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>

tests/test_parse_timing.py

@@ -0,0 +1,30 @@
+from scripts.parse_timing import parse_line, aggregate
+
+SAMPLE = (
+    "2026-06-11 14:35:00 - vllm_batcher - _process_jobs - line:119 - INFO - "
+    "surya_timing_summary request_id=a,b batch_size=2 events=["
+    "{'name': 'openai_chat_completion', 'duration_ms': 100.0, 'metadata': {'token_count': 40}}, "
+    "{'name': 'openai_chat_completion', 'duration_ms': 300.0, 'metadata': {'token_count': 60}}, "
+    "{'name': 'recognition_manager_generate', 'duration_ms': 450.0}]"
+)
+
+
+def test_parse_line_extracts_events_and_batch_size():
+    rec = parse_line(SAMPLE)
+    assert rec["batch_size"] == 2
+    assert len(rec["events"]) == 3
+    assert rec["events"][0]["metadata"]["token_count"] == 40
+
+
+def test_parse_line_returns_none_for_unrelated_line():
+    assert parse_line("2026-06-11 - foo - bar - line:1 - INFO - request_start id=x") is None
+
+
+def test_aggregate_counts_and_sums_by_span_name():
+    rec = parse_line(SAMPLE)
+    agg = aggregate([rec])
+    chat = agg["openai_chat_completion"]
+    assert chat["count"] == 2
+    assert chat["total_ms"] == 400.0
+    assert chat["mean_ms"] == 200.0
+    assert chat["total_tokens"] == 100