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/schemas.py

@@ -0,0 +1,46 @@
+import base64
+import binascii
+from typing import Any, Literal, Optional
+
+from pydantic import (
+    BaseModel,
+    Field,
+    PrivateAttr,
+    StrictBool,
+    field_validator,
+    model_validator,
+)
+
+
+class Info(BaseModel):
+    _decoded_file: bytes = PrivateAttr(default=b"")
+
+    file: str  # base64 string
+    type: Literal["png", "jpg", "jpeg", "gif"]  # image type like 'png', 'jpg'
+    skip_text_detection: Optional[StrictBool] = Field(False)
+    skip_table_detection: Optional[StrictBool] = Field(False)
+    recognize_math: Optional[StrictBool] = Field(False)
+    ocr_with_boxes: Optional[StrictBool] = Field(True)
+
+    @field_validator("type", mode="before")
+    @classmethod
+    def normalize_type(cls, value: str) -> str:
+        if not isinstance(value, str):
+            raise ValueError("type must be a string")
+        return value.lower()
+
+    @model_validator(mode="after")
+    def decode_base64_file(self):
+        if not self.file:
+            raise ValueError("file must be a non-empty base64 string")
+        try:
+            self._decoded_file = base64.b64decode(self.file, validate=True)
+        except binascii.Error as exc:
+            raise ValueError("file must be valid base64") from exc
+        return self
+
+
+class ApiResponse(BaseModel):
+    data: Any
+    message: str
+    code: int