Suya OCR API — vLLM-backed, OpenAI-compatible OCR service

FastAPI service wrapping the Surya-OCR-2 model (datalab-to) served through vLLM:
legacy /v1/api/ai/* endpoints, an OpenAI-compatible /v1/chat/completions endpoint,
a coalescing request batcher, a local OCR CLI, Docker packaging, multilingual
example outputs, and quantization/concurrency benchmarks.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

scripts/parse_timing.py

@@ -0,0 +1,68 @@
+"""Parse surya_timing_summary lines from logs/info.log and aggregate spans.
+
+Usage:
+    python scripts/parse_timing.py logs/info.log
+"""
+from __future__ import annotations
+
+import argparse
+import ast
+import json
+import re
+from collections import defaultdict
+from pathlib import Path
+from typing import Any, Dict, List, Optional
+
+_BATCH_RE = re.compile(r"batch_size=(\d+)")
+_REQID_RE = re.compile(r"request_id=(\S+)")
+
+
+def parse_line(line: str) -> Optional[Dict[str, Any]]:
+    if "surya_timing_summary" not in line or "events=" not in line:
+        return None
+    events_str = line.split("events=", 1)[1].strip()
+    try:
+        events = ast.literal_eval(events_str)
+    except (ValueError, SyntaxError):
+        return None
+    batch_match = _BATCH_RE.search(line)
+    reqid_match = _REQID_RE.search(line)
+    return {
+        "request_id": reqid_match.group(1) if reqid_match else "-",
+        "batch_size": int(batch_match.group(1)) if batch_match else 0,
+        "events": events,
+    }
+
+
+def aggregate(records: List[Dict[str, Any]]) -> Dict[str, Dict[str, float]]:
+    stats: Dict[str, Dict[str, float]] = defaultdict(
+        lambda: {"count": 0, "total_ms": 0.0, "mean_ms": 0.0, "total_tokens": 0}
+    )
+    for rec in records:
+        for event in rec["events"]:
+            s = stats[event["name"]]
+            s["count"] += 1
+            s["total_ms"] += event.get("duration_ms", 0.0)
+            tokens = (event.get("metadata") or {}).get("token_count")
+            if tokens:
+                s["total_tokens"] += tokens
+    for s in stats.values():
+        s["mean_ms"] = round(s["total_ms"] / s["count"], 2) if s["count"] else 0.0
+        s["total_ms"] = round(s["total_ms"], 2)
+    return dict(stats)
+
+
+def main() -> None:
+    parser = argparse.ArgumentParser(description="Aggregate surya timing spans from a log file.")
+    parser.add_argument("logfile", type=Path)
+    args = parser.parse_args()
+    records = [
+        rec
+        for rec in (parse_line(line) for line in args.logfile.read_text(encoding="utf-8").splitlines())
+        if rec is not None
+    ]
+    print(json.dumps({"records": len(records), "spans": aggregate(records)}, indent=2))
+
+
+if __name__ == "__main__":
+    main()