structured-eval 0.1.0__tar.gz

structured_eval-0.1.0/LICENSE

@@ -0,0 +1,201 @@
+                                 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
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   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.

structured_eval-0.1.0/PKG-INFO

@@ -0,0 +1,322 @@
+Metadata-Version: 2.4
+Name: structured-eval
+Version: 0.1.0
+Summary: The LLM Structured Output Evaluation Framework
+License: Apache-2.0
+Project-URL: Homepage, https://github.com/kirillpechurin/structured-eval
+Project-URL: Repository, https://github.com/kirillpechurin/structured-eval
+Project-URL: Issues, https://github.com/kirillpechurin/structured-eval/issues
+Keywords: structured llm evaluation,structured eval,json eval,llm,structured output,pydantic
+Classifier: Development Status :: 4 - Beta
+Classifier: Intended Audience :: Developers
+Classifier: License :: OSI Approved :: Apache Software License
+Classifier: Operating System :: OS Independent
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3.13
+Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
+Classifier: Topic :: Software Development :: Testing
+Classifier: Typing :: Typed
+Requires-Python: >=3.12
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: pydantic>=2.5.0
+Provides-Extra: yaml
+Requires-Dist: pyyaml>=6.0; extra == "yaml"
+Provides-Extra: fuzzy
+Requires-Dist: rapidfuzz>=3.0.0; extra == "fuzzy"
+Provides-Extra: jsonschema
+Requires-Dist: jsonschema>=4.20.0; extra == "jsonschema"
+Provides-Extra: rules
+Requires-Dist: jsonpath-ng>=1.6.0; extra == "rules"
+Provides-Extra: diff
+Requires-Dist: deepdiff>=7.0.0; extra == "diff"
+Provides-Extra: align
+Requires-Dist: scipy>=1.13.0; extra == "align"
+Provides-Extra: deepeval
+Requires-Dist: deepeval>=3.0.0; extra == "deepeval"
+Provides-Extra: langsmith
+Requires-Dist: langsmith>=0.8.0; extra == "langsmith"
+Provides-Extra: all
+Requires-Dist: structured-eval[align,diff,fuzzy,jsonschema,rules,yaml]; extra == "all"
+Dynamic: license-file
+
+# structured-eval
+
+**A declarative, field-level evaluation framework for LLM structured outputs (JSON/YAML).**
+
+Getting an LLM to return well-formed JSON is mostly a solved problem — it parses,
+it fits the schema, the types line up. But well-formed isn't the same as *right*.
+The shape can be flawless while a price is wrong, a date is invented, or a status
+quietly contradicts the rest of the record. Structural checks wave all of that
+through.
+
+structured-eval looks at what those checks skip: **the values themselves**. It
+scores your output field by field, so you don't just learn *that* a response is
+off — you see *which* fields matched and which didn't, where to look first, and,
+across a dataset, which fields your model keeps getting wrong.
+
+structured-eval lets you check not just that the JSON is valid, but that the
+data itself is correct.
+
+## The gap it closes
+
+Correctness is a ladder — each level assumes the ones below it:
+
+|           |                                                      |
+|-----------|------------------------------------------------------|
+| **L0–L3** | structure: parses · types · required · no extras     |
+| **L4**    | values are close to expected                         |
+| **L5**    | values are grounded in the source (no hallucination) |
+| **L6**    | fields are logically consistent with one another     |
+
+L0–L3 is where most tools stop. **L4–L6 is where structured-eval earns its keep.**
+See [the introduction](docs/introduction.md) for the full ladder.
+
+## Install
+
+```bash
+pip install structured-eval          # core depends only on Pydantic
+pip install "structured-eval[all]"   # + YAML, fuzzy, schema, rules, scipy alignment…
+```
+
+Optional features live behind [extras](docs/getting-started.md#install) — install
+only what you need.
+
+## Quick start
+
+A model extracted a course record; you have the canonical one to check it against.
+The two are *structurally* identical — same keys, mixed types, a nested array of
+objects — but several values are off. A small config says *how* to judge each
+field, and the report tells you exactly where the output stands:
+
+```python
+from structured_eval import evaluate
+from structured_eval.models import EvalConfig, FieldConfig
+from structured_eval.metrics import Numeric, TokenF1
+
+expected = {
+    "course_id": "COURSE-101",
+    "title": "Introduction to Python",
+    "published": True,
+    "duration_hours": 12,
+    "rating": 4.8,
+    "modules": [
+        {"name": "Basics", "lessons": 5},
+        {"name": "Functions", "lessons": 4},
+        {"name": "Classes", "lessons": 3},
+    ],
+}
+
+actual = {
+    "course_id": "COURSE-101",
+    "title": "Intro to Python",  # paraphrased
+    "published": True,
+    "duration_hours": 10,  # off by 2
+    "rating": 4.5,  # off by 0.3
+    "modules": [
+        {"name": "Basics", "lessons": 5},
+        {"name": "Functions", "lessons": 4},
+    ],  # "Classes" module missing
+}
+
+config = EvalConfig(fields={
+    "title": FieldConfig(metrics=[TokenF1()]),                # reward paraphrases
+    "duration_hours": FieldConfig(metrics=[Numeric(tolerance=2)]),
+    "rating": FieldConfig(metrics=[Numeric(tolerance=0.5)]),  # close enough is fine
+})
+
+report = evaluate(actual, expected, config)
+
+report.score  # 0.8889  — close, with the gaps pinpointed
+report.field_scores["title"].score  # 0.6667  — paraphrase gets partial credit
+report.field_scores["duration_hours"].score  # 1.0     — within tolerance
+report.field_scores["modules"].score  # 0.6667  — 2 of 3 modules recovered
+report.field_scores["modules[0]"].score  # 1.0     — first module is spot-on
+```
+
+Every field is scored — nested objects and array elements included — so you see
+not a single pass/fail but exactly which fields hold up and which don't.
+
+### Sensible default metrics
+
+The config is optional. structured-eval ships a default metric for every node
+type, so you only configure the fields where the default isn't what you want —
+the rest just work. With no config at all, the same data is scored by those
+defaults:
+
+```python
+report = evaluate(actual, expected)  # no config
+
+report.score  # 0.4444  — scored entirely by the defaults
+report.field_scores["title"].score  # 0.0  — exact match: "Intro to Python" ≠ "Introduction to Python"
+```
+
+Each node type gets a structural default, and every node's headline score (its
+*representative*) defaults to the mean of its own metrics:
+
+| Node                 | Default metric   | What it does                                   |
+|----------------------|------------------|------------------------------------------------|
+| scalar (leaf)        | `ExactMatch`     | the value must match exactly                   |
+| object               | `ObjectAccuracy` | mean correctness of its fields                 |
+| array                | `ArrayAccuracy`  | mean correctness of its aligned elements       |
+| any node (headline)  | `MeanScore`      | the node's representative = mean of its metrics |
+
+Exact match is a strict baseline — it punishes every paraphrase and rounded value
+as wrong, which is why the no-config score is low. Tuning metrics per field, as in
+the first example, is how you tell the evaluator what "close enough" means for
+*your* data. The defaults and the representative score are covered in
+[the evaluation model](docs/core-concepts/evaluation-model.md) and the
+[metric catalog](docs/metrics/index.md).
+
+## Explore — every level of correctness
+
+structured-eval covers the whole ladder, L0 through L6. Each level has a tool and
+a concept page behind it:
+
+| Level               | The question                        | Reach for                                                   | Learn more                                                             |
+|---------------------|-------------------------------------|-------------------------------------------------------------|------------------------------------------------------------------------|
+| **L0–L3** structure | does it parse / fit the schema?     | `SchemaValidity`                                            | [schema validity](docs/metrics/catalog/schema-validity.md)             |
+| **L4** values       | is each value right?                | field metrics — `ExactMatch`, `Numeric`, `TokenF1`, `Fuzzy` | [comparison is a metric](docs/core-concepts/comparison-is-a-metric.md) |
+| **L4** roll-up      | how do fields & elements aggregate? | `ObjectF1` / `ArrayF1`, alignment, weights                  | [array alignment](docs/core-concepts/array-alignment.md)               |
+| **L5** faithfulness | is it grounded in the source?       | `FieldFaithfulness(source=…)`                               | [field faithfulness](docs/metrics/catalog/field_faithfulness.md)       |
+| **L6** logic        | are fields mutually consistent?     | `RulePassRate` + `Rule` DSL                                 | [rule pass rate](docs/metrics/catalog/rule-pass-rate.md)               |
+
+**L0–L3 — structure.** Validate against a Pydantic model or JSON Schema, with no
+ground-truth answer:
+
+```python
+from pydantic import BaseModel
+from structured_eval import evaluate
+from structured_eval.models import EvalConfig
+from structured_eval.metrics import SchemaValidity
+
+
+class Course(BaseModel):
+    title: str
+    duration_hours: int
+
+
+report = evaluate(
+    actual={"title": "ML", "duration_hours": "twelve"},
+    expected=None,
+    config=EvalConfig(key_metric=SchemaValidity(Course))
+)
+report.score  # 0.0
+report.metrics["schema_validity"].root().extra["schema_errors"]
+#   {'type_errors': ['duration_hours'], 'missing_required': [], 'extra_fields': []}
+```
+
+**L4 — values.** Pick *how* each field is judged — exact match is just the default
+(*comparison is a metric*):
+
+```python
+from structured_eval import evaluate
+from structured_eval.models import EvalConfig, FieldConfig
+from structured_eval.metrics import Numeric, TokenF1
+
+report = evaluate(
+    actual={"title": "Intro to Python", "duration_hours": 11},
+    expected={"title": "Introduction to Python", "duration_hours": 12},
+    config=EvalConfig(fields={
+        "title": FieldConfig(metrics=[TokenF1()]),  # token overlap
+        "duration_hours": FieldConfig(metrics=[Numeric(tolerance=2)]),
+    }),
+)
+report.field_scores["title"].score  # 0.6667  — partial credit for a paraphrase
+report.field_scores["duration_hours"].score  # 1.0     — within tolerance
+```
+
+Fields roll up into objects and arrays with precision / recall / F1, and arrays are
+[aligned](docs/core-concepts/array-alignment.md) by index, key, or optimally:
+
+```python
+from structured_eval import evaluate
+from structured_eval.models import EvalConfig
+from structured_eval.metrics import ObjectF1
+
+report = evaluate(
+    actual={"a": 1, "b": 9},
+    expected={"a": 1, "b": 2, "c": 3},
+    config=EvalConfig(metrics=[ObjectF1()])
+)
+
+report.metrics["object_f1"].root()  # 0.4
+```
+
+**L5 — faithfulness.** Catch hallucinations by checking each value against its
+source — no `expected` required:
+
+```python
+from structured_eval import evaluate
+from structured_eval.models import EvalConfig
+from structured_eval.metrics import FieldFaithfulness
+
+report = evaluate(
+    actual={"title": "Introduction to Python", "duration_hours": 40},
+    expected=None,
+    config=EvalConfig(metrics=[FieldFaithfulness()]),
+    source="Course: Introduction to Python. Duration: 12 hours.",
+)
+
+report.metrics["field_faithfulness"].by_path  # {'title': 1.0, 'duration_hours': 0.0 ← 40 ≠ 12}
+```
+
+**L6 — logic.** Assert cross-field business rules with a small DSL:
+
+```python
+from structured_eval import evaluate
+from structured_eval.models import EvalConfig
+from structured_eval.metrics import Rule, RulePassRate
+
+report = evaluate(
+    actual={"subtotal": 100, "tax": 20, "total": 130},
+    expected=None,
+    config=EvalConfig(key_metric=RulePassRate([Rule("$.total").eq("$.subtotal + $.tax")]))
+)
+
+report.score  # 0.0  — 130 ≠ 120
+```
+
+## Scale it
+
+Evaluate a whole dataset, or measure how stable a prompt is across repeated runs:
+
+```python
+from structured_eval import evaluate_batch, evaluate_consistency
+from structured_eval.models import Sample
+
+# one report per sample + dataset aggregates
+batch = evaluate_batch([
+    Sample(actual={"x": 1}, expected={"x": 1}),
+    Sample(actual={"x": 1}, expected={"x": 2}),
+])
+batch.score  # 0.5
+batch.perfect_response_rate  # 0.5
+
+# repeated runs of the same prompt → which fields drift?
+runs = [
+    Sample(actual={"sentiment": "positive", "score": 0.9}, expected={"sentiment": "positive", "score": 0.9}),
+    Sample(actual={"sentiment": "positive", "score": 0.9}, expected={"sentiment": "positive", "score": 0.9}),
+    Sample(actual={"sentiment": "neutral", "score": 0.9}, expected={"sentiment": "positive", "score": 0.9}),
+]
+report = evaluate_consistency(runs, variance_threshold=0.05)
+report.stable_fields  # ['score']
+report.unstable_fields  # ['sentiment']  — flipped on one run
+```
+
+## Documentation
+
+- **[Introduction](docs/introduction.md)** — the L0–L6 ladder and why values matter.
+- **[Getting started](docs/getting-started.md)** — install → first evaluation →
+  reading and tuning the report.
+- **Core concepts** — [the evaluation model](docs/core-concepts/evaluation-model.md) ·
+  [comparison is a metric](docs/core-concepts/comparison-is-a-metric.md) ·
+  [array alignment](docs/core-concepts/array-alignment.md)
+- **[Evaluation functions](docs/evaluation/index.md)** — `evaluate`,
+  `evaluate_batch`, `evaluate_consistency`.
+- **[Metric catalog](docs/metrics/index.md)** — every metric, plus how to write your own.
+
+## License
+
+Apache-2.0 — see [LICENSE](LICENSE).