release: 0.18.0

.release-please-manifest.json

@@ -1,3 +1,3 @@
 {
-  ".": "0.17.1"
+  ".": "0.18.0"
 }

.stats.yml

@@ -1,4 +1,4 @@
 configured_endpoints: 5
-openapi_spec_url: https://storage.googleapis.com/stainless-sdk-openapi-specs/isaacus%2Fisaacus-fd2419c8cf8666e821f55c4cc5a5402900c443e2db59b3e53cfb46fd5bb52316.yml
-openapi_spec_hash: a92a4436234b019c9c23feb11b1f2f28
+openapi_spec_url: https://storage.googleapis.com/stainless-sdk-openapi-specs/isaacus%2Fisaacus-46de7b353c33e2b6c139e564caeec9e0462ad714121690f65167a7943e325000.yml
+openapi_spec_hash: 6527287f9709a8741c9cc5b4181d7bb1
 config_hash: 9040e7359f066240ad536041fb2c5185

CHANGELOG.md

@@ -1,5 +1,23 @@
 # Changelog
 
+## 0.18.0 (2026-02-11)
+
+Full Changelog: [v0.17.1...v0.18.0](https://github.com/isaacus-dev/isaacus-python/compare/v0.17.1...v0.18.0)
+
+### Features
+
+* **api:** add chunking mode to enrichment ([20a3231](https://github.com/isaacus-dev/isaacus-python/commit/20a3231576f04f67e35db65de497ea89fccafdb0))
+
+
+### Chores
+
+* **internal:** bump dependencies ([233ae5c](https://github.com/isaacus-dev/isaacus-python/commit/233ae5cb808d97971cfffb68d2b5191c2a4d7cfa))
+
+
+### Documentation
+
+* **api:** update `creation` date type definition ([66db3ea](https://github.com/isaacus-dev/isaacus-python/commit/66db3eac21d416b05e0da12e4f22871a03ef3ebc))
+
 ## 0.17.1 (2026-02-07)
 
 Full Changelog: [v0.17.0...v0.17.1](https://github.com/isaacus-dev/isaacus-python/compare/v0.17.0...v0.17.1)

pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "isaacus"
-version = "0.17.1"
+version = "0.18.0"
 description = "The official Python library for the isaacus API"
 dynamic = ["readme"]
 license = "Apache-2.0"

requirements-dev.lock

@@ -12,14 +12,14 @@
 -e file:.
 aiohappyeyeballs==2.6.1
     # via aiohttp
-aiohttp==3.13.2
+aiohttp==3.13.3
     # via httpx-aiohttp
     # via isaacus
 aiosignal==1.4.0
     # via aiohttp
 annotated-types==0.7.0
     # via pydantic
-anyio==4.12.0
+anyio==4.12.1
     # via httpx
     # via isaacus
 argcomplete==3.6.3
@@ -31,7 +31,7 @@ attrs==25.4.0
     # via nox
 backports-asyncio-runner==1.2.0
     # via pytest-asyncio
-certifi==2025.11.12
+certifi==2026.1.4
     # via httpcore
     # via httpx
 colorlog==6.10.1
@@ -61,15 +61,15 @@ httpx==0.28.1
     # via httpx-aiohttp
     # via isaacus
     # via respx
-httpx-aiohttp==0.1.9
+httpx-aiohttp==0.1.12
     # via isaacus
 humanize==4.13.0
     # via nox
 idna==3.11
     # via anyio
     # via httpx
     # via yarl
-importlib-metadata==8.7.0
+importlib-metadata==8.7.1
 iniconfig==2.1.0
     # via pytest
 markdown-it-py==3.0.0
@@ -82,14 +82,14 @@ multidict==6.7.0
 mypy==1.17.0
 mypy-extensions==1.1.0
     # via mypy
-nodeenv==1.9.1
+nodeenv==1.10.0
     # via pyright
 nox==2025.11.12
 packaging==25.0
     # via dependency-groups
     # via nox
     # via pytest
-pathspec==0.12.1
+pathspec==1.0.3
     # via mypy
 platformdirs==4.4.0
     # via virtualenv
@@ -115,13 +115,13 @@ python-dateutil==2.9.0.post0
     # via time-machine
 respx==0.22.0
 rich==14.2.0
-ruff==0.14.7
+ruff==0.14.13
 six==1.17.0
     # via python-dateutil
 sniffio==1.3.1
     # via isaacus
 time-machine==2.19.0
-tomli==2.3.0
+tomli==2.4.0
     # via dependency-groups
     # via mypy
     # via nox
@@ -141,7 +141,7 @@ typing-extensions==4.15.0
     # via virtualenv
 typing-inspection==0.4.2
     # via pydantic
-virtualenv==20.35.4
+virtualenv==20.36.1
     # via nox
 yarl==1.22.0
     # via aiohttp

requirements.lock

@@ -12,21 +12,21 @@
 -e file:.
 aiohappyeyeballs==2.6.1
     # via aiohttp
-aiohttp==3.13.2
+aiohttp==3.13.3
     # via httpx-aiohttp
     # via isaacus
 aiosignal==1.4.0
     # via aiohttp
 annotated-types==0.7.0
     # via pydantic
-anyio==4.12.0
+anyio==4.12.1
     # via httpx
     # via isaacus
 async-timeout==5.0.1
     # via aiohttp
 attrs==25.4.0
     # via aiohttp
-certifi==2025.11.12
+certifi==2026.1.4
     # via httpcore
     # via httpx
 distro==1.9.0
@@ -43,7 +43,7 @@ httpcore==1.0.9
 httpx==0.28.1
     # via httpx-aiohttp
     # via isaacus
-httpx-aiohttp==0.1.9
+httpx-aiohttp==0.1.12
     # via isaacus
 idna==3.11
     # via anyio

src/isaacus/_version.py

@@ -1,4 +1,4 @@
 # File generated from our OpenAPI spec by Stainless. See CONTRIBUTING.md for details.
 
 __title__ = "isaacus"
-__version__ = "0.17.1"  # x-release-please-version
+__version__ = "0.18.0"  # x-release-please-version

src/isaacus/resources/enrichments.py

@@ -49,7 +49,7 @@ def create(
         *,
         model: Literal["kanon-2-enricher"],
         texts: Union[SequenceNotStr[str], str],
-        overflow_strategy: Optional[Literal["auto", "drop_end"]] | Omit = omit,
+        overflow_strategy: Optional[Literal["auto", "drop_end", "chunk"]] | Omit = omit,
         # 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,
@@ -71,9 +71,15 @@ def create(
 
           overflow_strategy: The strategy for handling content exceeding the model's maximum input length.
 
-              `auto` currently behaves the same as `drop_end`, dropping excess tokens from the
-              end of input. In the future, `auto` may implement more sophisticated strategies
-              such as chunking and context-aware stitching.
+              `auto`, which is the default and recommended setting, currently behaves the same
+              as `chunk`, which intelligently breaks the input up into smaller chunks and then
+              stitches the results back together into a single prediction. In the future
+              `auto` may implement even more sophisticated strategies for handling long
+              contexts such as leveraging chunk overlap and/or a specialized stitching model.
+
+              `chunk` breaks the input up into smaller chunks that fit within the model's
+              context window and then intelligently merges the results into a single
+              prediction at the cost of a minor accuracy drop.
 
               `drop_end` drops tokens from the end of input exceeding the model's maximum
               input length.
@@ -131,7 +137,7 @@ async def create(
         *,
         model: Literal["kanon-2-enricher"],
         texts: Union[SequenceNotStr[str], str],
-        overflow_strategy: Optional[Literal["auto", "drop_end"]] | Omit = omit,
+        overflow_strategy: Optional[Literal["auto", "drop_end", "chunk"]] | Omit = omit,
         # 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,
@@ -153,9 +159,15 @@ async def create(
 
           overflow_strategy: The strategy for handling content exceeding the model's maximum input length.
 
-              `auto` currently behaves the same as `drop_end`, dropping excess tokens from the
-              end of input. In the future, `auto` may implement more sophisticated strategies
-              such as chunking and context-aware stitching.
+              `auto`, which is the default and recommended setting, currently behaves the same
+              as `chunk`, which intelligently breaks the input up into smaller chunks and then
+              stitches the results back together into a single prediction. In the future
+              `auto` may implement even more sophisticated strategies for handling long
+              contexts such as leveraging chunk overlap and/or a specialized stitching model.
+
+              `chunk` breaks the input up into smaller chunks that fit within the model's
+              context window and then intelligently merges the results into a single
+              prediction at the cost of a minor accuracy drop.
 
               `drop_end` drops tokens from the end of input exceeding the model's maximum
               input length.

src/isaacus/types/enrichment_create_params.py

@@ -25,12 +25,18 @@ class EnrichmentCreateParams(TypedDict, total=False):
     No more than 8 texts can be enriched in a single request.
     """
 
-    overflow_strategy: Optional[Literal["auto", "drop_end"]]
+    overflow_strategy: Optional[Literal["auto", "drop_end", "chunk"]]
     """The strategy for handling content exceeding the model's maximum input length.
 
-    `auto` currently behaves the same as `drop_end`, dropping excess tokens from the
-    end of input. In the future, `auto` may implement more sophisticated strategies
-    such as chunking and context-aware stitching.
+    `auto`, which is the default and recommended setting, currently behaves the same
+    as `chunk`, which intelligently breaks the input up into smaller chunks and then
+    stitches the results back together into a single prediction. In the future
+    `auto` may implement even more sophisticated strategies for handling long
+    contexts such as leveraging chunk overlap and/or a specialized stitching model.
+
+    `chunk` breaks the input up into smaller chunks that fit within the model's
+    context window and then intelligently merges the results into a single
+    prediction at the cost of a minor accuracy drop.
 
     `drop_end` drops tokens from the end of input exceeding the model's maximum
     input length.

src/isaacus/types/ilgs/v1/date.py

@@ -26,8 +26,8 @@ class Date(BaseModel):
     mentioned in a document that does not fit into a supported type, it will not be
     extracted.
 
-    `creation` denotes the date the document was created. There may only be one
-    `creation` date per document.
+    `creation` denotes the date the document was created or last updated. There may
+    only be one `creation` date per document.
 
     `signature` denotes the date the document was signed.