fix(decoder): auto-detect gzip magic bytes for responses without Content-Encoding header

Co-Authored-By: syed.khadeer@airbyte.io <cloud-support@airbyte.io>

Author: devin-ai-integration[bot]

airbyte_cdk/sources/declarative/decoders/composite_raw_decoder.py

@@ -11,6 +11,8 @@
 from io import BufferedIOBase, TextIOWrapper
 from typing import Any, List, Optional
 
+GZIP_MAGIC_BYTES = b"\x1f\x8b"
+
 import orjson
 import requests
 
@@ -35,15 +37,22 @@ def parse(self, data: BufferedIOBase) -> PARSER_OUTPUT_TYPE:
         """
         Decompress gzipped bytes and pass decompressed data to the inner parser.
 
-        IMPORTANT:
-            - If the data is not gzipped, reset the pointer and pass the data to the inner parser as is.
-
-        Note:
-            - The data is not decoded by default.
+        Auto-detects gzip content by checking for magic bytes (1f 8b) at the start of the data.
+        If the data is not gzip-compressed, it is passed directly to the inner parser as-is.
+        This handles APIs that return gzip-compressed bodies without setting the Content-Encoding header.
         """
+        header = data.read(2)
+        if not header:
+            return
 
-        with gzip.GzipFile(fileobj=data, mode="rb") as gzipobj:
-            yield from self.inner_parser.parse(gzipobj)
+        remaining = data.read()
+        full_data = io.BytesIO(header + remaining)
+
+        if header == GZIP_MAGIC_BYTES:
+            with gzip.GzipFile(fileobj=full_data, mode="rb") as gzipobj:
+                yield from self.inner_parser.parse(gzipobj)
+        else:
+            yield from self.inner_parser.parse(full_data)
 
 
 @dataclass

airbyte_cdk/sources/declarative/parsers/model_to_component_factory.py

@@ -2634,15 +2634,16 @@ def create_gzip_decoder(
         gzip_parser: GzipParser = ModelToComponentFactory._get_parser(model, config)  # type: ignore  # based on the model, we know this will be a GzipParser
 
         if self._emit_connector_builder_messages:
-            # This is very surprising but if the response is not streamed,
-            # CompositeRawDecoder calls response.content and the requests library actually uncompress the data as opposed to response.raw,
-            # which uses urllib3 directly and does not uncompress the data.
-            return CompositeRawDecoder(gzip_parser.inner_parser, False)
+            # When not streaming, CompositeRawDecoder uses response.content which the requests
+            # library auto-decompresses when Content-Encoding is set. However, some APIs return
+            # gzip data without Content-Encoding headers. Using gzip_parser (which auto-detects
+            # gzip magic bytes) ensures decompression works in both cases.
+            return CompositeRawDecoder(gzip_parser, False)
 
         return CompositeRawDecoder.by_headers(
             [({"Content-Encoding", "Content-Type"}, _compressed_response_types, gzip_parser)],
             stream_response=True,
-            fallback_parser=gzip_parser.inner_parser,
+            fallback_parser=gzip_parser,
         )
 
     @staticmethod