chunk-invariance 0.1.0__tar.gz

chunk_invariance-0.1.0/LICENSE

@@ -0,0 +1,190 @@
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+Copyright 2026 Ninad Phalak
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.

chunk_invariance-0.1.0/PKG-INFO

@@ -0,0 +1,111 @@
+Metadata-Version: 2.4
+Name: chunk-invariance
+Version: 0.1.0
+Summary: Assert that a streaming filter gives the same output however its input is split into chunks.
+Author-email: Ninad Phalak <ninadphalak@gmail.com>
+Maintainer-email: Ninad Phalak <ninadphalak@gmail.com>
+License-Expression: Apache-2.0
+Project-URL: Homepage, https://github.com/ninadphalak/LLM-Shield-Proxy/tree/main/chunk-invariance
+Project-URL: Repository, https://github.com/ninadphalak/LLM-Shield-Proxy
+Project-URL: Bug Tracker, https://github.com/ninadphalak/LLM-Shield-Proxy/issues
+Keywords: streaming,testing,redaction,guardrails,chunk-boundary,property-based-testing
+Classifier: Development Status :: 4 - Beta
+Classifier: Intended Audience :: Developers
+Classifier: Framework :: Hypothesis
+Classifier: Framework :: Pytest
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.9
+Classifier: Programming Language :: Python :: 3.10
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3.13
+Classifier: Topic :: Software Development :: Testing
+Classifier: Typing :: Typed
+Requires-Python: >=3.9
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Provides-Extra: hypothesis
+Requires-Dist: hypothesis>=6; extra == "hypothesis"
+Dynamic: license-file
+
+# chunk-invariance
+
+A streaming filter (a redactor, a guardrail, a rewriter) is correct only if, for every way of
+splitting an input into chunks, the concatenated streamed output equals the output of filtering
+the whole input at once. This package turns that rule into one assertion.
+
+```python
+from chunk_invariance import assert_chunk_invariant
+
+def redact(text):                 # your filter on the whole input
+    return text.replace("sk-AbCd1234", "[KEY]")
+
+def redact_stream(chunks):        # your filter on a stream: takes chunks, yields chunks
+    for chunk in chunks:
+        yield redact(chunk)       # redacts each chunk alone
+
+def test_streaming_redaction():
+    assert_chunk_invariant(redact_stream, redact, "use sk-AbCd1234 now")
+```
+
+That test fails, and names the smallest split that breaks it:
+
+```
+chunk_invariance.core.ChunkInvarianceError: stream filter is not chunk-invariant
+  input:    'use sk-AbCd1234 now'
+  split:    ['use s', 'k-AbCd1234 now'] (cut at 5)
+  whole:    'use [KEY] now'
+  streamed: 'use sk-AbCd1234 now'
+  split #6 in the order tried; generated splits run fewest parts first
+```
+
+```bash
+pip install chunk-invariance
+pip install "chunk-invariance[hypothesis]"   # optional Hypothesis strategies
+```
+
+## What it catches
+
+- A filter that evaluates each chunk alone, so a value cut across two chunks matches neither half.
+- A filter that prepends a fixed carryover from the previous chunk. The match fires, but the start
+  of the value was already emitted with the previous chunk.
+- A hold-back that is shorter than the pattern, a missing end-of-stream flush, and output that is
+  corrupted rather than leaked (any difference from the whole-input output fails).
+
+`chunk_invariance.examples` has one correct filter (`hold_back`) and the two wrong designs
+(`design_a`, `design_b`) for the same pattern, as a worked comparison.
+
+## What it does not catch
+
+It tests a filter you can call in-process. It does not test a running gateway, where the chunks
+are decided by an upstream model, a network and an SSE parser. For that, use
+[`pii-leak-benchmark`](https://pypi.org/project/pii-leak-benchmark/), which sends
+fragmented responses through the gateway and checks what reaches the client.
+
+It checks the splits you ask for. The default is the whole input plus every single cut, which
+catches a value cut in two. `splits="all"` tries every split of a short input (up to 16
+characters); `splits=3` tries every split into at most three parts; an explicit list of chunk
+lists runs exactly those. It cannot prove a filter correct for inputs you did not give it.
+
+## API
+
+| Name | Does |
+|---|---|
+| `assert_chunk_invariant(stream_filter, whole_filter, text, *, splits="all-two-part")` | Raises `ChunkInvarianceError` (an `AssertionError`) on the first failing split, fewest parts first. Returns the number of splits checked. Works on `str` and `bytes`. |
+| `assert_chunk_invariant_async(...)` | The same, for an async generator filter, inside an async test. |
+| `two_part_splits(text)` | Every split into two non-empty parts. |
+| `all_splits(text, max_parts=None)` | Every split into at most `max_parts` non-empty parts, the whole input first. |
+| `per_chunk(process, *, state=dict, flush=None)` | Adapts `process(chunk, state) -> output` and `flush(state) -> output`, with fresh state per stream. |
+| `from_async(async_stream_filter)` | Adapts an async generator filter for the synchronous assertion. |
+| `strategies.splits_of(text, max_parts=None)` | Hypothesis strategy: random splits of `text`, shrinking toward fewer cuts. |
+| `strategies.text_and_splits(texts, max_parts=None)` | Hypothesis strategy: `(text, chunks)` pairs. |
+
+`stream_filter` is called once per split, so it must start from fresh state each time.
+
+## Background
+
+The invariant and the two wrong designs are described in Phalak, N. (2026), *Split-Boundary
+Leaks in Streaming Guardrails*, Zenodo, [10.5281/zenodo.22909585](https://doi.org/10.5281/zenodo.22909585).
+
+Standard library only. Apache-2.0. A TypeScript version with the same API is published on npm
+as `chunk-invariance`.

chunk_invariance-0.1.0/README.md

@@ -0,0 +1,81 @@
+# chunk-invariance
+
+A streaming filter (a redactor, a guardrail, a rewriter) is correct only if, for every way of
+splitting an input into chunks, the concatenated streamed output equals the output of filtering
+the whole input at once. This package turns that rule into one assertion.
+
+```python
+from chunk_invariance import assert_chunk_invariant
+
+def redact(text):                 # your filter on the whole input
+    return text.replace("sk-AbCd1234", "[KEY]")
+
+def redact_stream(chunks):        # your filter on a stream: takes chunks, yields chunks
+    for chunk in chunks:
+        yield redact(chunk)       # redacts each chunk alone
+
+def test_streaming_redaction():
+    assert_chunk_invariant(redact_stream, redact, "use sk-AbCd1234 now")
+```
+
+That test fails, and names the smallest split that breaks it:
+
+```
+chunk_invariance.core.ChunkInvarianceError: stream filter is not chunk-invariant
+  input:    'use sk-AbCd1234 now'
+  split:    ['use s', 'k-AbCd1234 now'] (cut at 5)
+  whole:    'use [KEY] now'
+  streamed: 'use sk-AbCd1234 now'
+  split #6 in the order tried; generated splits run fewest parts first
+```
+
+```bash
+pip install chunk-invariance
+pip install "chunk-invariance[hypothesis]"   # optional Hypothesis strategies
+```
+
+## What it catches
+
+- A filter that evaluates each chunk alone, so a value cut across two chunks matches neither half.
+- A filter that prepends a fixed carryover from the previous chunk. The match fires, but the start
+  of the value was already emitted with the previous chunk.
+- A hold-back that is shorter than the pattern, a missing end-of-stream flush, and output that is
+  corrupted rather than leaked (any difference from the whole-input output fails).
+
+`chunk_invariance.examples` has one correct filter (`hold_back`) and the two wrong designs
+(`design_a`, `design_b`) for the same pattern, as a worked comparison.
+
+## What it does not catch
+
+It tests a filter you can call in-process. It does not test a running gateway, where the chunks
+are decided by an upstream model, a network and an SSE parser. For that, use
+[`pii-leak-benchmark`](https://pypi.org/project/pii-leak-benchmark/), which sends
+fragmented responses through the gateway and checks what reaches the client.
+
+It checks the splits you ask for. The default is the whole input plus every single cut, which
+catches a value cut in two. `splits="all"` tries every split of a short input (up to 16
+characters); `splits=3` tries every split into at most three parts; an explicit list of chunk
+lists runs exactly those. It cannot prove a filter correct for inputs you did not give it.
+
+## API
+
+| Name | Does |
+|---|---|
+| `assert_chunk_invariant(stream_filter, whole_filter, text, *, splits="all-two-part")` | Raises `ChunkInvarianceError` (an `AssertionError`) on the first failing split, fewest parts first. Returns the number of splits checked. Works on `str` and `bytes`. |
+| `assert_chunk_invariant_async(...)` | The same, for an async generator filter, inside an async test. |
+| `two_part_splits(text)` | Every split into two non-empty parts. |
+| `all_splits(text, max_parts=None)` | Every split into at most `max_parts` non-empty parts, the whole input first. |
+| `per_chunk(process, *, state=dict, flush=None)` | Adapts `process(chunk, state) -> output` and `flush(state) -> output`, with fresh state per stream. |
+| `from_async(async_stream_filter)` | Adapts an async generator filter for the synchronous assertion. |
+| `strategies.splits_of(text, max_parts=None)` | Hypothesis strategy: random splits of `text`, shrinking toward fewer cuts. |
+| `strategies.text_and_splits(texts, max_parts=None)` | Hypothesis strategy: `(text, chunks)` pairs. |
+
+`stream_filter` is called once per split, so it must start from fresh state each time.
+
+## Background
+
+The invariant and the two wrong designs are described in Phalak, N. (2026), *Split-Boundary
+Leaks in Streaming Guardrails*, Zenodo, [10.5281/zenodo.22909585](https://doi.org/10.5281/zenodo.22909585).
+
+Standard library only. Apache-2.0. A TypeScript version with the same API is published on npm
+as `chunk-invariance`.

chunk_invariance-0.1.0/chunk_invariance/__init__.py

@@ -0,0 +1,26 @@
+"""Test that a streaming filter gives the same output however its input is split into chunks.
+
+Standard library only. The optional Hypothesis strategies live in ``chunk_invariance.strategies``
+and need the ``hypothesis`` extra.
+"""
+
+from .adapters import from_async, per_chunk
+from .core import (
+    ChunkInvarianceError,
+    all_splits,
+    assert_chunk_invariant,
+    assert_chunk_invariant_async,
+    two_part_splits,
+)
+
+__version__ = "0.1.0"
+
+__all__ = [
+    "ChunkInvarianceError",
+    "all_splits",
+    "assert_chunk_invariant",
+    "assert_chunk_invariant_async",
+    "from_async",
+    "per_chunk",
+    "two_part_splits",
+]

chunk_invariance-0.1.0/chunk_invariance/adapters.py

@@ -0,0 +1,54 @@
+"""Turn common streaming-filter shapes into the ``stream_filter`` the assertion calls."""
+
+from __future__ import annotations
+
+import asyncio
+from typing import Any, AsyncIterable, AsyncIterator, Callable, Iterable, Iterator, Optional
+
+
+def per_chunk(
+    process: Callable[[Any, Any], Any],
+    *,
+    state: Callable[[], Any] = dict,
+    flush: Optional[Callable[[Any], Any]] = None,
+) -> Callable[[Iterable[Any]], Iterator[Any]]:
+    """Adapt ``process(chunk, state) -> output`` plus an optional ``flush(state) -> output``.
+
+    ``state`` is a factory called once per stream, so every split starts clean. ``flush`` is
+    called after the last chunk; leave it out only if the filter really has no end-of-stream step.
+    """
+
+    def stream_filter(chunks: Iterable[Any]) -> Iterator[Any]:
+        current = state()
+        for chunk in chunks:
+            yield process(chunk, current)
+        if flush is not None:
+            yield flush(current)
+
+    return stream_filter
+
+
+def from_async(
+    async_stream_filter: Callable[[AsyncIterable[Any]], AsyncIterable[Any]],
+) -> Callable[[Iterable[Any]], Iterator[Any]]:
+    """Adapt an async generator function that takes an async iterable of chunks.
+
+    Each stream runs to completion in its own event loop. Inside a running loop (an async test),
+    use ``assert_chunk_invariant_async`` instead.
+    """
+
+    async def feed(chunks: Iterable[Any]) -> AsyncIterator[Any]:
+        for chunk in chunks:
+            yield chunk
+
+    async def collect(chunks: Iterable[Any]) -> list:
+        return [piece async for piece in async_stream_filter(feed(chunks))]
+
+    def stream_filter(chunks: Iterable[Any]) -> Iterator[Any]:
+        try:
+            asyncio.get_running_loop()
+        except RuntimeError:
+            return iter(asyncio.run(collect(list(chunks))))
+        raise RuntimeError("from_async cannot run inside an event loop; use assert_chunk_invariant_async")
+
+    return stream_filter

chunk_invariance-0.1.0/chunk_invariance/core.py

@@ -0,0 +1,156 @@
+"""The invariant, the splits it is checked over, and the assertion."""
+
+from __future__ import annotations
+
+import inspect
+from itertools import combinations
+from typing import Any, AsyncIterable, AsyncIterator, Callable, Iterable, Iterator, List, Optional, Sequence, Union
+
+Text = Union[str, bytes]
+StreamFilter = Callable[[Iterable[Any]], Iterable[Any]]
+WholeFilter = Callable[[Any], Any]
+Splits = Union[str, int, Iterable[Sequence[Any]]]
+
+# `splits="all"` enumerates 2 ** (len(text) - 1) splits; past this length that is too many
+# to run in a test. Pass an int (the most parts per split) for longer inputs.
+ALL_SPLITS_MAX_LENGTH = 16
+
+
+class ChunkInvarianceError(AssertionError):
+    """The streamed output for one split differs from filtering the whole input at once."""
+
+    def __init__(self, text: Text, chunks: Sequence[Text], expected: Text, actual: Text, splits_tried: int) -> None:
+        self.text = text
+        self.chunks = list(chunks)
+        self.expected = expected
+        self.actual = actual
+        self.splits_tried = splits_tried
+        cuts = _cut_points(self.chunks)
+        super().__init__(
+            "stream filter is not chunk-invariant\n"
+            f"  input:    {text!r}\n"
+            f"  split:    {self.chunks!r} (cut at {', '.join(map(str, cuts)) or 'nothing: one chunk'})\n"
+            f"  whole:    {expected!r}\n"
+            f"  streamed: {actual!r}\n"
+            f"  split #{splits_tried} in the order tried; generated splits run fewest parts first"
+        )
+
+
+def _cut_points(chunks: Sequence[Text]) -> List[int]:
+    cuts, offset = [], 0
+    for chunk in chunks[:-1]:
+        offset += len(chunk)
+        cuts.append(offset)
+    return cuts
+
+
+def _split_at(text: Text, cuts: Sequence[int]) -> List[Text]:
+    bounds = [0, *cuts, len(text)]
+    return [text[a:b] for a, b in zip(bounds, bounds[1:])]
+
+
+def two_part_splits(text: Text) -> Iterator[List[Text]]:
+    """Every split of ``text`` into two non-empty parts, left cut first."""
+    for cut in range(1, len(text)):
+        yield [text[:cut], text[cut:]]
+
+
+def all_splits(text: Text, max_parts: Optional[int] = None) -> Iterator[List[Text]]:
+    """Every split of ``text`` into at most ``max_parts`` non-empty parts (all of them when None).
+
+    Ordered by number of parts, then by cut positions, so the first failure found is a minimal one.
+    The single-part split ``[text]`` comes first.
+    """
+    if max_parts is not None and max_parts < 1:
+        raise ValueError("max_parts must be at least 1")
+    positions = range(1, len(text))
+    most = len(text) if max_parts is None else min(max_parts, max(len(text), 1))
+    for parts in range(1, most + 1):
+        for cuts in combinations(positions, parts - 1):
+            yield _split_at(text, cuts)
+
+
+def _resolve_splits(text: Text, splits: Splits) -> Iterable[Sequence[Text]]:
+    if splits == "all-two-part":
+        return all_splits(text, 2)
+    if splits == "all":
+        if len(text) > ALL_SPLITS_MAX_LENGTH:
+            raise ValueError(
+                f'splits="all" is 2 ** {len(text) - 1} splits for a {len(text)}-character input; '
+                f"use it up to {ALL_SPLITS_MAX_LENGTH} characters, or pass an int (the most parts per split)"
+            )
+        return all_splits(text)
+    if isinstance(splits, bool):
+        raise TypeError("splits must be 'all-two-part', 'all', an int, or an iterable of chunk lists")
+    if isinstance(splits, int):
+        return all_splits(text, splits)
+    if isinstance(splits, (str, bytes)):
+        raise ValueError(f"unknown splits value {splits!r}; use 'all-two-part', 'all', an int, or chunk lists")
+    return splits
+
+
+def _join(text: Text, pieces: Iterable[Text]) -> Text:
+    return text[:0].join(pieces)
+
+
+def assert_chunk_invariant(
+    stream_filter: StreamFilter,
+    whole_filter: WholeFilter,
+    text: Text,
+    *,
+    splits: Splits = "all-two-part",
+) -> int:
+    """Assert that streaming ``text`` in any of ``splits`` gives what filtering it whole gives.
+
+    ``stream_filter`` takes an iterable of input chunks and returns or yields output chunks; it is
+    called once per split, so it must start from fresh state each call. ``whole_filter`` takes the
+    whole input. ``splits`` is ``"all-two-part"`` (the default: the input whole, then every single
+    cut), ``"all"`` (every split, for short inputs), an int (every split into at most that many
+    parts), or an iterable of explicit chunk lists, which must each join back to ``text``.
+
+    Raises ``ChunkInvarianceError`` on the first split, fewest parts first, whose concatenated
+    output differs. Returns the number of splits checked.
+    """
+    expected = whole_filter(text)
+    tried = 0
+    for chunks in _resolve_splits(text, splits):
+        chunks = list(chunks)
+        if _join(text, chunks) != text:
+            raise ValueError(f"explicit split {chunks!r} does not join back to the input")
+        tried += 1
+        actual = _join(text, stream_filter(iter(chunks)))
+        if actual != expected:
+            raise ChunkInvarianceError(text, chunks, expected, actual, tried)
+    return tried
+
+
+async def assert_chunk_invariant_async(
+    stream_filter: Callable[[AsyncIterable[Any]], AsyncIterable[Any]],
+    whole_filter: Callable[[Any], Any],
+    text: Text,
+    *,
+    splits: Splits = "all-two-part",
+) -> int:
+    """``assert_chunk_invariant`` for an async generator filter, for use inside an async test.
+
+    ``stream_filter`` takes an async iterable of chunks and yields output chunks. ``whole_filter``
+    may return the output or an awaitable of it.
+    """
+
+    async def feed(chunks: Sequence[Text]) -> AsyncIterator[Text]:
+        for chunk in chunks:
+            yield chunk
+
+    expected = whole_filter(text)
+    if inspect.isawaitable(expected):
+        expected = await expected
+    tried = 0
+    for chunks in _resolve_splits(text, splits):
+        chunks = list(chunks)
+        if _join(text, chunks) != text:
+            raise ValueError(f"explicit split {chunks!r} does not join back to the input")
+        tried += 1
+        actual = _join(text, [piece async for piece in stream_filter(feed(chunks))])
+        if actual != expected:
+            raise ChunkInvarianceError(text, chunks, expected, actual, tried)
+    return tried

chunk_invariance-0.1.0/chunk_invariance/examples.py

@@ -0,0 +1,75 @@
+"""Three streaming redactors for one pattern: one correct, two that leak across a chunk boundary.
+
+The pattern is a fixed-length key, ``sk-`` plus eight letters or digits, replaced by ``[KEY]``.
+``redact`` is the whole-input filter all three are compared against.
+
+- ``hold_back`` is correct. It holds back the last ``KEY_LENGTH - 1`` characters, the longest
+  tail that could still be the start of a key, and redacts before it emits.
+- ``design_a`` redacts each chunk on its own. A key cut in two is never seen whole.
+- ``design_b`` prepends a fixed carryover from the previous chunk and redacts the join. The match
+  fires, but the start of the key already left with the previous chunk.
+
+Run ``assert_chunk_invariant(design_a, redact, "use sk-AbCd1234 now")`` to see the failure.
+"""
+
+from __future__ import annotations
+
+import re
+from typing import Dict, Iterable, Iterator
+
+from .adapters import per_chunk
+
+KEY = re.compile(r"sk-[A-Za-z0-9]{8}")
+KEY_LENGTH = 11
+MASK = "[KEY]"
+
+
+def redact(text: str) -> str:
+    """The whole-input filter: every key replaced."""
+    return KEY.sub(MASK, text)
+
+
+def _hold_back_process(chunk: str, state: Dict[str, str]) -> str:
+    buffer = state.get("buffer", "") + chunk
+    # A key that starts before `safe` ends inside the buffer, so it is already complete.
+    safe = len(buffer) - (KEY_LENGTH - 1)
+    out, emitted = [], 0
+    for match in KEY.finditer(buffer):
+        if match.start() >= safe:
+            break
+        out.append(buffer[emitted : match.start()])
+        out.append(MASK)
+        emitted = match.end()
+    keep_from = max(emitted, safe)
+    out.append(buffer[emitted:keep_from])
+    state["buffer"] = buffer[keep_from:]
+    return "".join(out)
+
+
+def _hold_back_flush(state: Dict[str, str]) -> str:
+    rest, state["buffer"] = state.get("buffer", ""), ""
+    return redact(rest)
+
+
+hold_back = per_chunk(_hold_back_process, state=dict, flush=_hold_back_flush)
+hold_back.__doc__ = "Correct: holds back a possible key prefix and redacts before emitting."
+
+
+def design_a(chunks: Iterable[str]) -> Iterator[str]:
+    """Leaks: each chunk is redacted alone, so a key split across two chunks passes."""
+    for chunk in chunks:
+        yield redact(chunk)
+
+
+def design_b(chunks: Iterable[str], carry: int = KEY_LENGTH - 1) -> Iterator[str]:
+    """Leaks: prepends the last ``carry`` characters of the previous chunk and redacts the join.
+
+    The match now fires, but the carried characters were already emitted with the previous
+    chunk, so the start of a split key has gone out unredacted, and cutting the carry off the
+    redacted join corrupts what follows.
+    """
+    previous = ""
+    for chunk in chunks:
+        joined = previous + chunk
+        yield redact(joined)[len(previous) :]
+        previous = joined[-carry:] if carry else ""

chunk_invariance-0.1.0/chunk_invariance/strategies.py

@@ -0,0 +1,40 @@
+"""Hypothesis strategies for splits. Needs ``pip install chunk-invariance[hypothesis]``.
+
+    from hypothesis import given
+    from chunk_invariance import assert_chunk_invariant
+    from chunk_invariance.strategies import splits_of
+
+    @given(splits_of("use sk-AbCd1234 now"))
+    def test_filter(chunks):
+        assert_chunk_invariant(my_filter, redact, "".join(chunks), splits=[chunks])
+"""
+
+from __future__ import annotations
+
+from typing import Any, List, Optional
+
+try:
+    from hypothesis import strategies as st
+except ImportError as exc:  # pragma: no cover - exercised only without the extra
+    raise ImportError("chunk_invariance.strategies needs Hypothesis: pip install 'chunk-invariance[hypothesis]'") from exc
+
+
+def splits_of(text: Any, max_parts: Optional[int] = None) -> "st.SearchStrategy[List[Any]]":
+    """Random splits of one fixed input into non-empty parts; shrinks toward fewer cuts."""
+    positions = list(range(1, len(text)))
+    most = len(positions) if max_parts is None else max(max_parts - 1, 0)
+
+    def build(cuts: List[int]) -> List[Any]:
+        bounds = [0, *sorted(cuts), len(text)]
+        return [text[a:b] for a, b in zip(bounds, bounds[1:])]
+
+    if not positions or most == 0:
+        return st.just([text])
+    return st.lists(st.sampled_from(positions), unique=True, max_size=most).map(build)
+
+
+@st.composite
+def text_and_splits(draw: Any, texts: "st.SearchStrategy[Any]", max_parts: Optional[int] = None) -> Any:
+    """Draw an input from ``texts``, then a split of it. Returns ``(text, chunks)``."""
+    text = draw(texts)
+    return text, draw(splits_of(text, max_parts))

chunk_invariance-0.1.0/chunk_invariance.egg-info/PKG-INFO

@@ -0,0 +1,111 @@
+Metadata-Version: 2.4
+Name: chunk-invariance
+Version: 0.1.0
+Summary: Assert that a streaming filter gives the same output however its input is split into chunks.
+Author-email: Ninad Phalak <ninadphalak@gmail.com>
+Maintainer-email: Ninad Phalak <ninadphalak@gmail.com>
+License-Expression: Apache-2.0
+Project-URL: Homepage, https://github.com/ninadphalak/LLM-Shield-Proxy/tree/main/chunk-invariance
+Project-URL: Repository, https://github.com/ninadphalak/LLM-Shield-Proxy
+Project-URL: Bug Tracker, https://github.com/ninadphalak/LLM-Shield-Proxy/issues
+Keywords: streaming,testing,redaction,guardrails,chunk-boundary,property-based-testing
+Classifier: Development Status :: 4 - Beta
+Classifier: Intended Audience :: Developers
+Classifier: Framework :: Hypothesis
+Classifier: Framework :: Pytest
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.9
+Classifier: Programming Language :: Python :: 3.10
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3.13
+Classifier: Topic :: Software Development :: Testing
+Classifier: Typing :: Typed
+Requires-Python: >=3.9
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Provides-Extra: hypothesis
+Requires-Dist: hypothesis>=6; extra == "hypothesis"
+Dynamic: license-file
+
+# chunk-invariance
+
+A streaming filter (a redactor, a guardrail, a rewriter) is correct only if, for every way of
+splitting an input into chunks, the concatenated streamed output equals the output of filtering
+the whole input at once. This package turns that rule into one assertion.
+
+```python
+from chunk_invariance import assert_chunk_invariant
+
+def redact(text):                 # your filter on the whole input
+    return text.replace("sk-AbCd1234", "[KEY]")
+
+def redact_stream(chunks):        # your filter on a stream: takes chunks, yields chunks
+    for chunk in chunks:
+        yield redact(chunk)       # redacts each chunk alone
+
+def test_streaming_redaction():
+    assert_chunk_invariant(redact_stream, redact, "use sk-AbCd1234 now")
+```
+
+That test fails, and names the smallest split that breaks it:
+
+```
+chunk_invariance.core.ChunkInvarianceError: stream filter is not chunk-invariant
+  input:    'use sk-AbCd1234 now'
+  split:    ['use s', 'k-AbCd1234 now'] (cut at 5)
+  whole:    'use [KEY] now'
+  streamed: 'use sk-AbCd1234 now'
+  split #6 in the order tried; generated splits run fewest parts first
+```
+
+```bash
+pip install chunk-invariance
+pip install "chunk-invariance[hypothesis]"   # optional Hypothesis strategies
+```
+
+## What it catches
+
+- A filter that evaluates each chunk alone, so a value cut across two chunks matches neither half.
+- A filter that prepends a fixed carryover from the previous chunk. The match fires, but the start
+  of the value was already emitted with the previous chunk.
+- A hold-back that is shorter than the pattern, a missing end-of-stream flush, and output that is
+  corrupted rather than leaked (any difference from the whole-input output fails).
+
+`chunk_invariance.examples` has one correct filter (`hold_back`) and the two wrong designs
+(`design_a`, `design_b`) for the same pattern, as a worked comparison.
+
+## What it does not catch
+
+It tests a filter you can call in-process. It does not test a running gateway, where the chunks
+are decided by an upstream model, a network and an SSE parser. For that, use
+[`pii-leak-benchmark`](https://pypi.org/project/pii-leak-benchmark/), which sends
+fragmented responses through the gateway and checks what reaches the client.
+
+It checks the splits you ask for. The default is the whole input plus every single cut, which
+catches a value cut in two. `splits="all"` tries every split of a short input (up to 16
+characters); `splits=3` tries every split into at most three parts; an explicit list of chunk
+lists runs exactly those. It cannot prove a filter correct for inputs you did not give it.
+
+## API
+
+| Name | Does |
+|---|---|
+| `assert_chunk_invariant(stream_filter, whole_filter, text, *, splits="all-two-part")` | Raises `ChunkInvarianceError` (an `AssertionError`) on the first failing split, fewest parts first. Returns the number of splits checked. Works on `str` and `bytes`. |
+| `assert_chunk_invariant_async(...)` | The same, for an async generator filter, inside an async test. |
+| `two_part_splits(text)` | Every split into two non-empty parts. |
+| `all_splits(text, max_parts=None)` | Every split into at most `max_parts` non-empty parts, the whole input first. |
+| `per_chunk(process, *, state=dict, flush=None)` | Adapts `process(chunk, state) -> output` and `flush(state) -> output`, with fresh state per stream. |
+| `from_async(async_stream_filter)` | Adapts an async generator filter for the synchronous assertion. |
+| `strategies.splits_of(text, max_parts=None)` | Hypothesis strategy: random splits of `text`, shrinking toward fewer cuts. |
+| `strategies.text_and_splits(texts, max_parts=None)` | Hypothesis strategy: `(text, chunks)` pairs. |
+
+`stream_filter` is called once per split, so it must start from fresh state each time.
+
+## Background
+
+The invariant and the two wrong designs are described in Phalak, N. (2026), *Split-Boundary
+Leaks in Streaming Guardrails*, Zenodo, [10.5281/zenodo.22909585](https://doi.org/10.5281/zenodo.22909585).
+
+Standard library only. Apache-2.0. A TypeScript version with the same API is published on npm
+as `chunk-invariance`.

chunk_invariance-0.1.0/chunk_invariance.egg-info/SOURCES.txt

@@ -0,0 +1,14 @@
+LICENSE
+README.md
+pyproject.toml
+chunk_invariance/__init__.py
+chunk_invariance/adapters.py
+chunk_invariance/core.py
+chunk_invariance/examples.py
+chunk_invariance/py.typed
+chunk_invariance/strategies.py
+chunk_invariance.egg-info/PKG-INFO
+chunk_invariance.egg-info/SOURCES.txt
+chunk_invariance.egg-info/dependency_links.txt
+chunk_invariance.egg-info/requires.txt
+chunk_invariance.egg-info/top_level.txt

chunk_invariance-0.1.0/chunk_invariance.egg-info/dependency_links.txt

@@ -0,0 +1 @@
+

chunk_invariance-0.1.0/chunk_invariance.egg-info/requires.txt

@@ -0,0 +1,3 @@
+
+[hypothesis]
+hypothesis>=6

chunk_invariance-0.1.0/chunk_invariance.egg-info/top_level.txt

@@ -0,0 +1 @@
+chunk_invariance

chunk_invariance-0.1.0/pyproject.toml

@@ -0,0 +1,63 @@
+[build-system]
+requires = ["setuptools>=61.0", "wheel"]
+build-backend = "setuptools.build_meta"
+
+[project]
+name = "chunk-invariance"
+version = "0.1.0"
+description = "Assert that a streaming filter gives the same output however its input is split into chunks."
+readme = "README.md"
+requires-python = ">=3.9"
+license = "Apache-2.0"
+authors = [
+    {name = "Ninad Phalak", email = "ninadphalak@gmail.com"}
+]
+maintainers = [
+    {name = "Ninad Phalak", email = "ninadphalak@gmail.com"}
+]
+keywords = [
+    "streaming",
+    "testing",
+    "redaction",
+    "guardrails",
+    "chunk-boundary",
+    "property-based-testing",
+]
+classifiers = [
+    "Development Status :: 4 - Beta",
+    "Intended Audience :: Developers",
+    "Framework :: Hypothesis",
+    "Framework :: Pytest",
+    "Programming Language :: Python :: 3",
+    "Programming Language :: Python :: 3.9",
+    "Programming Language :: Python :: 3.10",
+    "Programming Language :: Python :: 3.11",
+    "Programming Language :: Python :: 3.12",
+    "Programming Language :: Python :: 3.13",
+    "Topic :: Software Development :: Testing",
+    "Typing :: Typed",
+]
+# No runtime dependencies. Hypothesis is optional and only chunk_invariance.strategies imports it.
+dependencies = []
+
+[project.optional-dependencies]
+hypothesis = ["hypothesis>=6"]
+
+[project.urls]
+Homepage = "https://github.com/ninadphalak/LLM-Shield-Proxy/tree/main/chunk-invariance"
+Repository = "https://github.com/ninadphalak/LLM-Shield-Proxy"
+"Bug Tracker" = "https://github.com/ninadphalak/LLM-Shield-Proxy/issues"
+
+[tool.setuptools.packages.find]
+include = ["chunk_invariance*"]
+
+[tool.setuptools.package-data]
+chunk_invariance = ["py.typed"]
+
+[tool.ruff]
+line-length = 120
+target-version = "py39"
+
+[tool.ruff.lint]
+select = ["E", "W", "F", "I"]
+ignore = ["E501"]

chunk_invariance-0.1.0/setup.cfg

@@ -0,0 +1,4 @@
+[egg_info]
+tag_build = 
+tag_date = 0
+