Repository navigation
Conversation
Add Rust comprehension (Part A)
Generated by .github/workflows/episteme.yml; the knowledge base is derived from the tree, not authored, and is regenerated after merge so branches never carry it and never conflict over it.
… with 8 updates Bumps the production-minor-patch group with 8 updates in the / directory: | Package | From | To | | --- | --- | --- | | [fastapi](https://github.com/fastapi/fastapi) | `0.141.1` | `0.142.2` | | [types-aiobotocore-s3](https://github.com/youtype/mypy_boto3_builder) | `3.9.1` | `3.9.2` | | [temporalio](https://github.com/temporalio/sdk-python) | `1.33.0` | `1.34.0` | | [sqlglot](https://github.com/tobymao/sqlglot) | `30.18.0` | `30.21.0` | | [opentelemetry-sdk](https://github.com/open-telemetry/opentelemetry-python) | `1.44.0` | `1.45.0` | | [opentelemetry-exporter-otlp-proto-http](https://github.com/open-telemetry/opentelemetry-python) | `1.44.0` | `1.45.0` | | [ruff](https://github.com/astral-sh/ruff) | `0.16.8` | `0.16.10` | | [mypy](https://github.com/python/mypy) | `2.3.1` | `2.4.0` | Updates `fastapi` from 0.141.1 to 0.142.2 - [Release notes](https://github.com/fastapi/fastapi/releases) - [Commits](fastapi/fastapi@0.141.1...0.142.2) Updates `types-aiobotocore-s3` from 3.9.1 to 3.9.2 - [Release notes](https://github.com/youtype/mypy_boto3_builder/releases) - [Commits](https://github.com/youtype/mypy_boto3_builder/commits) Updates `temporalio` from 1.33.0 to 1.34.0 - [Release notes](https://github.com/temporalio/sdk-python/releases) - [Changelog](https://github.com/temporalio/sdk-python/blob/main/CHANGELOG.md) - [Commits](temporalio/sdk-python@1.33.0...1.34.0) Updates `sqlglot` from 30.18.0 to 30.21.0 - [Commits](tobymao/sqlglot@v30.18.0...v30.21.0) Updates `opentelemetry-sdk` from 1.44.0 to 1.45.0 - [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases) - [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md) - [Commits](open-telemetry/opentelemetry-python@v1.44.0...v1.45.0) Updates `opentelemetry-exporter-otlp-proto-http` from 1.44.0 to 1.45.0 - [Release notes](https://github.com/open-telemetry/opentelemetry-python/releases) - [Changelog](https://github.com/open-telemetry/opentelemetry-python/blob/main/CHANGELOG.md) - [Commits](open-telemetry/opentelemetry-python@v1.44.0...v1.45.0) Updates `ruff` from 0.16.8 to 0.16.10 - [Release notes](https://github.com/astral-sh/ruff/releases) - [Changelog](https://github.com/astral-sh/ruff/blob/main/CHANGELOG.md) - [Commits](astral-sh/ruff@0.16.8...0.16.10) Updates `mypy` from 2.3.1 to 2.4.0 - [Changelog](https://github.com/python/mypy/blob/master/CHANGELOG.md) - [Commits](python/mypy@v2.3.1...v2.4.0) --- updated-dependencies: - dependency-name: fastapi dependency-version: 0.142.2 dependency-type: direct:production update-type: version-update:semver-minor dependency-group: production-minor-patch - dependency-name: types-aiobotocore-s3 dependency-version: 3.9.2 dependency-type: direct:production update-type: version-update:semver-patch dependency-group: production-minor-patch - dependency-name: temporalio dependency-version: 1.34.0 dependency-type: direct:production update-type: version-update:semver-minor dependency-group: production-minor-patch - dependency-name: sqlglot dependency-version: 30.21.0 dependency-type: direct:production update-type: version-update:semver-minor dependency-group: production-minor-patch - dependency-name: opentelemetry-sdk dependency-version: 1.45.0 dependency-type: direct:production update-type: version-update:semver-minor dependency-group: production-minor-patch - dependency-name: opentelemetry-exporter-otlp-proto-http dependency-version: 1.45.0 dependency-type: direct:production update-type: version-update:semver-minor dependency-group: production-minor-patch - dependency-name: ruff dependency-version: 0.16.10 dependency-type: direct:production update-type: version-update:semver-patch dependency-group: production-minor-patch - dependency-name: mypy dependency-version: 2.4.0 dependency-type: direct:production update-type: version-update:semver-minor dependency-group: production-minor-patch ... Signed-off-by: dependabot[bot] <support@github.com>
…on-minor-patch-4e16c1bde2 chore(deps): bump the production-minor-patch group across 1 directory with 8 updates
Add Cargo-aware Rust codegen and validation gates
Generated by .github/workflows/episteme.yml; the knowledge base is derived from the tree, not authored, and is regenerated after merge so branches never carry it and never conflict over it.
Generated by .github/workflows/episteme.yml; the knowledge base is derived from the tree, not authored, and is regenerated after merge so branches never carry it and never conflict over it.
fix(rust): use maintained Orchard grammar for pinned source
Generated by .github/workflows/episteme.yml; the knowledge base is derived from the tree, not authored, and is regenerated after merge so branches never carry it and never conflict over it.
docs(rust): sign off core support after upstream lint merge
Generated by .github/workflows/episteme.yml; the knowledge base is derived from the tree, not authored, and is regenerated after merge so branches never carry it and never conflict over it.
…/preflight SSPN-118. A project opts in by committing .spine/required-behavior.yaml: checks that must pass before a run can report itself verified, distinct from PreflightRunner (lint/type quality) and TestRunner (the model's own tests). ONTM-4 showed generated tests can pass while the feature under test never exercises its own default wiring; this closes that gap with checks the model does not author and cannot edit away. - contracts.py: RequiredBehaviorRunner Protocol, RequiredBehaviorResult, RequirementResult (carries environment_blocked, determined structurally from the exit code rather than text-sniffing output). - required_behavior.py: SubprocessRequiredBehaviorRunner (self-skips without a manifest, mirroring SubprocessPreflightRunner's own "no pyproject.toml" behavior) and StubRequiredBehaviorRunner. load_manifest() fails loud on a malformed manifest rather than treating it as "no manifest" -- an unusable opt-in should not read as unverified. - deps.py / worker.py: new SDLCDeps field, wired alongside preflight. Found and fixed in passing: exec_capture() (process.py) raises FileNotFoundError instead of returning a code when argv[0] doesn't exist -- every other caller's argv[0] is a Spine-controlled interpreter that's always present, so this never surfaced before. A manifest's command is project-authored and can simply be wrong; handled locally in required_behavior.py rather than changing the shared helper. Hidden-check loading (SSPN-118/D16) and the loop integration (SSPN-119) are separate tickets; this lands the Protocol/types/self-skip behavior on its own, consumable by both. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
… loop SSPN-119. required_behavior joins tests/types/coverage as its own independently-budgeted check in run_feature's implement -> author_tests -> refine loop, run after the coverage probe clears (the last, most specific gate before a run reports itself passed). Own spent/budgets entry, same reasoning the loop already applies to tests/types/coverage: a shared pool starves whichever check runs last. Defaults to max_refine's value. An environment-blocked requirement (missing fixture/command, not a code defect) stops the loop immediately rather than spending refine budget on something refine cannot fix by editing code -- same policy `tests` already has via environment_blocker(), applied structurally here via RequiredBehaviorResult.items[].environment_blocked instead of the text-pattern heuristic that function uses (SSPN-118/D21). Explicitly out of scope, per the ticket's own D22: _repair_after_revision and review_response.py's review-feedback loop. Both already drive refine() independently and are tracked as a named follow-up, not wired here. New tests (4): pass completes the run unchanged, fail-then-pass drives exactly one refine call, failure alone exhausts its own budget while tests/types stay green the whole time, environment-blocked stops with zero refine calls. All 118 pre-existing test_feature_runner.py tests stay green unmodified -- none construct a .spine/required-behavior.yaml, so the new kind is an inert no-op for every one of them, confirming SSPN-118's self-skip design holds end to end. Caught by the new tests, not by inspection: the first version of this wired required_behavior's "needs refine" branch so it fell through into the sibling coverage-budget block (same indentation, previously reachable only when `gaps` was non-empty) -- silently corrupting the coverage counter and calling author_tests with an empty gap list, which broke the loop on the very first required_behavior failure. Fixed with an explicit else so the two branches can no longer collide. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
…ctly SSPN-120. PKGCodegenGrounder.context_for_required_entry_points() resolves a required-behavior manifest's entry point against the PKG and surfaces OTHER implementations of the same base type via the existing implements_of()/implementors_of() primitives -- no new PKG fact type, no extractor change (checked: the closest existing concept, the PROVIDES edge, is Kotlin/C#-only; extending it for Python is its own track, not this ticket). Why this and not more context_for_spec: for an ONTM-4-shaped bug (a missing call, not a wrong one), showing the model the buggy symbol's own source proves nothing. A sibling that wires correctly -- a different rule type's loader that does call its compiler -- is what actually helps. Wired into LLMCodegenAdapter._grounding() additively: an empty lexical match from context_for_spec must not suppress a sibling finding, and vice versa (verified by test_grounding_is_additive_not_gated_on_lexical_ match, which deliberately uses a spec unrelated to the fixture's domain). Guarded with getattr(), the same defensive pattern context_for_symbols already uses, since not every CodegenGrounder test double implements it. Extends SSPN-118's manifest schema with an optional `entry_point` field (dotted path or short name) -- not present in the original ticket's Requirement dataclass, needed for this to have anything to resolve against. Backward compatible: existing manifests without it are unaffected. Tests: 6 new in test_grounding.py against a real on-disk fixture (common base + a buggy implementation + a correctly-wired sibling) proving the sibling surfaces and the target/base do not; 3 new in test_required_behavior.py for the entry_point field; 2 end-to-end against an ONTM-4-shaped fixture (default loader skips its compiler) proving the real (non-faked) runner fails on the broken version and passes on the fix, per this ticket's own exit criterion. Gate: ruff/mypy src tests (824 files) clean. tests/sdlc/ -- 1610 passed, 2 skipped (unrelated missing extras), 0 regressions. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
SSPN-121 (ai/spine half -- the spineharness docs/packaging half ships as its own PR per SSPN-122/D30). Two gaps the roadmap's own wording anticipated but the code didn't yet support. Telemetry (D27): StageUsage/TokenLedger only ever got an entry via record(stage, result: CompletionResult), called exclusively from real LLM completions. required_behavior (SSPN-119) makes zero LLM calls -- pure subprocess execution -- so it never appeared in the ledger, despite " deterministic work may have zero model tokens while still having runtime; label it accordingly" being the roadmap's own stated requirement. record_deterministic() + StageUsage.deterministic_seconds close the gap; wired into run_feature's loop around the required_behavior call. The worklog table gains a Deterministic (s) column so a zero-token row with real time reads as "ran real work", not "didn't run". Three-outcome reporting (D28): nothing in types.py distinguished more than a binary passed/failed verdict -- built from scratch, using data the loop already computes (spent). _attempt_outcome() labels a successful run as either "initial generated candidate passed with no correction needed" or "first completed autonomous workflow after N refine(s)", threaded into render_worklog's new Outcome line. The third outcome, a later human-assisted continuation, happens in a separate call (review_response.py) that review_response.py's own loop never re-checks required_behavior for (SSPN-119/D22's deliberate scope cut) -- ReviewResponse.required_behavior_note documents that explicitly on every instance, a standing fact rather than a blank field to source-dive for. Tests: 3 new in test_recording.py for the deterministic path, 2 new in test_feature_runner.py proving both outcome strings render in a real posted worklog, 1 new in test_review_response.py for the documented note. All passed on first run. Gate: ruff/mypy src tests (824 files) clean, full-repo ruff check clean. tests/sdlc/ + tests/core/llm/ -- 1656 passed, 2 skipped (unrelated missing extras), 0 regressions. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
… gates SSPN-122 gate-closing work. CHANGELOG.md gets one combined Unreleased entry covering SSPN-118-121's cumulative addition (one entry, not four, per this plan's "one MR" delivery model). STATE-OF-SPINE.md's source modules/test functions/test files counts refreshed to match the source (404/4659/392) -- state-numbers.py --check was failing on three stale claims, mostly from upstream's Rust codegen landing since this branch was cut, partly from this epic's own new module and tests. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
…output type SSPN-121 added ReviewResponse.required_behavior_note, and sdlc_address_review's handler spreads **result.__dict__ into its return dict -- the field was always going to flow through the MCP tool unannounced. Caught by the real cost of running the full test suite (not a scoped subset): tests/plugin/test_server.py's output-contract check failed on the undeclared key. Declared it on SdlcAddressReviewOut instead of filtering the field out of the response -- it's useful information for an MCP caller too, not something to hide. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
load_manifest() returns tuple[Requirement, ...] -- its length isn't statically known to be 1, so (requirement,) = load_manifest(tmp_path) is a real static-analysis risk (CodeQL: "Left hand side of assignment contains 1 variables, but right hand side is a tuple of length 0"), even though these two tests' own fixtures guarantee exactly one requirement at runtime. Indexes with an explicit length assertion instead -- same safety, no destructuring CodeQL can flag. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
PR #514's dependency-audit job failed on 3 new advisories against versions already locked on develop (confirmed: develop's own most recent audit run passed against these identical versions minutes earlier -- the CVEs were published between that run and this PR's, not introduced by this branch). All three are transitive (none pinned in pyproject.toml): multidict via aiohttp/aiobotocore/yarl, fsspec via huggingface-hub/torch (the asr extra), langgraph-sdk via langgraph. uv lock --upgrade-package resolved all three within existing constraints -- no pyproject.toml change needed: multidict 6.7.1 -> 6.9.1 (CVE-2026-104874's exact fix) fsspec 2026.4.0 -> 2026.9.0 langgraph-sdk 0.4.2 -> 0.4.5 Verified before committing: re-ran the exact `pip-audit` command the CI job uses -- "No known vulnerabilities found, 1 ignored" (the pre-existing, already-justified click/PYSEC-2026-2132 entry, unchanged). uv.lock diff touches only these 3 packages' versions, nothing else moved. Full repo ruff/mypy clean. Full test suite re-run after re-syncing: 5519 passed, 13 skipped (unrelated missing-tooling skips), 0 failed -- the fsspec/torch/huggingface-hub chain is deep enough that this needed real verification, not just a clean resolver run. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
feat(sdlc): required-behavior gates (SSPN-116)
Generated by .github/workflows/episteme.yml; the knowledge base is derived from the tree, not authored, and is regenerated after merge so branches never carry it and never conflict over it.
Minor: required-behavior gates add a fourth, independently-budgeted check to sdlc feature's native refine loop, re-verified after every refinement (SSPN-116) — a project opts in via .spine/required-behavior.yaml and a legacy project without one takes the exact code path it took before. Core Rust support (comprehension + Cargo-aware codegen/build/test) is signed off for supported host environments, closing the installed-parser grammar gate via a pinned tree-sitter-rust-orchard and the Synaptreesitter lint-baseline gate via its own upstream PR. Also fixes two stale release-cut gaps found while cutting this one: the Rust comprehension changelog entry was misfiled under the already-shipped 3.53.0 section instead of Unreleased, and capability-matrix.md's header plus gap4-adoption-distribution-roadmap.md's workflow tag were still stamped 3.52.0 from two releases ago. Stamps pyproject + the lockfile's own entry, the three plugin manifests, the reusable-workflow tags (including the one in gap4-adoption-distribution- roadmap.md), CHANGELOG (Unreleased -> 3.54.0), README "What's new", CLI_REFERENCE, STATE-OF-SPINE (including its stale Rust-gate-open prose), capability-matrix, SPEC-INDEX's recount line, and the re-rendered architecture + knowledge-foundation SVG/PNG. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Generated by .github/workflows/episteme.yml; the knowledge base is derived from the tree, not authored, and is regenerated after merge so branches never carry it and never conflict over it.
![github-advanced-security[bot]](https://avatars.githubusercontent.com/in/57789?s=60&v=4){kind=small}
mbinder45
approved these changes
Oct 6, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Promotes
developtomainfor the 3.54.0 release..spine/required-behavior.yaml— checks the model does not author and cannot edit away, re-verified after every refinement as a fourth independently-budgeted check insdlc feature's native loop. Closes the ONTM-4 gap where a generated test could pass without exercising a feature's real default wiring. A legacy/token-only project without a manifest takes the exactrefine()path it took before — additive, not a branch. Merged via PR #514.capability-matrix.md's header plusgap4-adoption-distribution-roadmap.md's workflow tag were still stamped 3.52.0 from two releases ago.Test plan
ruff format --check .,mypy src tests(828 files), fullpytestsuite (5519 passed, 13 skipped, 51 deselected)--checkgates green:mcp-tools.py,matrix-count.py,state-numbers.py, both SVG renderersorchestrator pkg accuracy --check— 0 gated regressionsorchestrator pkg verify— 0 errorsscripts/sdlc_shapes.py— all 4 repository shapes holdorchestrator understand .— episteme builds clean (not committed; regenerates post-merge per.github/workflows/episteme.yml)developpush (ea8579a9→check,Dependency audit,CodeQL,Episteme)security scancheck on this promotion PR🤖 Generated with Claude Code