feat: add extraction formats parameter to sdk

Signed-off-by: Aleš Kalfas <kalfas.ales@gmail.com>

Author: aleskalfas

apps/agentstack-sdk-py/src/agentstack_sdk/platform/file.py

@@ -16,12 +16,14 @@
 from agentstack_sdk.util.file import LoadedFile, LoadedFileWithUri, PlatformFileUrl
 from agentstack_sdk.util.utils import filter_dict
 
+ExtractionFormatLiteral = typing.Literal["markdown", "vendor_specific_json"]
+
 
 class ExtractedFileInfo(pydantic.BaseModel):
     """Information about an extracted file."""
 
     file_id: str
-    format: typing.Literal["markdown", "vendor_specific_json"] | None
+    format: ExtractionFormatLiteral | None
 
 
 class Extraction(pydantic.BaseModel):
@@ -195,6 +197,7 @@ async def load_json_content(
     async def create_extraction(
         self: File | str,
         *,
+        formats: list[ExtractionFormatLiteral] | None = None,
         client: PlatformClient | None = None,
         context_id: str | None | Literal["auto"] = "auto",
     ) -> Extraction:
@@ -207,6 +210,7 @@ async def create_extraction(
                     await platform_client.post(
                         url=f"/api/v1/files/{file_id}/extraction",
                         params=context_id and {"context_id": context_id},
+                        json={"settings": {"formats": formats}} if formats else None,
                     )
                 )
                 .raise_for_status()

docs/agent-development/rag.mdx

@@ -99,7 +99,7 @@ async def extract_file(file: File):
 
 #### Extraction Formats
 
-Text extraction produces two extraction formats:
+Text extraction produces two extraction formats and you can request either subset by passing `formats` to `create_extraction` (e.g., `["markdown"]` if you only need plain text):
 
 - __markdown__: The extracted text formatted as Markdown (`file.load_text_content()`)
 - __vendor_specific_json__: The Docling-specific JSON format containing document structure (`file.load_json_content()`)