evalkeep 0.1.0__tar.gz

evalkeep-0.1.0/LICENSE

@@ -0,0 +1,202 @@
+
+                                 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.

evalkeep-0.1.0/PKG-INFO

@@ -0,0 +1,221 @@
+Metadata-Version: 2.4
+Name: evalkeep
+Version: 0.1.0
+Summary: Turn production agent failures into a small, reviewed regression suite.
+Keywords: llm,evals,evaluation,regression-testing,ai-agents,testing,promptfoo
+Author: Rakshita Devurkar
+Author-email: Rakshita Devurkar <13130544+rakshita-devurkar@users.noreply.github.com>
+License-Expression: Apache-2.0
+License-File: LICENSE
+Classifier: Development Status :: 3 - Alpha
+Classifier: Environment :: Console
+Classifier: Intended Audience :: Developers
+Classifier: Operating System :: OS Independent
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3.13
+Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
+Classifier: Topic :: Software Development :: Quality Assurance
+Classifier: Topic :: Software Development :: Testing
+Classifier: Typing :: Typed
+Requires-Dist: click>=8.5.0
+Requires-Dist: numpy>=2.4.6
+Requires-Dist: pydantic>=2.13.5
+Requires-Dist: pyyaml>=6.0.3
+Requires-Dist: rich>=15.0.0
+Requires-Dist: typer>=0.27.2
+Requires-Dist: anthropic>=1.0.0 ; extra == 'anthropic'
+Requires-Python: >=3.11
+Project-URL: Homepage, https://github.com/rakshita-devurkar/evalkeep
+Project-URL: Repository, https://github.com/rakshita-devurkar/evalkeep
+Project-URL: Issues, https://github.com/rakshita-devurkar/evalkeep/issues
+Project-URL: Changelog, https://github.com/rakshita-devurkar/evalkeep/blob/main/CHANGELOG.md
+Project-URL: Documentation, https://github.com/rakshita-devurkar/evalkeep/blob/main/docs/pipeline.md
+Provides-Extra: anthropic
+Description-Content-Type: text/markdown
+
+# Evalkeep
+
+**Stop fixing the same agent bug twice.** Evalkeep turns production failures
+into a small, reviewed regression suite, and tells you whether a fix held — or
+that the evidence is too thin to say.
+
+Existing eval tools *execute* tests. The hard part is deciding which of
+thousands of production traces deserve permanent coverage. Evalkeep owns that
+decision:
+
+```
+trace → failure evidence → failure family → representative case
+      → reviewed regression test → runner execution → trustworthy comparison
+```
+
+**Evalkeep does not run your agent and is not an eval framework.** It generates
+tests, delegates execution to [Promptfoo](https://promptfoo.dev), and compares
+baseline against candidate. It sits upstream of your eval runner, not next to it.
+
+## Install
+
+```bash
+git clone https://github.com/rakshita-devurkar/evalkeep && cd evalkeep
+uv sync
+```
+
+Node.js is needed only for `evalkeep run`, which shells out to Promptfoo.
+
+## Quick start
+
+One command takes a trace file to a review queue. Offline, no API key:
+
+```bash
+uv run evalkeep demo .
+uv run evalkeep init
+uv run evalkeep from-traces refund-agent/traces.jsonl
+```
+
+```
+5 trace(s) ingested
+3 failure(s) found  explicit_status x3, failed_evaluator x1, negative_feedback x2
+2 failure famil(ies)
+3 with enough evidence for a regression test
+3 of them only forbid the mistake that was observed; say what should have happened at review.
+
+Review them: evalkeep review  (3 pending)
+```
+
+That second qualifier is the honest state of a first run: nothing in a trace
+says what the agent *should* have done. Describing the failures closes it, by
+hand at review or with an analyzer configured. It stops at review on purpose —
+approving a test is a judgement.
+
+## On real agent data
+
+The example above is five traces. Here is the same pipeline on
+[tau-bench trajectories](https://huggingface.co/datasets/AgentSuite/tau-bench-trajectories):
+165 retail and airline customer-service tasks per model, each scored by
+comparing the final database state against the expected one. Real tool calls,
+and an independent verdict — the two halves a regression suite needs.
+
+```bash
+python tau-bench/prepare.py                      # ~8 MB, two models
+uv run evalkeep from-traces tau-bench/Qwen3-235B-A22B-FP8.traces.jsonl
+```
+
+```
+165 trace(s) ingested
+90 failure(s) found  explicit_status x90, failed_evaluator x90
+4 failure famil(ies)
+```
+
+Four families, from 90 failures: retail exchange and refund flows, airline
+reservation changes, and two shapes of giving up and escalating to a human.
+Build a test per failure, approve them, and run two recorded models against the
+suite one of them produced:
+
+```bash
+uv run evalkeep dataset build --all
+uv run evalkeep review
+uv run evalkeep targets add baseline  --type python --function call_api \
+  --path tau-bench/replay_Qwen3_235B_A22B_FP8.py
+uv run evalkeep targets add candidate --type python --function call_api \
+  --path tau-bench/replay_claude_4_5_sonnet_thinking_off.py
+uv run evalkeep run --target baseline && uv run evalkeep run --target candidate
+uv run evalkeep compare
+```
+
+```
+compared                           89
+baseline pass rate               4.5%
+candidate pass rate             42.7%
+difference                     +38.2%
+p-value                        0.0000
+95% interval         +27.2% to +49.2%
+McNemar's exact test: the change is unlikely to be chance.
+
+1 test(s) excluded and not counted in any rate above.
+```
+
+Baseline scoring 4.5% is the control: the tests came from its own failures, so
+it should fail nearly all of them. The four it passes are the documented
+weakness of deriving a test with nothing describing the failure — assertions
+target the last tool call, which is sometimes a harmless lookup. The excluded
+test is one whose target raised rather than answered, and it is kept out of
+every rate rather than counted as a failure.
+
+## Doing it stage by stage
+
+`from-traces` runs five commands in order. Each is a real decision with its own
+evidence, and you will want them separately once you are tuning a suite:
+
+```bash
+uv run evalkeep ingest traces.jsonl   # validate, redact, store
+uv run evalkeep detect                # evidence-backed failures
+uv run evalkeep analyze               # describe them (or: failures label)
+uv run evalkeep discover              # embed, cluster, pick representatives
+uv run evalkeep dataset build         # draft a test per representative
+```
+
+Re-running `from-traces` is safe: it skips traces it already has and rebuilds
+drafts, but never touches a test you have reviewed.
+
+## Bring your own traces
+
+```bash
+uv run evalkeep ingest spans.json --format otlp        # OpenTelemetry / OpenInference
+uv run evalkeep ingest runs.jsonl --format langsmith   # LangSmith
+```
+
+Adapters read files, never APIs — no credentials, any vendor tier. OpenTelemetry
+covers the most ground, since Langfuse, Braintrust and Phoenix all ingest OTLP.
+`evalkeep demo` writes an example export in each format.
+
+## Commands
+
+| Stage | Commands |
+| --- | --- |
+| Set up | `init`, `targets add/list/show/remove` |
+| Ingest | `ingest`, `trace list/show` |
+| Detect | `detect`, `failures list/show/confirm/dismiss/add` |
+| Analyze | `analyze`, `failures label` |
+| Group | `discover`, `clusters list/show/rename/merge/split/dismiss/restore` |
+| Build | `dataset build/list/show` |
+| Review | `review`, `dataset approve/reject/edit` |
+| Run | `export`, `run --target ...`, `runs list/show` |
+| Compare | `compare`, `baseline promote/show` |
+
+## What it guarantees
+
+- **Values are redacted before storage** — in memory, with no path around it.
+  Identifiers can be [pseudonymized](docs/security.md#identifiers) too.
+- **Automation never overwrites human judgement.** Re-running any stage
+  refreshes derived data and leaves your reviews, labels and edits alone.
+- **Nothing is exported without approval.** Generated tests are drafts.
+- **A test that never ran is not a test that failed.** Timeouts and crashed
+  providers are excluded from every rate, so an outage cannot read as a regression.
+- **One lucky pass is not a fix.** `run --repetitions N` reports a per-case
+  verdict; a case that only sometimes passes is flaky, never passing.
+- **Score changes are not overclaimed.** McNemar's exact test, and no confidence
+  interval when the sample cannot support one.
+
+## More
+
+[How it works](docs/pipeline.md) · [Privacy and security](docs/security.md) ·
+[Roadmap](docs/roadmap.md) · [Contributing](CONTRIBUTING.md) ·
+[Changelog](CHANGELOG.md)
+
+```bash
+uv sync && uv run pytest              # 985 tests
+uv run ruff check . && uv run mypy    # lint and strict types
+```
+
+`EVALKEEP_E2E=1 uv run pytest` also runs the suite against real Promptfoo.
+Exit codes: `0` success, `1` ran but some records were rejected, `2` could not
+run.
+
+0.1 is feature-complete and not yet released to PyPI. Multi-turn replay and
+longitudinal failure history are still open; see the
+[roadmap](docs/roadmap.md).
+
+## License
+
+Apache-2.0. See [LICENSE](LICENSE).

evalkeep-0.1.0/README.md

@@ -0,0 +1,184 @@
+# Evalkeep
+
+**Stop fixing the same agent bug twice.** Evalkeep turns production failures
+into a small, reviewed regression suite, and tells you whether a fix held — or
+that the evidence is too thin to say.
+
+Existing eval tools *execute* tests. The hard part is deciding which of
+thousands of production traces deserve permanent coverage. Evalkeep owns that
+decision:
+
+```
+trace → failure evidence → failure family → representative case
+      → reviewed regression test → runner execution → trustworthy comparison
+```
+
+**Evalkeep does not run your agent and is not an eval framework.** It generates
+tests, delegates execution to [Promptfoo](https://promptfoo.dev), and compares
+baseline against candidate. It sits upstream of your eval runner, not next to it.
+
+## Install
+
+```bash
+git clone https://github.com/rakshita-devurkar/evalkeep && cd evalkeep
+uv sync
+```
+
+Node.js is needed only for `evalkeep run`, which shells out to Promptfoo.
+
+## Quick start
+
+One command takes a trace file to a review queue. Offline, no API key:
+
+```bash
+uv run evalkeep demo .
+uv run evalkeep init
+uv run evalkeep from-traces refund-agent/traces.jsonl
+```
+
+```
+5 trace(s) ingested
+3 failure(s) found  explicit_status x3, failed_evaluator x1, negative_feedback x2
+2 failure famil(ies)
+3 with enough evidence for a regression test
+3 of them only forbid the mistake that was observed; say what should have happened at review.
+
+Review them: evalkeep review  (3 pending)
+```
+
+That second qualifier is the honest state of a first run: nothing in a trace
+says what the agent *should* have done. Describing the failures closes it, by
+hand at review or with an analyzer configured. It stops at review on purpose —
+approving a test is a judgement.
+
+## On real agent data
+
+The example above is five traces. Here is the same pipeline on
+[tau-bench trajectories](https://huggingface.co/datasets/AgentSuite/tau-bench-trajectories):
+165 retail and airline customer-service tasks per model, each scored by
+comparing the final database state against the expected one. Real tool calls,
+and an independent verdict — the two halves a regression suite needs.
+
+```bash
+python tau-bench/prepare.py                      # ~8 MB, two models
+uv run evalkeep from-traces tau-bench/Qwen3-235B-A22B-FP8.traces.jsonl
+```
+
+```
+165 trace(s) ingested
+90 failure(s) found  explicit_status x90, failed_evaluator x90
+4 failure famil(ies)
+```
+
+Four families, from 90 failures: retail exchange and refund flows, airline
+reservation changes, and two shapes of giving up and escalating to a human.
+Build a test per failure, approve them, and run two recorded models against the
+suite one of them produced:
+
+```bash
+uv run evalkeep dataset build --all
+uv run evalkeep review
+uv run evalkeep targets add baseline  --type python --function call_api \
+  --path tau-bench/replay_Qwen3_235B_A22B_FP8.py
+uv run evalkeep targets add candidate --type python --function call_api \
+  --path tau-bench/replay_claude_4_5_sonnet_thinking_off.py
+uv run evalkeep run --target baseline && uv run evalkeep run --target candidate
+uv run evalkeep compare
+```
+
+```
+compared                           89
+baseline pass rate               4.5%
+candidate pass rate             42.7%
+difference                     +38.2%
+p-value                        0.0000
+95% interval         +27.2% to +49.2%
+McNemar's exact test: the change is unlikely to be chance.
+
+1 test(s) excluded and not counted in any rate above.
+```
+
+Baseline scoring 4.5% is the control: the tests came from its own failures, so
+it should fail nearly all of them. The four it passes are the documented
+weakness of deriving a test with nothing describing the failure — assertions
+target the last tool call, which is sometimes a harmless lookup. The excluded
+test is one whose target raised rather than answered, and it is kept out of
+every rate rather than counted as a failure.
+
+## Doing it stage by stage
+
+`from-traces` runs five commands in order. Each is a real decision with its own
+evidence, and you will want them separately once you are tuning a suite:
+
+```bash
+uv run evalkeep ingest traces.jsonl   # validate, redact, store
+uv run evalkeep detect                # evidence-backed failures
+uv run evalkeep analyze               # describe them (or: failures label)
+uv run evalkeep discover              # embed, cluster, pick representatives
+uv run evalkeep dataset build         # draft a test per representative
+```
+
+Re-running `from-traces` is safe: it skips traces it already has and rebuilds
+drafts, but never touches a test you have reviewed.
+
+## Bring your own traces
+
+```bash
+uv run evalkeep ingest spans.json --format otlp        # OpenTelemetry / OpenInference
+uv run evalkeep ingest runs.jsonl --format langsmith   # LangSmith
+```
+
+Adapters read files, never APIs — no credentials, any vendor tier. OpenTelemetry
+covers the most ground, since Langfuse, Braintrust and Phoenix all ingest OTLP.
+`evalkeep demo` writes an example export in each format.
+
+## Commands
+
+| Stage | Commands |
+| --- | --- |
+| Set up | `init`, `targets add/list/show/remove` |
+| Ingest | `ingest`, `trace list/show` |
+| Detect | `detect`, `failures list/show/confirm/dismiss/add` |
+| Analyze | `analyze`, `failures label` |
+| Group | `discover`, `clusters list/show/rename/merge/split/dismiss/restore` |
+| Build | `dataset build/list/show` |
+| Review | `review`, `dataset approve/reject/edit` |
+| Run | `export`, `run --target ...`, `runs list/show` |
+| Compare | `compare`, `baseline promote/show` |
+
+## What it guarantees
+
+- **Values are redacted before storage** — in memory, with no path around it.
+  Identifiers can be [pseudonymized](docs/security.md#identifiers) too.
+- **Automation never overwrites human judgement.** Re-running any stage
+  refreshes derived data and leaves your reviews, labels and edits alone.
+- **Nothing is exported without approval.** Generated tests are drafts.
+- **A test that never ran is not a test that failed.** Timeouts and crashed
+  providers are excluded from every rate, so an outage cannot read as a regression.
+- **One lucky pass is not a fix.** `run --repetitions N` reports a per-case
+  verdict; a case that only sometimes passes is flaky, never passing.
+- **Score changes are not overclaimed.** McNemar's exact test, and no confidence
+  interval when the sample cannot support one.
+
+## More
+
+[How it works](docs/pipeline.md) · [Privacy and security](docs/security.md) ·
+[Roadmap](docs/roadmap.md) · [Contributing](CONTRIBUTING.md) ·
+[Changelog](CHANGELOG.md)
+
+```bash
+uv sync && uv run pytest              # 985 tests
+uv run ruff check . && uv run mypy    # lint and strict types
+```
+
+`EVALKEEP_E2E=1 uv run pytest` also runs the suite against real Promptfoo.
+Exit codes: `0` success, `1` ran but some records were rejected, `2` could not
+run.
+
+0.1 is feature-complete and not yet released to PyPI. Multi-turn replay and
+longitudinal failure history are still open; see the
+[roadmap](docs/roadmap.md).
+
+## License
+
+Apache-2.0. See [LICENSE](LICENSE).