pathlib import Path
+import tempfile
+import threading
+
+TERMINAL = {"COMPLETED", "FAILED", "CANCELLED", "TIMED_OUT"}
+FIELDS = {"record_id", "execution_id", "worker_id", "participant_id", "endpoint_id", "job_id", "status", "created_at", "updated_at", "error", "prompt_id"}
+
+
+class JobLedger:
+    def __init__(self, path):
+        self.path = Path(path)
+        self._lock = threading.RLock()
+
+    def _read(self):
+        if self.path.is_symlink():
+            raise RuntimeError("Unsafe serverless job journal path")
+        if not self.path.exists():
+            return {}
+        try:
+            data = json.loads(self.path.read_text(encoding="utf-8"))
+            if not isinstance(data, dict) or any(not isinstance(v, dict) or set(v) - FIELDS for v in data.values()):
+                raise ValueError()
+            return data
+        except (ValueError, OSError):
+            raise RuntimeError("Serverless job journal is unreadable; preserve it for recovery before submitting more work") from None
+
+    def all(self):
+        with self._lock:
+            return list(self._read().values())
+
+    def update(self, record_id, **fields):
+        if set(fields) - FIELDS:
+            raise ValueError("Unexpected serverless journal field")
+        with self._lock:
+            data = self._read()
+            data.setdefault(record_id, {"record_id": record_id}).update(fields)
+            # Keep every unresolved job, plus a bounded terminal history.
+            completed = sorted((k for k, v in data.items() if v.get("status") in TERMINAL), key=lambda k: data[k].get("updated_at", 0))
+            for key in completed[:-200]:
+                data.pop(key)
+            if self.path.parent.is_symlink():
+                raise RuntimeError("Unsafe serverless journal directory")
+            self.path.parent.mkdir(mode=0o700, parents=True, exist_ok=True)
+            fd, temporary = tempfile.mkstemp(prefix=".jobs-", dir=self.path.parent)
+            try:
+                with os.fdopen(fd, "w", encoding="utf-8") as stream:
+                    json.dump(data, stream, allow_nan=False)
+                    stream.flush()
+                    os.fsync(stream.fileno())
+                os.replace(temporary, self.path)
+            finally:
+                if os.path.exists(temporary):
+                    os.unlink(temporary)
+            return dict(data[record_id])