chore(09-08-2026): update schemas

.github/workflows/e2e.yml

@@ -25,4 +25,4 @@ jobs:
       - name: Run E2E tests
         env:
           REDUCTO_API_KEY: ${{ secrets.REDUCTO_API_KEY }}
-        run: uv run pytest tests/e2e/ -v -x -o "addopts="
+        run: uv run pytest tests/e2e/ -v -x -o "addopts=" -n auto --dist loadscope

CONTRIBUTING.md

@@ -71,6 +71,12 @@ $ pip install ./path-to-wheel-file.whl
 $ ./scripts/test
 ```
 
+End-to-end tests in `tests/e2e/` call the live Reducto API and are excluded by default. They need `REDUCTO_API_KEY`. Run them in parallel, one worker per test class:
+
+```sh
+$ REDUCTO_API_KEY=... uv run pytest tests/e2e/ -o "addopts=" -n auto --dist loadscope
+```
+
 ## Linting and formatting
 
 This repository uses [ruff](https://github.com/astral-sh/ruff) to format the code.

api.md

@@ -11,12 +11,16 @@ from reducto.types import (
     AsyncPipelineResponse,
     AsyncSplitResponse,
     BaseProcessingOptions,
+    ChartResponse,
     Chunking,
     ChunkingConfig,
     ClassifyResponse,
     DirectWebhookConfig,
+    DocumentProperties,
     EditResponse,
     EnrichConfig,
+    ErrorCode,
+    ErrorDetail,
     ExperimentalProcessingOptions,
     ExtractResponse,
     FigureAgentic,
@@ -31,6 +35,7 @@ from reducto.types import (
     TableAgentic,
     TableSummaryConfig,
     TextAgentic,
+    URLResult,
     Upload,
     WebhookConfigNew,
 )
@@ -41,12 +46,13 @@ from reducto.types import (
 Types:
 
 ```python
-from reducto.types import APIVersionResponse
+from reducto.types import APIVersionResponse, ClientDeleteUploadResponse
 ```
 
 Methods:
 
 - <code title="get /version">client.<a href="./src/reducto/_client.py">api_version</a>() -> str</code>
+- <code title="delete /upload/{file_id}">client.<a href="./src/reducto/_client.py">delete_upload</a>(file_id) -> <a href="./src/reducto/types/client_delete_upload_response.py">ClientDeleteUploadResponse</a></code>
 - <code title="post /upload">client.<a href="./src/reducto/_client.py">upload</a>(\*\*<a href="src/reducto/types/client_upload_params.py">params</a>) -> <a href="./src/reducto/types/shared/upload.py">Upload</a></code>
 
 # Parse
@@ -154,11 +160,12 @@ Methods:
 Types:
 
 ```python
-from reducto.types import JobGetResponse, JobGetAllResponse
+from reducto.types import JobDeleteResponse, JobGetResponse, JobGetAllResponse
 ```
 
 Methods:
 
 - <code title="post /cancel/{job_id}">client.job.<a href="./src/reducto/resources/job.py">cancel</a>(job_id) -> object</code>
+- <code title="delete /job/{job_id}">client.job.<a href="./src/reducto/resources/job.py">delete</a>(job_id, \*\*<a href="src/reducto/types/job_delete_params.py">params</a>) -> <a href="./src/reducto/types/job_delete_response.py">JobDeleteResponse</a></code>
 - <code title="get /job/{job_id}">client.job.<a href="./src/reducto/resources/job.py">get</a>(job_id) -> <a href="./src/reducto/types/job_get_response.py">JobGetResponse</a></code>
 - <code title="get /jobs">client.job.<a href="./src/reducto/resources/job.py">get_all</a>(\*\*<a href="src/reducto/types/job_get_all_params.py">params</a>) -> <a href="./src/reducto/types/job_get_all_response.py">JobGetAllResponse</a></code>

scripts/spec_drift.py

@@ -7,6 +7,10 @@
 and walked in parallel, comparing JSON property names, types, enum values, and
 required-ness.
 
+`spec/drift-allowlist.json` lists known, intentional deviations with a reason.
+Matching items are reported as allowed and do not fail the check. Entries that
+match nothing are flagged as stale.
+
 The check runs against the committed snapshot `spec/openapi.json` by default, so
 it is reproducible and needs no network. `--live` checks against the public URL.
 `--update-snapshot` fetches the live spec, rewrites the snapshot, then checks.
@@ -31,7 +35,7 @@
 import typing_extensions
 from typing import Any, Dict, List, Tuple, Union, Literal, Optional
 from pathlib import Path
-from dataclasses import field, asdict, dataclass
+from dataclasses import field, asdict, replace, dataclass
 from typing_extensions import Required, Annotated, NotRequired, get_args, get_origin, is_typeddict
 
 REPO = Path(__file__).resolve().parent.parent
@@ -43,6 +47,7 @@
 
 LIVE_SPEC = "https://reducto.ai/openapi.json"
 SNAPSHOT = REPO / "spec" / "openapi.json"
+ALLOWLIST = REPO / "spec" / "drift-allowlist.json"
 HTTP_VERBS = ("get", "post", "put", "patch", "delete")
 HTTP_METHODS = {name: verb for verb in HTTP_VERBS for name in (verb, "_" + verb)}
 SUCCESS_CODES = ("200", "201", "202")
@@ -64,6 +69,7 @@ class Shape:
     items: Optional[Shape] = None
     members: List[Shape] = field(default_factory=lambda: [])
     label: str = ""
+    has_default: bool = False
 
 
 ANY = Shape("any")
@@ -117,6 +123,12 @@ def resolve(self, node: JsonDict) -> JsonDict:
         return node
 
     def shape(self, node: JsonDict, seen: typing.FrozenSet[str] = frozenset()) -> Shape:
+        shape = self.shape_of(node, seen)
+        if "default" in node and not shape.has_default:
+            shape = replace(shape, has_default=True)
+        return shape
+
+    def shape_of(self, node: JsonDict, seen: typing.FrozenSet[str]) -> Shape:
         label = ""
         if "$ref" in node:
             label = node["$ref"].rsplit("/", 1)[-1]
@@ -485,9 +497,13 @@ def compare_required(self, spec: Shape, sdk: Shape, name: str, loc: str) -> None
         if spec_req == sdk_req:
             return
         # Response models express nullable fields as Optional[...] = None, so
-        # "required but nullable" in the spec is not drift there.
+        # "required but nullable" in the spec is not drift there. A response
+        # field with a server default is always present, so the sdk may
+        # require it.
         if self.ignore_null and spec_req and is_nullable(spec.props[name]):
             return
+        if self.ignore_null and sdk_req and spec.props[name].has_default:
+            return
         self.report(loc, "required", f"spec {req_word(spec_req)}, sdk {req_word(sdk_req)}")
 
     def compare_union(self, spec: Shape, sdk: Shape, loc: str) -> None:
@@ -682,9 +698,59 @@ def display(source: str) -> str:
     return source
 
 
-def print_report(drifts: List[Drift], endpoints: List[Endpoint], spec: Spec, source: str) -> None:
+@dataclass
+class Allowed:
+    endpoint: str
+    location: str
+    kind: str
+    reason: str
+    detail: Optional[str] = None
+
+    def matches(self, d: Drift) -> bool:
+        if (self.endpoint, self.location, self.kind) != (d.endpoint, d.location, d.kind):
+            return False
+        return self.detail is None or self.detail == d.detail
+
+
+def load_allowlist(path: Path) -> List[Allowed]:
+    if not path.exists():
+        return []
+    entries = typing.cast(List[JsonDict], json.loads(path.read_text()))
+    return [Allowed(**e) for e in entries]
+
+
+def split_allowed(drifts: List[Drift], allowlist: List[Allowed]) -> Tuple[List[Drift], List[Drift], List[Allowed]]:
+    active: List[Drift] = []
+    allowed: List[Drift] = []
+    used: typing.Set[int] = set()
+    for d in drifts:
+        hit = next((i for i, a in enumerate(allowlist) if a.matches(d)), None)
+        if hit is None:
+            active.append(d)
+        else:
+            allowed.append(d)
+            used.add(hit)
+    stale = [a for i, a in enumerate(allowlist) if i not in used]
+    return active, allowed, stale
+
+
+def print_report(
+    drifts: List[Drift],
+    allowed: List[Drift],
+    stale: List[Allowed],
+    endpoints: List[Endpoint],
+    spec: Spec,
+    source: str,
+) -> None:
     print(f"Spec: {display(source)} (version {spec.version})")
     print(f"Checked {len(endpoints)} sdk endpoints.")
+    if allowed:
+        print(f"{len(allowed)} allowed drift item(s), see {ALLOWLIST.relative_to(REPO)}:")
+        for d in allowed:
+            loc = f" {d.location}" if d.location else ""
+            print(f"  {d.endpoint} [{d.kind}]{loc}: {d.detail}")
+    for a in stale:
+        print(f"Stale allowlist entry matches nothing: {a.endpoint} [{a.kind}] {a.location}", file=sys.stderr)
     if not drifts:
         print("No drift found.")
         return
@@ -735,13 +801,19 @@ def main() -> int:
 
     spec = Spec(load_doc(source))
     endpoints = discover_endpoints()
-    drifts = [d for d in run(spec, endpoints) if d.kind not in args.ignore]
+    found = [d for d in run(spec, endpoints) if d.kind not in args.ignore]
+    drifts, allowed, stale = split_allowed(found, load_allowlist(ALLOWLIST))
 
     if args.json:
-        payload = {"spec": source, "spec_version": spec.version, "drifts": [asdict(d) for d in drifts]}
+        payload = {
+            "spec": source,
+            "spec_version": spec.version,
+            "drifts": [asdict(d) for d in drifts],
+            "allowed": [asdict(d) for d in allowed],
+        }
         print(json.dumps(payload, indent=2))
     else:
-        print_report(drifts, endpoints, spec, source)
+        print_report(drifts, allowed, stale, endpoints, spec, source)
     return 0 if not drifts or args.warn_only else 1
 
 

spec/drift-allowlist.json

@@ -0,0 +1,9 @@
+[
+  {
+    "endpoint": "POST /extract",
+    "location": "response",
+    "kind": "extra",
+    "detail": "sdk union member object(7 fields) has no spec counterpart",
+    "reason": "The API returns ExtractResponse (response_type 'extract') for accounts not on V3 extract. The spec lists only V3ExtractResponse and AsyncExtractResponse for this endpoint. ExtractRunResponse includes ExtractResponse so those responses parse. Remove once the spec adds ExtractResponse to POST /extract."
+  }
+]

src/reducto/_client.py

@@ -26,6 +26,7 @@
 from ._utils import (
     is_given,
     extract_files,
+    path_template,
     maybe_transform,
     deepcopy_minimal,
     get_async_library,
@@ -49,6 +50,7 @@
     make_request_options,
 )
 from .types.shared.upload import Upload
+from .types.client_delete_upload_response import ClientDeleteUploadResponse
 
 if TYPE_CHECKING:
     from .resources import job, edit, parse, split, extract, webhook, classify, pipeline
@@ -312,6 +314,39 @@ def api_version(
             cast_to=str,
         )
 
+    def delete_upload(
+        self,
+        file_id: str,
+        *,
+        # Use the following arguments if you need to pass additional parameters to the API that aren't available via kwargs.
+        # The extra values given here take precedence over values defined on the client or passed to this method.
+        extra_headers: Headers | None = None,
+        extra_query: Query | None = None,
+        extra_body: Body | None = None,
+        timeout: float | httpx.Timeout | None | NotGiven = not_given,
+    ) -> ClientDeleteUploadResponse:
+        """
+        Delete Uploaded File
+
+        Args:
+          extra_headers: Send extra headers
+
+          extra_query: Add additional query parameters to the request
+
+          extra_body: Add additional JSON properties to the request
+
+          timeout: Override the client-level default timeout for this request, in seconds
+        """
+        if not file_id:
+            raise ValueError(f"Expected a non-empty value for `file_id` but received {file_id!r}")
+        return self.delete(
+            path_template("/upload/{file_id}", file_id=file_id),
+            options=make_request_options(
+                extra_headers=extra_headers, extra_query=extra_query, extra_body=extra_body, timeout=timeout
+            ),
+            cast_to=ClientDeleteUploadResponse,
+        )
+
     def upload(
         self,
         *,
@@ -623,6 +658,39 @@ async def api_version(
             cast_to=str,
         )
 
+    async def delete_upload(
+        self,
+        file_id: str,
+        *,
+        # Use the following arguments if you need to pass additional parameters to the API that aren't available via kwargs.
+        # The extra values given here take precedence over values defined on the client or passed to this method.
+        extra_headers: Headers | None = None,
+        extra_query: Query | None = None,
+        extra_body: Body | None = None,
+        timeout: float | httpx.Timeout | None | NotGiven = not_given,
+    ) -> ClientDeleteUploadResponse:
+        """
+        Delete Uploaded File
+
+        Args:
+          extra_headers: Send extra headers
+
+          extra_query: Add additional query parameters to the request
+
+          extra_body: Add additional JSON properties to the request
+
+          timeout: Override the client-level default timeout for this request, in seconds
+        """
+        if not file_id:
+            raise ValueError(f"Expected a non-empty value for `file_id` but received {file_id!r}")
+        return await self.delete(
+            path_template("/upload/{file_id}", file_id=file_id),
+            options=make_request_options(
+                extra_headers=extra_headers, extra_query=extra_query, extra_body=extra_body, timeout=timeout
+            ),
+            cast_to=ClientDeleteUploadResponse,
+        )
+
     async def upload(
         self,
         *,
@@ -711,6 +779,9 @@ def __init__(self, client: Reducto) -> None:
         self.api_version = to_raw_response_wrapper(
             client.api_version,
         )
+        self.delete_upload = to_raw_response_wrapper(
+            client.delete_upload,
+        )
         self.upload = to_raw_response_wrapper(
             client.upload,
         )
@@ -773,6 +844,9 @@ def __init__(self, client: AsyncReducto) -> None:
         self.api_version = async_to_raw_response_wrapper(
             client.api_version,
         )
+        self.delete_upload = async_to_raw_response_wrapper(
+            client.delete_upload,
+        )
         self.upload = async_to_raw_response_wrapper(
             client.upload,
         )
@@ -835,6 +909,9 @@ def __init__(self, client: Reducto) -> None:
         self.api_version = to_streamed_response_wrapper(
             client.api_version,
         )
+        self.delete_upload = to_streamed_response_wrapper(
+            client.delete_upload,
+        )
         self.upload = to_streamed_response_wrapper(
             client.upload,
         )
@@ -897,6 +974,9 @@ def __init__(self, client: AsyncReducto) -> None:
         self.api_version = async_to_streamed_response_wrapper(
             client.api_version,
         )
+        self.delete_upload = async_to_streamed_response_wrapper(
+            client.delete_upload,
+        )
         self.upload = async_to_streamed_response_wrapper(
             client.upload,
         )