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>

surya/endpoint/app.py

@@ -0,0 +1,97 @@
+import json
+import logging
+import logging.config
+import time
+import uuid
+
+import yaml
+from fastapi import FastAPI, Request
+from fastapi.exceptions import RequestValidationError
+from fastapi.responses import JSONResponse
+
+from surya.endpoint.legacy import router as legacy_router
+from surya.endpoint.openai import router as openai_router
+from surya.endpoint.service import _error_response
+
+with open("logger.yaml", "r", encoding="utf-8") as f:
+    config = yaml.safe_load(f.read())
+    logging.config.dictConfig(config)
+
+logger = logging.getLogger(__name__)
+
+app = FastAPI(
+    title="OCR SERVICE API SERVICE",
+    version="1.0",
+    docs_url="/v1/api/ai/swagger",
+    openapi_url="/v1/api/ai/openapi.json",
+)
+
+
+@app.exception_handler(RequestValidationError)
+async def validation_exception_handler(request: Request, exc: RequestValidationError):
+    request_id = getattr(request.state, "request_id", "-")
+    logger.warning(
+        "request_validation_failed request_id=%s method=%s path=%s errors=%s",
+        request_id,
+        request.method,
+        request.url.path,
+        exc.errors(),
+    )
+    return JSONResponse(
+        status_code=200,
+        content={
+            "data": [],
+            "message": json.dumps({"error info": exc.errors()}, default=str),
+            "code": 422,
+        },
+    )
+
+
+@app.middleware("http")
+async def allow_openapi_unauthorized(request: Request, call_next):
+    if request.url.path == "/v1/api/ai/openapi.json":
+        response = await call_next(request)
+        return response
+    return await call_next(request)
+
+
+@app.middleware("http")
+async def log_request_response(request, call_next):
+    request_id = uuid.uuid4().hex
+    request.state.request_id = request_id
+    start = time.perf_counter()
+    client = request.client.host if request.client else "-"
+    logger.info(
+        "request_start request_id=%s method=%s path=%s client=%s",
+        request_id,
+        request.method,
+        request.url.path,
+        client,
+    )
+    try:
+        response = await call_next(request)
+    except Exception as e:
+        duration_ms = (time.perf_counter() - start) * 1000
+        logger.exception(
+            "request_unhandled_exception request_id=%s method=%s path=%s duration_ms=%.2f",
+            request_id,
+            request.method,
+            request.url.path,
+            duration_ms,
+            exc_info=True,
+        )
+        return JSONResponse(status_code=500, content=_error_response(e))
+    duration_ms = (time.perf_counter() - start) * 1000
+    logger.info(
+        "request_end request_id=%s method=%s path=%s status_code=%s duration_ms=%.2f",
+        request_id,
+        request.method,
+        request.url.path,
+        response.status_code,
+        duration_ms,
+    )
+    return response
+
+
+app.include_router(legacy_router)
+app.include_router(openai_router)