flowharness-evaluation 0.1.0__tar.gz

flowharness_evaluation-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.

flowharness_evaluation-0.1.0/PKG-INFO

@@ -0,0 +1,26 @@
+Metadata-Version: 2.4
+Name: flowharness-evaluation
+Version: 0.1.0
+Summary: Evaluation context (FLOWHARNESS_SPEC.md §6.6) — eval suites, golden-set corpus, signed eval runs, the gate service
+License-Expression: Apache-2.0
+License-File: LICENSE
+Classifier: Development Status :: 3 - Alpha
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3.13
+Requires-Dist: flowharness-core==0.1.0
+Requires-Dist: sqlalchemy>=2.0
+Requires-Dist: alembic>=1.13
+Requires-Python: >=3.12
+Project-URL: Homepage, https://github.com/flowharness-ai
+Description-Content-Type: text/markdown
+
+# flowharness-evaluation
+
+Evaluation bounded context — owns eval suites, golden-set corpus, signed eval runs, gate policies, and the gate service (verdict + scoring logic). Provides the domain layer for evaluating AI surfaces against a committed test suite, applying gate policies, and producing signed evaluation reports.
+
+This is an **internal library package** of the FlowHarness toolchain, published to PyPI as a transitive dependency of [`flowharness`](https://github.com/flowharness-ai) (the scan CLI) and [`flowharness-ci-runner`](https://github.com/flowharness-ai). Install those packages instead of depending on `flowharness-evaluation` directly.
+
+> **Note:** this package currently installs a top-level module named `evaluation`. Prefer `uvx` (isolated environments) over `pip install` into a shared venv until the 0.2 module-namespace rename lands.
+
+**License:** Apache-2.0  
+**Homepage:** https://github.com/flowharness-ai

flowharness_evaluation-0.1.0/README.md

@@ -0,0 +1,10 @@
+# flowharness-evaluation
+
+Evaluation bounded context — owns eval suites, golden-set corpus, signed eval runs, gate policies, and the gate service (verdict + scoring logic). Provides the domain layer for evaluating AI surfaces against a committed test suite, applying gate policies, and producing signed evaluation reports.
+
+This is an **internal library package** of the FlowHarness toolchain, published to PyPI as a transitive dependency of [`flowharness`](https://github.com/flowharness-ai) (the scan CLI) and [`flowharness-ci-runner`](https://github.com/flowharness-ai). Install those packages instead of depending on `flowharness-evaluation` directly.
+
+> **Note:** this package currently installs a top-level module named `evaluation`. Prefer `uvx` (isolated environments) over `pip install` into a shared venv until the 0.2 module-namespace rename lands.
+
+**License:** Apache-2.0  
+**Homepage:** https://github.com/flowharness-ai

flowharness_evaluation-0.1.0/pyproject.toml

@@ -0,0 +1,41 @@
+[project]
+name = "flowharness-evaluation"
+version = "0.1.0"
+readme = "README.md"
+description = "Evaluation context (FLOWHARNESS_SPEC.md §6.6) — eval suites, golden-set corpus, signed eval runs, the gate service"
+requires-python = ">=3.12"
+license = "Apache-2.0"
+license-files = ["LICENSE"]
+classifiers = [
+    "Development Status :: 3 - Alpha",
+    "Programming Language :: Python :: 3.12",
+    "Programming Language :: Python :: 3.13",
+]
+dependencies = [
+    "flowharness-core==0.1.0",
+    "sqlalchemy>=2.0",
+    "alembic>=1.13",
+]
+
+[project.urls]
+Homepage = "https://github.com/flowharness-ai"
+
+[build-system]
+requires = ["uv_build>=0.9.0,<0.10.0"]
+build-backend = "uv_build"
+
+[tool.uv.build-backend]
+module-name = "evaluation"
+source-exclude = [
+    "src/evaluation/api",
+    "src/evaluation/application",
+    "src/evaluation/infrastructure",
+]
+wheel-exclude = [
+    "evaluation/api",
+    "evaluation/application",
+    "evaluation/infrastructure",
+]
+
+[tool.uv.sources.flowharness-core]
+workspace = true

flowharness_evaluation-0.1.0/pyproject.toml.orig

@@ -0,0 +1,51 @@
+[project]
+name = "flowharness-evaluation"
+version = "0.1.0"
+readme = "README.md"
+description = "Evaluation context (FLOWHARNESS_SPEC.md §6.6) — eval suites, golden-set corpus, signed eval runs, the gate service"
+requires-python = ">=3.12"
+license = "Apache-2.0"
+license-files = ["LICENSE"]
+classifiers = [
+    "Development Status :: 3 - Alpha",
+    "Programming Language :: Python :: 3.12",
+    "Programming Language :: Python :: 3.13",
+]
+# Internal deps pin ==0.1.0: the six OSS packages version in lockstep (release.yml builds
+# them from one tag), and [tool.uv.sources] workspace=true overrides these locally.
+dependencies = ["flowharness-core==0.1.0", "sqlalchemy>=2.0", "alembic>=1.13"]
+
+[project.urls]
+Homepage = "https://github.com/flowharness-ai"
+
+[build-system]
+requires = ["uv_build>=0.9.0,<0.10.0"]
+build-backend = "uv_build"
+
+[tool.uv.build-backend]
+module-name = "evaluation"
+# --- Pruned public surface (decision 2026-07-30: funnel-open / platform-closed) -------
+# `flowharness-evaluation` is published only because `flowharness-ci` reads the public gate
+# formats from it (domain.gate_policy / domain.report / domain.errors — verified as the whole
+# of its closure). So the dists ship `evaluation/__init__.py` + `evaluation/domain/` + py.typed
+# and nothing else: `application/` (service layer, UoW, staging), `infrastructure/`
+# (SQLAlchemy tables, repositories, migrations) and `api/` are the platform's, not the OSS
+# surface. `evaluation/migrations/` (the alembic revisions = the platform DB schema) already
+# sits outside `src/` and uv_build never ships it — verified against the baseline build.
+# Zero code change needed: `evaluation/__init__.py` imports nothing and
+# `evaluation/domain/__init__.py` imports only from `evaluation.domain`.
+# BOTH lists are required — see the note in platform/core/pyproject.toml.
+# Enforced by tests/packaging/test_public_surface_manifest.py.
+source-exclude = [
+    "src/evaluation/api",
+    "src/evaluation/application",
+    "src/evaluation/infrastructure",
+]
+wheel-exclude = [
+    "evaluation/api",
+    "evaluation/application",
+    "evaluation/infrastructure",
+]
+
+[tool.uv.sources]
+flowharness-core = { workspace = true }

flowharness_evaluation-0.1.0/src/evaluation/__init__.py

@@ -0,0 +1,7 @@
+"""Evaluation bounded context (FLOWHARNESS_SPEC.md §6.6).
+
+Owns eval suites, golden-set corpus, signed eval runs, the gate service.
+
+Near-empty four-layer package (domain / application / infrastructure / api) so
+the import-linter dependency graph is real from day one (§8).
+"""

flowharness_evaluation-0.1.0/src/evaluation/domain/__init__.py

@@ -0,0 +1,9 @@
+"""Evaluation domain layer (FLOWHARNESS_SPEC.md §6.6).
+
+Dataclass value objects, entities, and domain services. Imports stdlib + core
+only — enforced by import-linter (dependency rule 4, §8).
+"""
+
+from evaluation.domain.errors import EvaluationError
+
+__all__ = ["EvaluationError"]

flowharness_evaluation-0.1.0/src/evaluation/domain/corpus.py

@@ -0,0 +1,25 @@
+"""Corpus implementations for Evaluation (FLOWHARNESS_SPEC §6.6, D9a-9).
+
+`GoldenSet` is the p1 `core.Corpus`: a curated set of eval-case assets, plus a CONTENT-ADDRESSED
+`corpus_ref` so "same corpus" comparisons are structural — two runs over the same case set share a
+`corpus_ref`, which is what makes "a run binds an exact (candidate, cell, corpus)" enforceable. The
+platform derives the ref from the suite at ingest, so a runner cannot forge which corpus it ran
+against. `ReplayWindow` (the second `Corpus`, over real `GenerationTrace`s) lands in p3.
+"""
+
+import hashlib
+from dataclasses import dataclass
+
+from core import canonical_json
+
+
+@dataclass(frozen=True, slots=True)
+class GoldenSet:
+    case_asset_ids: tuple[str, ...]
+
+    def cases(self) -> tuple[str, ...]:  # the core.Corpus port: the case specs to run against
+        return tuple(sorted(self.case_asset_ids))
+
+    def corpus_ref(self) -> str:
+        body = canonical_json(list(self.cases())).encode("utf-8")
+        return "goldenset:" + hashlib.sha256(body).hexdigest()

flowharness_evaluation-0.1.0/src/evaluation/domain/errors.py

@@ -0,0 +1,11 @@
+"""Evaluation domain errors (FLOWHARNESS_SPEC.md §8.1, P&G ch.1).
+
+The context owns a DomainError subclass so invariant failures carry a stable,
+machine-readable code instead of leaking as bare ValueErrors across the boundary.
+"""
+
+from core import DomainError
+
+
+class EvaluationError(DomainError):
+    """Base of the evaluation context's domain-error hierarchy."""

flowharness_evaluation-0.1.0/src/evaluation/domain/eval_run.py

@@ -0,0 +1,120 @@
+"""EvalRun aggregate root (FLOWHARNESS_SPEC §6.6).
+
+A federated run an external CI runner uploads: it lands `RECEIVED`, then the ingestion UoW resolves
+it to `VERIFIED` or `REJECTED` (HMAC + scope + idempotency) — synchronously, in one transaction. The
+run binds an exact `(candidate, cell, corpus_ref)` so comparisons stay honest; `corpus_ref` is
+platform-derived from the suite (a runner cannot forge it). A terminal run cannot transition again.
+Subclasses `core.EventRecorder`; invariants raise `EvaluationError`. Insert-only (no `row_version`):
+the row is written once in its final state.
+"""
+
+from dataclasses import dataclass, field
+from enum import Enum
+
+from core import CandidateRef, Cell, Event, EventRecorder
+from evaluation.domain.errors import EvaluationError
+from evaluation.domain.events import EvalRunRejected, EvalRunVerified
+from evaluation.domain.report import EvalReport
+
+
+class RunStatus(Enum):
+    RECEIVED = "received"
+    VERIFIED = "verified"
+    REJECTED = "rejected"
+
+
+@dataclass(frozen=True, slots=True)
+class RunnerIdentity:
+    runner_ref: str  # the AGENT token actor_id (e.g. "apitoken:…")
+    nonce: str  # the runner-supplied nonce — the idempotency/replay axis
+
+    def __post_init__(self) -> None:
+        if not self.runner_ref or not self.nonce:
+            raise EvaluationError("runner_identity_incomplete")
+
+
+@dataclass(slots=True)
+class EvalRun(EventRecorder):
+    run_id: str
+    org_id: str
+    suite_id: str
+    candidate: CandidateRef
+    cell: Cell
+    corpus_ref: str
+    runner: RunnerIdentity
+    report: EvalReport
+    status: RunStatus = RunStatus.RECEIVED
+    rejection_reason: str | None = None
+    sequence_no: int = 0
+    # PR-P1.5 (§4.8.2): context-diff fields. Pure passthrough to the row/event (no domain rule),
+    # mirroring how corpus_ref/runner transit the aggregate. cch is journaled; changed_paths is
+    # row-only (K6). Both additive-defaulted so every existing constructor stays valid.
+    context_change_hash: str | None = None
+    changed_paths: tuple[str, ...] | None = None
+    _events: list[Event] = field(default_factory=list, init=False)
+
+    @classmethod
+    def receive(
+        cls,
+        *,
+        run_id: str,
+        org_id: str,
+        suite_id: str,
+        candidate: CandidateRef,
+        cell: Cell,
+        corpus_ref: str,
+        runner: RunnerIdentity,
+        report: EvalReport,
+        context_change_hash: str | None = None,
+        changed_paths: tuple[str, ...] | None = None,
+    ) -> "EvalRun":
+        return cls(
+            run_id=run_id,
+            org_id=org_id,
+            suite_id=suite_id,
+            candidate=candidate,
+            cell=cell,
+            corpus_ref=corpus_ref,
+            runner=runner,
+            report=report,
+            context_change_hash=context_change_hash,
+            changed_paths=changed_paths,
+        )
+
+    def _guard_received(self) -> None:
+        if self.status is not RunStatus.RECEIVED:
+            raise EvaluationError("eval_run_terminal")
+
+    def mark_verified(self) -> None:
+        self._guard_received()
+        self.status = RunStatus.VERIFIED
+        self._record(
+            EvalRunVerified(
+                org_id=self.org_id,
+                run_id=self.run_id,
+                suite_id=self.suite_id,
+                candidate_kind=self.candidate.kind.value,
+                candidate_ref=self.candidate.ref,
+                cell_harness=self.cell.harness,
+                cell_model=self.cell.model,
+                corpus_ref=self.corpus_ref,
+                report_hash=self.report.content_hash(),
+                pass_rate_bp=self.report.pass_rate_bp.value,
+                regressions_count=self.report.regressions_count,
+                new_passing_count=self.report.new_passing_count,
+                context_change_hash=self.context_change_hash,  # None ⇒ omitted from payload (D5)
+            )
+        )
+
+    def mark_rejected(self, reason: str) -> None:
+        self._guard_received()
+        self.status = RunStatus.REJECTED
+        self.rejection_reason = reason
+        self._record(
+            EvalRunRejected(
+                org_id=self.org_id,
+                run_id=self.run_id,
+                suite_id=self.suite_id,
+                rejection_reason=reason,
+            )
+        )

flowharness_evaluation-0.1.0/src/evaluation/domain/events.py

@@ -0,0 +1,89 @@
+"""Evaluation domain events (FLOWHARNESS_SPEC §4.2, §6.6): past-tense facts the aggregates raise.
+
+Lean, float-free payloads (ids + ints + bp-as-int + bool). `EvalSuiteCreated` carries the suite's
+shape (counts + thresholds, not the case bodies). `EvalRunVerified`/`EvalRunRejected` (S4) carry the
+run's identity + the report's `content_hash` + bp/counts — the gate's evidence, never raw secrets.
+"""
+
+from dataclasses import dataclass
+
+from core import Event
+
+
+@dataclass(frozen=True, slots=True)
+class EvalSuiteCreated(Event):
+    org_id: str
+    suite_id: str
+    playbook_id: str
+    case_count: int
+    min_pass_rate_bp: int
+    max_regressions: int
+    vs_baseline: bool
+
+
+@dataclass(frozen=True, slots=True)
+class EvalRunVerified(Event):
+    org_id: str
+    run_id: str
+    suite_id: str
+    candidate_kind: str
+    candidate_ref: str
+    cell_harness: str
+    cell_model: str
+    corpus_ref: str
+    report_hash: str  # the EvalReport content_hash — re-verifiable evidence (never a raw body)
+    pass_rate_bp: int
+    regressions_count: int
+    new_passing_count: int
+    # PR-P1.5 (§4.8.2): additive optional cch. OMITTED from the journal payload when None (skip-None
+    # in journal_mapping._payload) → byte-identical entry_hash for runs that don't carry it (§9.3).
+    context_change_hash: str | None = None
+
+
+@dataclass(frozen=True, slots=True)
+class EvalRunRejected(Event):
+    org_id: str
+    run_id: str
+    suite_id: str
+    rejection_reason: str
+
+
+# Gate-decision facts (§6.6, D9b-6): one per outcome so consumers (Catalog/Rollout/Signals) can
+# subscribe per-verdict. Evidence-only payload — `run_count` (how many verified runs the decision
+# weighed), never any report body or secret. `gate_eval_id` is the minted aggregate the event hangs
+# off (sequence_no 0, a fresh id → no collision with the per-aggregate/per-org journal chain).
+@dataclass(frozen=True, slots=True)
+class _GateEvent(Event):
+    org_id: str
+    gate_eval_id: str
+    candidate_kind: str
+    candidate_ref: str
+    verdict: str  # the core.Verdict value, redundant-but-explicit per the distinct event type
+    run_count: int
+    policy_version: str
+
+
+@dataclass(frozen=True, slots=True)
+class GatePolicySet(Event):
+    org_id: str
+    scope: str
+    required_cells: tuple[tuple[str, str], ...]  # (harness, model) pairs — full point-in-time audit
+    min_pass_rate_bp: int
+    max_regressions: int
+    human_judged: bool
+    version: str
+
+
+@dataclass(frozen=True, slots=True)
+class GatePassed(_GateEvent):
+    pass
+
+
+@dataclass(frozen=True, slots=True)
+class GateFailed(_GateEvent):
+    pass
+
+
+@dataclass(frozen=True, slots=True)
+class GateNeedsHuman(_GateEvent):
+    pass

flowharness_evaluation-0.1.0/src/evaluation/domain/gate_policy.py

@@ -0,0 +1,135 @@
+"""GatePolicy aggregate + the pure gate decision (FLOWHARNESS_SPEC §6.6, §4.2.7; PR9b D9b-1..D9b-4).
+
+A per-org policy document interpreted by a generic engine (§4.2.7): it carries the gate's `scope`
+(AGGREGATE vs PER_CELL) + thresholds, and `decide` turns verified-run evidence into a `Verdict` with
+no I/O — so the whole gate logic is unit-testable here. The safe default (`human_judged=True`)
+forces NEEDS_HUMAN until an org opts into automation (D9b-4); no evidence also defers to a human,
+never a silent PASS/FAIL.
+
+The decision only weighs a run's `pass_rate_bp` + `regressions_count`, declared here as the
+structural `RunEvidence` contract so the domain needn't import the application read model
+(`VerifiedRunRow` satisfies it) — the layering rule (domain imports stdlib + core only) holds.
+"""
+
+from dataclasses import dataclass, field
+from enum import Enum
+from typing import Protocol
+
+from core import BasisPoints, Cell, Event, EventRecorder, Verdict
+from evaluation.domain.errors import EvaluationError
+from evaluation.domain.events import GatePolicySet
+
+
+class RunEvidence(Protocol):
+    """The minimal verified-run shape `decide` weighs — owned by the domain (the PR9a application
+    `VerifiedRunRow` satisfies it structurally, so the domain stays application-free)."""
+
+    # Read-only @property members (not variable annotations): the conforming
+    # VerifiedRunRow is a frozen, slotted dataclass, so its attributes are read-only
+    # — a settable Protocol variable member never matches them (mypy's
+    # settable-vs-read-only rule). Declaring these read-only makes VerifiedRunRow
+    # conform without any runtime change (a Protocol body is never executed).
+    @property
+    def pass_rate_bp(self) -> BasisPoints: ...
+    @property
+    def regressions_count(self) -> int: ...
+
+
+class GateScope(Enum):
+    AGGREGATE = "aggregate"  # judge ALL the candidate's verified runs (any cell)
+    PER_CELL = "per_cell"  # every cell in required_cells must pass
+
+
+def _runs_pass(runs: tuple[RunEvidence, ...], min_bp: BasisPoints, max_reg: int) -> bool:
+    # conservative: every verified run in the slot must individually clear BOTH thresholds
+    return all(
+        r.pass_rate_bp.value >= min_bp.value and r.regressions_count <= max_reg for r in runs
+    )
+
+
+@dataclass(slots=True)
+class GatePolicy(EventRecorder):
+    org_id: str
+    scope: GateScope = GateScope.AGGREGATE
+    required_cells: tuple[Cell, ...] = ()
+    min_pass_rate_bp: BasisPoints = field(default_factory=lambda: BasisPoints(10_000))
+    max_regressions: int = 0
+    human_judged: bool = True  # the default policy forces NEEDS_HUMAN (safe until configured)
+    version: str = "evaluation.gate.v1"
+    row_version: int = 0
+    sequence_no: int = 0
+    _events: list[Event] = field(default_factory=list, init=False)
+
+    def __post_init__(self) -> None:
+        self._validate(self.scope, self.required_cells, self.max_regressions)
+
+    @staticmethod
+    def _validate(scope: GateScope, required_cells: tuple[Cell, ...], max_regressions: int) -> None:
+        if max_regressions < 0:
+            raise EvaluationError("gate_policy_max_regressions_negative")
+        if scope is GateScope.PER_CELL and not required_cells:
+            raise EvaluationError("gate_policy_per_cell_no_cells")
+
+    def configure(
+        self,
+        *,
+        scope: GateScope,
+        required_cells: tuple[Cell, ...],
+        min_pass_rate_bp: BasisPoints,
+        max_regressions: int,
+        human_judged: bool,
+    ) -> None:
+        """Reconfigure the org's policy and record `GatePolicySet`. Validates the new config BEFORE
+        mutating, so a rejected reconfigure leaves the AR unchanged (fail-atomic). Mutation goes
+        through this method (which records the event, bumping `sequence_no`) so a reconfigure is
+        never silently lost — the persistence layer dirty-tracks on the recorded write."""
+        self._validate(scope, required_cells, max_regressions)  # before mutating → fail-atomic
+        self.scope = scope
+        self.required_cells = required_cells
+        self.min_pass_rate_bp = min_pass_rate_bp
+        self.max_regressions = max_regressions
+        self.human_judged = human_judged
+        self._record(
+            GatePolicySet(
+                org_id=self.org_id,
+                scope=scope.value,
+                # full cell identities (not a count): the policy is a REPEATED mutation, so the
+                # journal records which cells each reconfigure required (point-in-time audit)
+                required_cells=tuple((c.harness, c.model) for c in required_cells),
+                min_pass_rate_bp=min_pass_rate_bp.value,
+                max_regressions=max_regressions,
+                human_judged=human_judged,
+                version=self.version,
+            )
+        )
+
+    def decide(self, runs_by_cell: dict[Cell | None, tuple[RunEvidence, ...]]) -> Verdict:
+        """runs_by_cell: for AGGREGATE, one entry keyed None = all the candidate's runs; for
+        PER_CELL, one entry per required Cell (missing/empty = no evidence for that cell)."""
+        if self.human_judged:
+            return Verdict.NEEDS_HUMAN
+        if self.scope is GateScope.AGGREGATE:
+            runs = runs_by_cell.get(None, ())
+            if not runs:
+                return Verdict.NEEDS_HUMAN  # no evidence → defer to a human
+            return (
+                Verdict.PASS
+                if _runs_pass(runs, self.min_pass_rate_bp, self.max_regressions)
+                else Verdict.FAIL
+            )
+        # PER_CELL — only the required cells matter (evidence keyed under any other cell is
+        # ignored, so an unrelated cell can never sink the gate). Precedence is by `required_cells`
+        # order: the first required cell that is missing → NEEDS_HUMAN, else the first that fails →
+        # FAIL. Both branches are conservative (no silent auto-PASS) and deterministic (ordered).
+        for cell in self.required_cells:
+            runs = runs_by_cell.get(cell, ())
+            if not runs:
+                return Verdict.NEEDS_HUMAN  # a required cell has no evidence → defer
+            if not _runs_pass(runs, self.min_pass_rate_bp, self.max_regressions):
+                return Verdict.FAIL
+        return Verdict.PASS
+
+
+def DEFAULT_GATE_POLICY(org_id: str) -> GatePolicy:
+    """The safe baseline used until an org configures one: `human_judged=True` → NEEDS_HUMAN."""
+    return GatePolicy(org_id=org_id)

flowharness_evaluation-0.1.0/src/evaluation/domain/report.py

@@ -0,0 +1,52 @@
+"""EvalReport — the immutable, content-addressed result a runner uploads (FLOWHARNESS_SPEC §6.6).
+
+Float-free by construction: the pass rate is integer `BasisPoints`, the counts are ints — so the
+report's `content_hash` is deterministic + byte-portable (the canonical guard rejects any float),
+which makes the stored evidence re-verifiable. `CaseResult.outcome` is golden-mode `passed|failed`;
+replay's `improved|unchanged|regressed` arrives in p3.
+"""
+
+from dataclasses import dataclass
+from enum import Enum
+
+from core import BasisPoints, ContentHashable
+from evaluation.domain.errors import EvaluationError
+
+
+class Outcome(Enum):
+    PASSED = "passed"
+    FAILED = "failed"
+
+
+@dataclass(frozen=True, slots=True)
+class CaseResult:
+    case_ref: str
+    outcome: Outcome
+    diff_ref: str | None
+
+    def __post_init__(self) -> None:
+        if not self.case_ref:
+            raise EvaluationError("eval_case_ref_blank")
+
+
+@dataclass(frozen=True, slots=True)
+class EvalReport(ContentHashable):
+    per_case: tuple[CaseResult, ...]
+    pass_rate_bp: BasisPoints
+    regressions_count: int
+    new_passing_count: int
+
+    def __post_init__(self) -> None:
+        if self.regressions_count < 0 or self.new_passing_count < 0:
+            raise EvaluationError("eval_report_count_negative")
+
+    def _hash_payload(self) -> object:
+        return {
+            "pass_rate_bp": self.pass_rate_bp.value,
+            "regressions_count": self.regressions_count,
+            "new_passing_count": self.new_passing_count,
+            "per_case": [
+                {"case_ref": c.case_ref, "outcome": c.outcome.value, "diff_ref": c.diff_ref}
+                for c in self.per_case
+            ],
+        }

flowharness_evaluation-0.1.0/src/evaluation/domain/signing.py

@@ -0,0 +1,20 @@
+"""HMAC-SHA256 over the canonical report bytes, keyed by the runner's raw bearer token (D9a-2).
+
+Pure + stdlib only. The platform recomputes the tag with the raw token it already receives to
+authenticate the request, so there is no separate signing-key store; the raw key never leaves the
+edge boundary. Comparison is constant-time. A malformed/empty signature simply fails to verify.
+"""
+
+from core import canonical_json
+
+# sign_report/verify_signature were lifted to core.signing (D3); re-exported here with zero behavior
+# change so every existing evaluation.domain.signing importer keeps working.
+from core.signing import sign_report, verify_signature
+from evaluation.domain.report import EvalReport
+
+__all__ = ["canonical_report_bytes", "sign_report", "verify_signature"]
+
+
+def canonical_report_bytes(report: EvalReport) -> bytes:
+    # the exact bytes the runner signs: the canonical JSON of the report's hash projection
+    return canonical_json(report._hash_payload()).encode("utf-8")

flowharness_evaluation-0.1.0/src/evaluation/domain/suite.py

@@ -0,0 +1,83 @@
+"""EvalSuite aggregate root (FLOWHARNESS_SPEC §6.6).
+
+A suite binds a playbook's curated eval-case assets to its publish thresholds. `Thresholds` is
+float-free (integer `BasisPoints` + ints + a bool). Subclasses `core.EventRecorder`; invariants
+raise `EvaluationError`. Authoring suites may start with no cases; suite mutation (add/retire cases,
+rebaseline) lands when composition screens do.
+"""
+
+from dataclasses import dataclass, field
+
+from core import BasisPoints, Event, EventRecorder
+from evaluation.domain.errors import EvaluationError
+from evaluation.domain.events import EvalSuiteCreated
+
+
+@dataclass(frozen=True, slots=True)
+class Thresholds:
+    min_pass_rate_bp: BasisPoints
+    max_regressions: int
+    vs_baseline: bool
+
+    def __post_init__(self) -> None:
+        if self.min_pass_rate_bp.value > 10_000:
+            raise EvaluationError("basis_points_out_of_range")
+        if self.max_regressions < 0:
+            raise EvaluationError("eval_threshold_negative")
+
+
+@dataclass(slots=True)
+class EvalSuite(EventRecorder):
+    suite_id: str
+    org_id: str
+    playbook_id: str
+    case_asset_ids: tuple[str, ...]
+    thresholds: Thresholds
+    baseline_run: str | None = None
+    name: str | None = None
+    slug: str | None = None
+    description: str | None = None
+    tags: tuple[str, ...] = ()
+    row_version: int = 0
+    sequence_no: int = 0
+    _events: list[Event] = field(default_factory=list, init=False)
+
+    @classmethod
+    def create(
+        cls,
+        *,
+        suite_id: str,
+        org_id: str,
+        playbook_id: str,
+        case_asset_ids: tuple[str, ...],
+        thresholds: Thresholds,
+        baseline_run: str | None = None,
+        name: str | None = None,
+        slug: str | None = None,
+        description: str | None = None,
+        tags: tuple[str, ...] = (),
+    ) -> "EvalSuite":
+        suite = cls(
+            suite_id=suite_id,
+            org_id=org_id,
+            playbook_id=playbook_id,
+            case_asset_ids=case_asset_ids,
+            thresholds=thresholds,
+            baseline_run=baseline_run,
+            name=name,
+            slug=slug,
+            description=description,
+            tags=tags,
+        )
+        suite._record(
+            EvalSuiteCreated(
+                org_id=org_id,
+                suite_id=suite_id,
+                playbook_id=playbook_id,
+                case_count=len(case_asset_ids),
+                min_pass_rate_bp=thresholds.min_pass_rate_bp.value,
+                max_regressions=thresholds.max_regressions,
+                vs_baseline=thresholds.vs_baseline,
+            )
+        )
+        return suite